hash stringlengths 32 32 | doc_id stringlengths 5 12 | section stringlengths 5 1.47k | content stringlengths 0 6.67M |
|---|---|---|---|
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.12.3 Service Flows | 1. The UAV A takes a real-time picture with its 4-way 4K full-angle camera;
2. The picture is transmitted to the forest fire monitoring centre via the 5G network with satellite access network. This would require high data rate (e.g., 120Mbit/s) in UL direction.
3. The forest fire monitoring centre uses the AI system to determine whether there is a fire, according to the received picture. In case of fire, the forest fire monitoring centre will request the position of the UAV.
4. After receiving the positioning service request, the 5G network detects an error that the negotiated location delivery latency cannot be guaranteed. Then, it provides a lower position accuracy to ensure latency.
5. If the forest fire monitoring centre decides to adjust the route of UAV A, it will send adjustment commands to the UAV via the 5G network with satellite access network, which requires low delay and high reliability in DL direction. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.12.4 Post-conditions | UAV adjusts its route. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.12.5 Existing features partly or fully covering use case functionality | There are a few position requirements specified in 3GPP TS 22.125 [15], which have been described as:
[R-5.1-009] The 3GPP system should enable an MNO to augment the data sent to a UTM with the following: network-based positioning information of UAV and UAV controller.
NOTE 1: This augmentation may be trust-based (i.e. the MNO informs the UTM that the UAV position information is trusted) or it may be additional location information based on network information, such as OTDOA, cell coordinates, synchronization source, etc.
NOTE 2: This requirement will not be applied to the case which the UAS and UTM has direct control communication connection without going through MNO, such as OTDOA, cell coordinates, synchronization source, etc.
[R-5.1-012] The 3GPP system shall enable a UAS to update a UTM with the live location information of a UAV and its UAV controller.
[R-5.1-013] The 3GPP network should be able to provide supplement location information of UAV and its controller to a UTM.
NOTE 3: This supplement may be trust-based (i.e. the MNO informs the UTM that the UAV position information is trusted) or it may be additional location information based on network information.
There are also a few position requirements specified in 3GPP TS 22.261 [2], which have been described as:
The 5G system shall support mechanisms to determine the UE’s position-related data for period when the UE is outside the coverage of 3GPP RAT-dependent positioning technologies but within the 5G positioning service area (e.g. within the coverage of satellite access).
In 3GPP TS 22.071 [x], the following location service requirements are captured:
The precision of the location shall be network design dependent, i.e., should be an operator’s choice. This precision requirement may vary from one part of a network to another.
About horizontal accuracy:
The LCS service shall provide techniques that allow operators to deploy networks that can provide at least the level of accuracy required by the regional regulatory bodies.
10m-50m: Asset Location, route guidance, navigation
About vertical accuracy:
For Value Added Services, and PLMN Operator Services, the following is applicable:
- When providing a Location Estimate, the LCS Server may provide the vertical location of a UE in terms of either absolute height/depth or relative height/depth to local ground level. The LCS Server shall allow an LCS Client to specify or negotiate the required vertical accuracy. The LCS Server shall normally attempt to satisfy or approach as closely as possible the requested or negotiated accuracy when other quality of service parameters are not in conflict.
- The vertical accuracy may range from about three metres (e.g. to resolve within 1 floor of a building) to hundreds of metres.
About location delivery latency:
Location Delivery Latency (Time to First Fix) is set at a maximum of 30 seconds from the time the user initiates an emergency service call to the time it is available at the location information center. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.12.6 Potential New Requirements needed to support the use case | [PR 5.12.6-001] The 5G system with satellite access shall be able to support suitable positioning mechanisms for UAV services.
[PR 5.12.6-002] The 5G system with satellite access shall be able to support positioning services and to provide information to a UE on delivered performance of positioning services. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.13 Use case on Enhanced Positioning Service using Satellite Access | |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.13.1 Description | During natural disasters, the recovery of communication services and the acquisition of the survivors’ location are important to aid the rescue activities. Normally, satellite communication networks and standalone GNSS are utilized to serve communication and positioning independently. With the integration of satellite access in 5G systems, it’s possible to provide both communication and positioning services by 5G system together to address the cases that GNSS cannot provide reliable positioning service (e.g. poor GNSS signal, limited visible satellites). Meanwhile, the positioning performance like accuracy can be improved with the assistance of 5G satellites (e.g. LEO), network information, etc. [16]
Indonesia is famous for its extraordinary natural landscapes attracting millions of tourists around the world. Meanwhile, it is widely recognized as one of the most disaster-prone countries in the world according to data released by the United Nations International Disaster Reduction Agency (UN-ISDR) [17]. A disaster (e.g. earthquake, tsunamis) will impact tens of millions of people, who may get support from the sustainable and reliable communication service and positioning service of the 5G system. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.13.2 Pre-conditions | Bali, Indonesia is covered by terrestrial access network of Operator TerrA and satellite access network of Operator SatA, which shares the core network of TerrA deployed in Jakarta with mutual agreement. 5G communication service and positioning service have been launched all through Indonesia.
GNSS (e.g. GPS, BeiDou) are allowed to use in Indonesia, but are independent from 5G satellite constellation.
It is assumed that UEs with the subscription of TerrA network are capable of 5G satellite access. Some of them are incapable of GNSS receivers and some are integrated with different types of GNSS receivers. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.13.3 Service Flows | A tsunami strikes Bali and has destroyed most infrastructures and terrestrial access networks. The core network is still in service without damage.
The awake survivors will initiate an emergency call or send an emergency message to report personal information, injuries, and surrounding conditions to Indonesia Rescue Center for rescue requests through SatA access network. During the interaction, the precise location information (e.g. accurate latitude) of the survivors is requested to report to Rescue Center from UE or/and the network with the help of 3GPP positioning methods or non-3GPP positioning methods (e.g. GNSS) within the requested response time of local regulatory requirements. The location of the survivors and rescue personnel will be sent to the survivors as well for preparation.
All powered-on UEs will autonomously update registration in TerrA’s network using satellite access. The network identifies the areas where the devices are located and authorizes Rescue Center to fetch real-time devices’ location and tracing log during the Rescue UAV or Helicopter searching for the survivors in a coma or not able to report emergency information actively. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.13.4 Post-conditions | The locations of the awake survivors are identified as compliant with regulatory requirements.
The location information of powered-on devices is available in Rescue vehicles. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.13.5 Existing features partly or fully covering use case functionality | Regarding TS 22.261[2] as below, UE using satellite access shall have the capability to offer the location, and 5G system needs to determine the location for service, but not consider the situation that the location cannot be decided by UE alone.
A UE supporting satellite access shall be able to provide or assist in providing its location to the 5G network.
A 5G system with satellite access shall be able to determine a UE's location in order to provide service (e.g. route traffic, support emergency calls) in accordance with the governing national or regional regulatory requirements applicable to that UE.
The legacy requirements for positioning service defined in TS 22.261 [2] are not well adapted to all types of UE with satellite access considering the satellite characteristics (e.g. latency).
The 5G system shall provide 5G positioning services in compliance with regulatory requirements.
The 5G network shall be able to request the UE to provide its position-related-data on request—together with the accuracy of its position—triggered by an event or periodically and to request the UE to stop providing its position-related data periodically.
The 5G System with satellite access shall be able to negotiate the positioning methods according to the operator's policy or the application’s requirements or the user's preferences and shall support mechanisms to allow the network or the UE to trigger this negotiation. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.13.6 Potential New Requirements needed to support the use case | [PR 5.13.6-001] Subject to regulatory requirements and operator’s policy, the 5G system with satellite access shall be able to support 3GPP positioning methods for UEs using only satellite access.
[PR 5.13.6-002] The 5G system with satellite access shall be able to negotiate the positioning methods according to 3GPP RAT and UE positioning capability, the availability of non-3GPP positioning technologies (e.g. GNSS) and shall support mechanisms to allow the network or UE to trigger the negotiation.
[PR 5.13.6-003] Subject to regulatory requirements, the 5G system with satellite access shall be able to provide positioning services (e.g. with the availability of 99%, the accuracy of several kilometers) independently of UE’s GNSS capability when the UE is using only satellite access.
NOTE: The regulatory requirements for positioning (e.g. service requirements of Public Safety by GSMA [22]) could be taken into account. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.14 Use case on service continuity for UE-to-UE communication between satellites | |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.14.1 Description | The provision of Internet services using mega-constellations of LEO (Low Earth Orbit) satellites is a promising solution on the path to the future mobile communication systems. LEO is the Earth-centered orbit with an altitude in the range of 350km and 2000km above sea level. The LEO satellites at 600km altitude travel at a speed of about 7.8km/sec [18]. Due to the fast movement of LEO satellite, the service duration of a satellite for the coverage with 1000km diameter is less than 3 minutes. Therefore, guaranteeing robust service continuity and satisfactory user experience is the most critical issue in LEO satellite system.
In some countries, the state government operates Aviation Branches for Forest Protection Service to fight forest fires and assist in search and rescue missions [19]. The helicopters are part of the Aviation Branch and used for fire detection and firefighting, dropping water, and moving firefighters and equipment to rural and remote locations. In these locations, there may be no terrestrial network, so the helicopters and firefighters can collaborate with each other by communicating in the help of satellite. Moreover, since it takes several hours or days to complete their missions, it should be considered how to ensure the continuity of communication service using satellite access across multiple NGSO satellites.
Furthermore, the draft report of ITU-R for IMT2020-satellite requirement for mobility interruption time in satellite radio interfaces is 50ms [20].
Figure 5.14.1-1: Example of service continuity for UE-to-UE communication between satellites without going through the ground network |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.14.2 Pre-conditions | Satellite operator Sat-OP has deployed NGSO satellites and has an agreement with Terrestrial Operator Ter-OP to provide communication services for UEs under satellite coverage.
Firefighter A and B have signed contract with Sat-OP for communication services using satellite access. Thus, their devices can communicate with each other directly via satellite without going through the ground network.
Firefighter A and B move to the rural or remote area in which there is no terrestrial network, but the satellites operated by Sat-OP can provide communicate services. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.14.3 Service Flows | 1. Firefighter A and B make a phone call or exchange some data (e.g. pictures, video streams) during their work. Then, their data traffic is routed through satellite Sat-1.
2. During the communication service, if Firefighter B is located in the coverage of satellite Sat-2, Firefighter B has a connection to satellite Sat-2 and the communication between Firefighter A and B is provided by satellite Sat-1 and Sat-2 through inter satellite link.
3. After some time, if satellite Sat-2 serves the area in which Firefighter A and B are located, the satellite Sat-2 takes over the data sessions for Firefighter A and B from satellite Sat-1, and then the data traffic is routed through satellite Sat-2. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.14.4 Post-conditions | Firefighter A and B can finish the phone call or data exchange without any discontinuation of communication service with the support of multiple NGSO satellites. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.14.5 Existing features partly or fully covering the use case functionality | Regarding TS 22.261 [2], satellite access and satellite connectivity are supported in Rel-18, as
The 5G system shall be able to provide services using satellite access.
The 5G core network shall support collection of charging information based on the access type (e.g. 3GPP, non-3GPP, satellite access).
For a 5G system with satellite access, the following requirements apply:
- A 5G system with satellite access shall enable roaming of UE supporting both satellite access and terrestrial access between 5G satellite networks and 5G terrestrial networks. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.14.6 Potential New Requirements needed to support the use case | [PR 5.14.6-001] Subject to regulatory requirements and operator’s policy, the 5G system with satellite access shall be able to support the establishment of a communication path between UEs via one or multiple serving satellites without going through the ground network.
[PR 5.14.6-002] Subject to regulatory requirements and operator’s policy, the 5G system with satellite access shall be able to support service continuity of a communication between UEs without going through the ground network when the UE communication path moves between serving satellites.
[PR 5.14.6-003] Subject to regulatory requirements and operator’s policy, the 5G system with satellite access shall support service continuity of a communication between UEs without going through the ground network when the communication path between UEs via one or multiple serving satellites extends across several satellites (through inter satellite links). |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.15 Use case on service continuity for UE-to-UE communication in case of mobility between satellite and terrestrial network | |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.15.1 Description | UAM (Urban Air Mobility) refers to a safe and efficient air transport system. UAM is used for transporting passengers or cargo in urban or suburban areas. Recently, in some countries, telecom operators have already started the collaboration with aviation companies for UAM business building from airframes to service platforms [21].
In order to control and manage the UAM body for safe and sound travel, the UAM vehicles should receive various information about the movement of other flying vehicles, climate conditions, location, and so on. Additionally, the UAM vehicle can provide the in-flight Internet service allowing the passengers to communicate with the users in the remote networks and on other flying vehicles as well.
Even though the UAM vehicles generally operate at an altitude less than 1km, they may fly over the air out of terrestrial network coverage. Thus, the commercialization of UAM depends on the establishment of a telecommunication network service including LEO satellite communications. While flying out of terrestrial network coverage, the vehicles can communicate with each other via satellite without going through the ground network. But, as a vehicle approaches the ground and hence has a connection to the terrestrial network, the communication between vehicles via satellite should be continuously provided through the satellite and terrestrial network.
Figure 5.15.1-1: Service continuity for UE-to-UE communication in case of mobility between satellite and terrestrial network without going through the ground network |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.15.2 Pre-conditions | UAM company UAM-Co operates many UAM vehicles in urban and suburban areas.
UAM company UAM-Co contracts with Terrestrial Operator Ter-OP to provide communication services for the devices on UAM vehicles.
UAM company UAM-Co also have signed contract with Sat-OP for communication services via satellite access. Their devices on UAM vehicles can communicate with each other directly via satellite without going through the ground network.
Satellite Operator Sat-OP has an agreement with Terrestrial Operator Ter-OP to provide communication services for UEs under satellite coverage.
The devices on UAM vehicle A and B have a subscription with the Terrestrial Operator Ter-OP. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.15.3 Service Flows | 1. The devices on UAM vehicle A and B register with the Ter-OP network.
2. UAM vehicle A is flying out of Ter-OP network coverage, thus its device has a connection to the satellite operated by Sat-OP.
3. UAM vehicle B is ready to fly in the ground station, and hence its device has a connection to the Ter-OP network.
4. Before or Just after taking off, UAM vehicle B needs to gather the information on the movement of other flying vehicles including vehicle A. The data traffic between UAM vehicle A and B is routed though the satellite and terrestrial networks.
5. UAM vehicle B keeps gathering the movement information of vehicle A even after it moves out of Ter-OP network coverage. Since the information exchange between UAM vehicles should be performed in real time (with very low latency), the vehicles communicate with each other via satellite directly without going through the ground network.
6. After then, as UAM vehicle B approaches the ground, it has a connection to the terrestrial network.
7. The traffic between UAM vehicle A and B is going through the satellite and terrestrial network.
8. As a result, the communication between UAM vehicle A and B keep going without any discontinuation of service. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.15.4 Post-conditions | User A and B can finish the exchange of their movement information without any discontinuation of communication service regardless of their roaming between satellite and terrestrial network. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.15.5 Existing features partly or fully covering the use case functionality | 3GPP TS 22.261 [2],
clause 6.2.4 includes roaming related requirements in diverse mobility management:
For a 5G system with satellite access, the following requirements apply:
- A 5G system with satellite access shall enable roaming of UE supporting both satellite access and terrestrial access between 5G satellite networks and 5G terrestrial networks.
clause 6.3.2.3 on satellite access includes the following requirement:
The 5G system shall be able to provide services using satellite access.
clause 9.1 on charging aspect includes the following requirement:
The 5G core network shall support collection of charging information based on the access type (e.g. 3GPP, non-3GPP, satellite access). |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.15.6 Potential New Requirements needed to support the use case | [PR 5.15.6-001] Subject to regulatory requirements and operator’s policy, the 5G system with satellite access shall support service continuity, when the UE communication path moves between 5G terrestrial access network and 5G satellite access network owned by the same operator or owned by different operators having an agreement. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.16 Use case on store and forward – emergency report | |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.16.1 Description | This use case illustrates the realization of a S&F service between a UE with satellite access and an Application Server for an emergency reporting service.
A description of store and forward operation is provided in Annex A.
Bob was sailing on an intercontinental containership, which sank for some exotic reason. Bob is now shipwrecked on a remote island and while he is not in immediate danger, he needs rescue within a matter of days as food and water is scarce.
A few items from the containership washed ashore with Bob, one of which is an IoT device from Company TrackingInc with a subscription to IoTSAT for the 5G IoT connectivity by satellite and IoTSAT is using a LEO constellation which supports S&F operation mode.
The IoT device allows Bob to send an emergency report including his position via the S&F network. A confirmation is received by the IoT device that the emergency report “went through” as soon as possible. As the indicator light by the emergency button of the IoT device starts blinking green, Bob knows that it is a matter of time before Alice rescues him. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.16.2 Pre-conditions | In the present use case, the emergency reporting UE is in a remote area with no ground stations available for feeder link connectivity and the emergency reporting UE is aware that IoTSAT constellation operates in S&F mode. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.16.3 Service Flows | 1. Bob is sailing on an intercontinental containership, which sinks.
2. Bob is ashore and finds an IoT device from Company TrackingInc with a subscription to IoTSAT for the 5G IoT connectivity by satellite.
3. Bob sends an emergency report including his position with the IoT device from Company TrackingInc through IoTSAT.
4. The emergency report from Bob is received by the fly-by satellite of the IoTSAT constellation and is stored in the satellite waiting to be delivered as there is no feeder link available in the area where Bob is ashore.
5. The satellite of the IoTSAT constellation is able to deliver the “emergency report” from Bob in a matter of seconds as soon as a first feeder link is available as it identified the service as emergency and there is no restriction to use any feeder link and ground station for such service.
6. Bob is informed that the emergency report has been delivered upon the next fly-by of a satellite from the IoTSAT constellation. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.16.4 Post-conditions | The emergency report generated by the IoT UE has been delivered successfully to the TrackingInc application server and forwarded to a service able to treat the report and a response has been forwarded to the IoT UE without relying on a continuous end-to-end network connectivity path between them. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.16.5 Existing features partly or fully covering the use case functionality | 3GPP TS 22.261 [2], clause 6.3.2.3 on satellite access includes the following requirements:
The 5G system shall be able to provide services using satellite access.
The 5G system with satellite access shall be able to support low power MIoT type of communications.
However, it is not sufficient in regards to S&F operation especially for the delivery of delay-tolerant/non-real-time IoT NTN services with NGSO satellites and considering here the case of emergency delivering in area where there is only a LEO constellation covering the device and using store and forward operation. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 5.16.6 Potential New Requirements needed to support the use case | [PR.5.16.6-001] The 5G system with satellite access, and when the satellite access is operating in store and forward mode, shall be able to inform an authorized UE about how long the data is expected to be stored before being delivered.
[PR.5.16.6-002] Subject to regulatory requirements and operator’s policy, a 5G system with satellite access supporting S&F Satellite operation shall be able to forward an emergency report as soon as there is a feeder link available and shall be able to notify as soon as possible the UE about the successful forwarding.
NOTE 1: Subject to regulation, the emergency report could have priority over other communication |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 6 Consolidated requirements | |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 6.1 Introduction | The following requirements represent a consolidation of the various potential requirements captured in the above use cases related to a 5G system with satellite access. |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 6.2 Store & Forward Satellite operation | The potential requirements corresponding to the support of S&F Satellite operation are listed in the table below.
Table 6.2-1 – Consolidated Requirements for S&F Satellite operation
CPR #
Consolidated Potential Requirement
Original PR #
Comment
CPR 6.2-1
Subject to operator’s policies, a 5G system with satellite access shall be able to support S&F Satellite operation for authorized UEs e.g. store data on the satellite when the feeder link is unavailable; and forward the data once the feeder link between the satellite and the ground segment becomes available.
[PR 5.1.6-001]
[PR.5.3.6-001]
[PR 5.4.6-001]
[PR.5.16.6-002]
CPR 6.2-2
A 5G system with satellite access shall be able to inform a UE whether S&F Satellite operation is applied.
[PR 5.1.6-002]
[PR.5.16.6-001]
CPR 6.2-3
Subject to operator’s policies, a 5G system with satellite access supporting S&F Satellite operation shall be able to allow the operator or a trusted 3rd party to apply, on a per UE and/or satellite basis, an S&F data retention period.
[PR 5.1.6-003]
[PR 5.2.6-002]
CPR 6.2-4
Subject to operator’s policies, a 5G system with satellite access supporting S&F Satellite operation shall be able to allow the operator or a trusted 3rd party to apply, on a per UE and/or satellite basis, an S&F data storage quota.
[PR 5.1.6-004]
[PR 5.2.6-003]
[PR.5.3.6-003]
[PR 5.4.6-004]
CPR 6.2-5
A 5G system with satellite access supporting S&F Satellite operation shall be able to support a mechanism to configure and provision specific required QoS and policies for UE’s data subject to store and forward operation (e.g. forwarding priority, acknowledgment policy).
[PR 5.1.6-005]
[PR 5.2.6-004]
CPR 6.2-6
A 5G system with satellite access supporting S&F Satellite operation shall be able to provide related information (e.g. estimated delivery time to the authorised 3rd party) to an authorized UE.
[PR.5.16.6-001]
CPR 6.2-7
A 5G system with satellite access shall be able to inform an authorised 3rd party whether S&F Satellite operation is applied for communication with a UE and to provide related information (e.g. estimated delivery time to the authorised UE).
[PR 5.2.6-001]
[PR 5.2.6-005]
CPR 6.2-8
Subject to operator’s policies, a 5G system with satellite access supporting S&F Satellite operation shall be able to support forwarding of the stored data from one satellite to another satellite (e.g., which has an available feeder link to the ground network), through ISLs.
NOTE: It is assumed that the satellite constellation knows which satellite has a feeder link available. However, this is outside the scope of 3GPP.
[PR.5.1.6-008]
[PR.5.3.6-002]
CPR 6.2-9
Subject to operator’s policies, a 5G system with satellite access supporting the S&F Satellite operation shall be able to support suitable means to resume communication between the satellite and the ground station once the feeder link becomes available.
[PR 5.1.6-007]
[PR 5.2.6-007]
CPR 6.2-10
A 5G system with satellite access supporting S&F Satellite operation shall support mechanisms for a UE to register with the network when the network is in S&F Satellite operation.
[PR.5.1.6-009]
[PR 5.4.6-002]
CPR 6.2-11
A 5G system with satellite access supporting S&F Satellite operation shall support mechanisms to authorize subscribers for receiving services when the network is in S&F Satellite operation.
[PR.5.1.6-010]
[PR 5.4.6-003] |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 6.3 UE-Satellite-UE communication | The potential requirements corresponding to the support of UE-Satellite-UE communication are listed in the table below.
Table 6.3-1 – Consolidated Requirements for UE-Satellite-UE communication
CPR #
Consolidated Potential Requirement
Original PR #
Comment
CPR 6.3-1
Subject to regulatory requirements and operator’s policy, a 5G system with satellite access shall support UE-Satellite-UE communication regardless of whether the feeder link is available or not.
[PR 5.6.6-003]
[PR 5.7.6-001]
[PR 5.14.6-001]
[PR 5.6.6-005]
CPR 6.3-2
Subject to regulatory requirements and operator’s policy, a 5G system with satellite access shall be able to support service continuity (with minimum service interruption) of a UE-Satellite-UE communication when the UE communication path moves between serving satellites (due to the movement of the UE and/or the satellites).
[PR 5.6.6-004]
[PR 5.14.6-002]
CPR 6.3-3
Subject to regulatory requirements and operator’s policy, a 5G system with satellite access shall support service continuity (with minimum service interruption) of a UE-Satellite-UE communication when the communication path between UEs extends to additional satellites (through ISLs).
[PR 5.14.6-003]
CPR 6.3-4
Subject to regulatory requirements and operator’s policy, a 5G system with satellite access shall be able to provide QoS control of a UE-Satellite-UE communication
[PR 5.8.6-001]
CPR 6.3-5
Subject to regulatory requirements and operator’s policy, a 5G system with satellite access shall be able to support different types of UE-Satellite-UE communication (e.g. voice, messaging, broadband, unicast, multicast, broadcast).
[PR 5.8.6-002]
CPR 6.3-6
Subject to regulatory requirements and operator’s policy, a 5G system with satellite access shall support service continuity (with minimum service interruption) of a UE-Satellite-UE communication when one UE communication path moves between a direct network connection via 5G terrestrial access network and an indirect network connection via a relay UE (using satellite access).
NOTE: It is assumed that the 5G terrestrial access network and the satellite access network belong to the same operator.
[PR 5.10.6-002] |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 6.4 GNSS independent operation & positioning enhancements for satellite access | The potential requirements corresponding to the support of GNSS independent operation & positioning enhancements for satellite access are listed in the table below.
Table 6.4-1 – Consolidated Requirements for GNSS independent operation & positioning enhancements for satellite access
CPR #
Consolidated Potential Requirement
Original PR #
Comment
CPR 6.4-1
Subject to the regulatory requirements and operator’s policy, a 5G system with satellite access shall be able to provide services to an authorized UE independently of the UE’s GNSS capability.
[PR 5.11.6-001] [PR 5.11.6-002]
CPR 6.4-2
Subject to the regulatory requirements and operator’s policy, a 5G system with satellite access shall be able to determine the location of a UE using only satellite access (e.g. based on 3GPP positioning technologies, based on the information from reliable and trusted sources) in order to provide services.
[PR 5.11.6-003]
CPR 6.4-3
Subject to regulatory requirements and operator’s policy, a 5G system with satellite access shall be able to support 3GPP positioning methods for UEs using only satellite access.
[PR 5.13.6-001] [PR 5.12.6-001]
CPR 6.4-4
A 5G system with satellite access shall be able to provide positioning service to a UE using only satellite access and the information on positioning services (e.g. supported positioning performance).
NOTE: UE can be with or without GNSS capabilities.
[PR 5.12.6-002]
[PR 5.13.6-003]
CPR 6.4-5
A 5G system with satellite access shall be able to support negotiation of positioning methods, between UE and network, according e.g. to 3GPP RAT and UE positioning capability, the availability of non-3GPP positioning technologies (e.g. GNSS).
[PR 5.13.6-002] |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 6.5 Other aspects for satellite access | The potential requirements corresponding to the support of enhancements of other aspects of satellite access are listed in the table below.
Table 6.5-1 – Consolidated Requirements for other aspects of satellite access
CPR #
Consolidated Potential Requirement
Original PR #
Comment
CPR 6.5-1
Subject to regulatory requirements and operator’s policies, a 5G system with satellite access shall be able to support an efficient communication path and resource utilization for a UE using only satellites access, e.g. to minimize the latencies introduced by satellite links involved.
[PR 5.5.6-001]
[PR 5.5.6-002]
[PR 5.15.6-001]
CPR 6.5-2
Subject to regulatory requirements and operator’s policies, a 5G system with satellite access shall be able to support collection of information on usage statistics and location of the UEs that are connected to the satellite.
[PR 5.9.6-001] |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 6.6 Security aspects | The potential requirements corresponding to the security aspect are listed in the table below.
Table 6.6-1 – Consolidated Requirements for security aspects
CPR #
Consolidated Potential Requirement
Original PR #
Comment
CPR 6.6-1
Subject to operator’s policies, a 5G system with satellite access supporting S&F Satellite operation shall be able to preserve security of the data stored and forwarded.
[PR.5.3.6-002]
[PR.5.1.6-008]
CPR 6.6-2
A 5G system with satellite access supporting S&F Satellite operation shall be able to support mechanisms to authorize a UE to use the S&F Satellite operation.
[PR.5.3.6-004]
[PR 5.4.6-003]
[PR 5.4.6-002]
[PR 5.1.6-006]
[PR 5.2.6-006]
CPR 6.6-3
A 5G system with satellite access shall be able to support mechanisms to authorize the UE-Satellite-UE communication, based on e.g., location information and subscription.
NOTE: UEs can use satellite access directly or via a relay UE (using satellite access assuming that the 5G system with satellite access is authorized to assign spectrum resources for the communication between remote UE and relay UE).
[PR 5.6.6-001]
[PR 5.10.6-001] |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 6.7 Charging aspects | The potential requirements corresponding to the charging aspects are listed in the table below.
Table 6.7-1 – Consolidated Requirements for charging aspects
CPR #
Consolidated Potential Requirement
Original PR #
Comment
CPR 6.7-1
A 5G system with satellite access supporting S&F Satellite operation shall be able to collect charging information per UE or per application (e.g., number of UEs, data volume, duration, involved satellites).
[PR 5.4.6-005]
[PR 5.4.6-006]
CPR 6.7-2
A 5G system with satellite access shall be able to collect charging information for a UE registered to a HPLMN or a VPLMN, for UE-Satellite-UE communication.
[PR 5.6.6-002]
[PR 5.7.6-002] |
1c7e630f4af53df4342c640c46cce842 | 22.865 | 7 Conclusions and recommendations | This technical report identifies several use cases and potential new requirements related to the 5G system with satellite access. The resulting service requirements have been consolidated in clause 6. It is recommended to consider the consolidated requirements identified in this TR as the baseline for the subsequent normative work. Annex A (informative): Store and Forward Satellite operation The Store and Forward Satellite operation in a 5G system with satellite access is intended to provide some level of communication service for UEs under satellite coverage with intermittent/temporary satellite connectivity (e.g. when the satellite is not connected via a feeder link or via ISL to the ground network) for delay-tolerant communication service. An example of “S&F Satellite operation” is illustrated in Figure A-1, in contrast to what could be considered the current assumption for the “normal/default Satellite operation” of a 5G system with satellite access. As shown in Figure A-1: • Under “normal/default Satellite operation” mode, signalling and data traffic exchange between a UE with satellite access and the remote ground network requires the service and feeder links to be active simultaneously, so that, at the time that the UE interacts over the service link with the satellite, there is a continuous end-to-end connectivity path between the UE, the satellite and the ground network. - In contrast, under “S&F Satellite operation” mode, the end-to-end exchange of signalling/data traffic is now handled as a combination of two steps not concurrent in time (Step A and B in Figure A-1). In Step A, signalling/data exchange between the UE and the satellite takes place, without the satellite being simultaneously connected to the ground network (i.e. the satellite is able to operate the service link without an active feeder link connection). In Step B, connectivity between the satellite and the ground network is established so that communication between the satellite and the ground network can take place. So, the satellite moves from being connected to the UE in step A to being connected to the ground network in step B. “Normal/default Satellite operation” mode “S&F Satellite operation” mode Figure A-1: Illustration of “normal/default operation” and “S&F operation” modes in a 5G system with satellite access. The concept of “S&F” service is widely used in the fields of delay-tolerant networking and disruption-tolerant networking. In 3GPP context, a service that could be assimilated to an S&F service is SMS, for which there is no need to have an end-to-end connectivity between the end-points (e.g. an end-point can be a UE and the other an application server) but only between the end-points and the SMSC which acts as an intermediate node in charge of storing and relying. The support of S&F Satellite operation is especially suited for the delivery of delay-tolerant/non-real-time IoT satellite services with NGSO satellites. Annex B (informative): Change history Change history Date Meeting TDoc CR Rev Cat Subject/Comment New version 2022-08 SA1#99e Inclusion of agreed pCRs: S1-222089; S1-222326; S1-222327; S1-222334; S1-222335; S1-222328; S1-222333; S1-222329; S1-222331; S1-222332; S1-222330; S1-222336 0.0.0 2022-11 SA1#100 Inclusion of agreed pCRs: S1-223531; S1-223392; S1-223393; S1-223533; S1-223535; S1-223639; S1-223715; S1-223638 0.2.0 2023-02 SA1#101 Inclusion of agreed pCRs: S1-230475; S1-230673; S1-230674; S1-230469; S1-230139; S1-230470; S1-230656; S1-230141; S1-230472; S1-230785; S1-230676; S1-230669; S1-230670; S1-230679 0.3.0 2023-03 SA#99 SP-230224 MCC clean-up for presentation to SA#99 1.0.0 2023-05 SA1#102 S1-231339 Inclusion of agreed pCRs: S1-231560, S1-231575, S1-231576, S1-231577, S1-231563, S1-231578, S1-231579, S1-231208, S1-231700, S1-231740, S1-231722, S1-231121, S1-231574, S1-231702, S1-231737, S1-231088 1.1.0 2023-06 SA#100 SP-230515 MCC clean-up for approval by SA#100 2.0.0 2023-06 SA#100 SP-230515 Raised to v.19.0.0 by MCC following approval by SA#100 19.0.0 2023-09 SA#101 SP-231025 0001 D Updates in scope, terms and overview 19.1.0 2023-09 SA#101 SP-231025 0002 1 B update of consolidation for TR 22.865 19.1.0 2023-09 SA#101 SP-231026 0003 3 C Updates on use case on Store and Forward-MO for TR 22.865 19.1.0 2023-09 SA#101 SP-231025 0004 3 B update of clause 5.16 19.1.0 2023-12 SA#102 SP-231409 0006 1 D Small editorial fixes to 22.865 19.2.0 |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 1 Scope | The objective of this document is to study the use cases with potential functional and performance requirements to support efficient AI/ML operations using direct device connection for various applications e.g. auto-driving, robot remote control, video recognition, etc.
The aspects addressed in the document includes:
- Identify the use cases for distributed AI inference;
- Identify the use cases for distributed/decentralized model training;
- Gap analysis to existing 5GS mechanism to support the distributed AI inference and model training. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 2 References | The following documents contain provisions which, through reference in this text, constitute provisions of the present document.
- References are either specific (identified by date of publication, edition number, version number, etc.) or non‑specific.
- For a specific reference, subsequent revisions do not apply.
- For a non-specific reference, the latest version applies. In the case of a reference to a 3GPP document (including a GSM document), a non-specific reference implicitly refers to the latest version of that document in the same Release as the present document.
[1] 3GPP TR 21.905: "Vocabulary for 3GPP Specifications".
[2] 3GPP TR 22.874, Study on traffic characteristics and performance requirements for AI/ML model transfer in 5GS (Release 18)
[3] 3GPP TS 22.104, Service requirements for cyber-physical control applications in vertical domains
[4] Huaijiang Zhu, Manali Sharma, Kai Pfeiffer, Marco Mezzavilla, Jia Shen, Sundeep Rangan, and Ludovic Righetti, “Enabling Remote Whole-body Control with 5G Edge Computing”, to appear, in Proc. 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems. Available at: https://arxiv.org/pdf/2008.08243.pdf
[5] B. Kehoe, S. Patil, P. Abbeel, and K. Goldberg, “A survey of research on cloud robotics and automation,” IEEE Transactions on automation science and engineering, vol. 12, no. 2, pp. 398–409, 2015.
[6] M. Chen, K. Huang, W. Saad, M. Bennis, A. V. Feljan, and H. V. Poor, “Distributed Learning in Wireless Networks: Recent Progress and Future Challenges”IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 39, NO. 12, DECEMBER 2021
[7] M. Chen, H. V. Poor, W. Saad, and S. Cui, “Wireless communications for collaborative federated learning,” IEEE Commun. Mag., vol. 58, no. 12, pp. 48–54, Dec. 2020
[8] Jianmin Chen, Xinghao Pan, Rajat Monga, Samy Bengio, Rafal Jozefowicz, “Revisiting Distributed Synchronous SGD,” arXiv preprint arXiv: 1604.00981, 2016
[9] Shuxin Zheng, Qi Meng, Taifang Wang, Wei Chen, Nenghai Yu, Zhi-Ming Ma, Tie-Yan Liu, “Asynchronous Stochastic Gradient Descent with Delay Compression” arXiv: 1609.08326, 2020
[10] 3GPP TR 21.905: "Service requirements for the 5G system".
[11] Yusuf Aytar, Carl Vondrick, Antonio Torralba: "SoundNet: Learning Sound Representations from Unlabeled Video", 27 Oct 2016.
[12] Iacovos Ioannou et al.: "Distributed Artificial Intelligence Solution for D2D Communication in 5G Networks", 20 Jan 2020.
[13] Pimmy Gandotra et al.: "Device-to-Device Communication in Cellular Networks: A Survey".
[14] Davide Villa et al.: "Internet of Robotic Things: Current Technologies, Applications, Challenges and Future Directions", 15 Jan 2021.
[15] Charles R. Qi et al.: "PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation", 10 Apr 2017.
[16] 3GPP TS 22.261: "Service requirements for the 5G system".
[17] 3GPP TS 23.303: "Proximity-based services (ProSe); Stage 2".
[18] 3GPP TS 22.104: "Service requirements for cyber-physical control applications in vertical domains; Stage 1".
[19] https://www.robots.ox.ac.uk/~vgg/software/vgg_face/.
[20] Y. Kang et al., "Neurosurgeon: Collaborative intelligence between the cloud and mobile edge", ACM SIGPLAN Notices, vol. 52, no. 4, pp. 615–629, 2017.
[21] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "ImageNet classification with deep convolutional neural networks", in Proc. NIPS, 2012, pp. 1097–1105.
[22] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition", in Proc. IEEE CVPR, Jun. 2016, pp. 770-778.
[23] Zhang Z, Wang S, Hong Y, et al. Distributed dynamic map fusion via federated learning for intelligent networked vehicles[C]//2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2021: 953-959.
[24] https://github.com/open-mmlab/OpenPCDet
[25] To Transfer or Not To Transfer, Massachusetts Institute of Technology, MIT, Michael T. Rosenstein, et al.
[26] Wang, J. et al. Easy Transfer Learning by Exploiting Intra-domain Structures. In 2019 IEEE International Conference on Multimedia and Expo (ICME), pages 1210-1215 IEEE.
[27] Wang K C, Fu Y, Li K, et al. Variational model inversion attacks[J]. Advances in Neural Information Processing Systems, 2021, 34: 9706-9719.
[28] Ming-Fang Chang, John Lambert, Patsorn Sangkloy, et. al. Argoverse: 3D Tracking and Forecasting with Rich Maps. arXiv:1911.02620v1 [cs.CV] 6 Nov 2019. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 3 Definitions, symbols and abbreviations | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 3.1 Definitions | For the purposes of the present document, the terms and definitions given in 3GPP TR 21.905 [1] and the following apply. A term defined in the present document takes precedence over the definition of the same term, if any, in 3GPP TR 21.905 [1].
Proximity-based work task offloading: based on 3rd party’s request, a relay UE receives data from a remote UE via direct device connection and performs calculation of a work task for the remote UE. The calculation result can be further sent to network server. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 3.2 Symbols | For the purposes of the present document, the following symbols apply:
<symbol> <Explanation> |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 3.3 Abbreviations | For the purposes of the present document, the abbreviations given in 3GPP TR 21.905 [1] and the following apply. An abbreviation defined in the present document takes precedence over the definition of the same abbreviation, if any, in 3GPP TR 21.905 [1].
<ACRONYM> <Explanation> |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 4 Overview | In TR 22.874, three types of AIML operations as below has been described
• AI/ML operation splitting between AI/ML endpoints;
• AI/ML model/data distribution and sharing over 5G system;
• Distributed/Federated Learning over 5G system.
For the phase-2 study, it continues to study how the 5GS can have more gains for each type of AIML operations when leveraging direct device connection. Thus, the following clause 5, 6, and 7 is to capture use cases corresponding to the three types of AIML operations considering the assistance of direct device connection. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5 Split AI/ML operation between AI/ML endpoints for AI inference by leveraging direct device connection | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.1 Proximity based work task offloading for AI/ML inference | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.1.1 Description | The model splitting is the most significant feature for AI inference. As some R18 use cases in TR 22.874[2] shows, the number of terminal computing layers and the amount of data transmission are corresponding to different model splitting points. For example, as figure 5.1-1 shows, the general trend is that the more layers the UE calculated, the less intermediate data needs to be transmitted to application server. In another word, when UE has low computation capacity (e.g. due to low battery), the application can change the splitting point to let UE calculate fewer layers while increasing the data rate in Uu for transmitting a higher load of intermediate data to network.
However, sometimes the data rate cannot be increased due to radio resource limitation, in such circumstances, UE with low computation capacity needs to offload the computation task to a proximity UE (likely a relay UE) but still keeping the computation service and let the proximity UE to send the calculated data to network. Thus, by offloading the work task using direct device connection, the original UE’s computation load will be released while the data rate in Uu interface will not necessarily be increased either, which leads to a more ideal performance.
Figure 5.1-1. Layer-level computation/communication resource evaluation for an AlexNet model (abstracted from subclause 5.1.1 in TR 22.874) |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.1.2 Pre-conditions | A UE uses the AI model (AlexNet) for image recognition. As predetermined by application, there are 5 alternative splitting points which are corresponding to intermediate data size and data rate, see reference [13-14] in TR 22.874, while fewer the layers being calculated implies fewer the workload being performed by UE. The specific values are shown in the table below (it is abstracted from clause 5.1 Split AI/ML image recognition in TR22.874).
Table 5.1-1: Required UL data rate for different split points of AlexNet model for video recognition @30FPS (Frame Per Second)
Split point
Approximate output data size (MByte)
Required UL data rate (Mbit/s)
Candidate split point 0
(Cloud-based inference)
0.15
36
Candidate split point 1
(after pool1 layer)
0.27
65
Candidate split point 2
(after pool2 layer)
0.17
41
Candidate split point 3
(after pool5 layer)
0.02
4.8
Candidate split point 4
(Device-based inference)
N/A
N/A |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.1.3 Service Flows | (a) no task offloading (b) task offloading by UE-B
Figure 5.1-2: Using direct device connection (sidelink) to realize the proximity-based work task offloading. In this case, the data rate on Uu need not be increased while the original UE’s computation load is offloaded
1. As shown in left(a) of Figure 5.1-2, UE-A is doing image recognition using Alexnet Model as described in clause 5.4.2. It selects splitting point-3 for the AI inference.
The E2E service latency (including image recognition latency and intermediate data transmission latency) is 1 second.
2. When the UE-A’s battery becomes low, it cannot afford the heavy work task for the AlexNet model (i.e. calculating layer 1-15 for AlexNet model in local side).
3. Being managed by 5G network, the UE-A discovers UE-B (a Customer Premise Equipment, CPE) which has installed the same model and is willing to take the offloading task from UE-A.
NOTE 1: The 5G network does not store UE-A and UE-B’s location data.
Then UE-A established the sidelink (direct device connection) to UE-B. During the sidelink establishment, the UE-B also gets the information of the total service latency (including the image recognition latency and intermediate data transmission latency) and the processing time consumed by UE-A for computing layer 1-4.
Since the UE-B has acquired the E2E service latency and the processing time consumed by UE-A, and also it knows its own processing time for computing layer 5-15, the UE-B can determine the QoS parameters applied to both Uu and Sidelink while keeping the E2E service latency same as the E2E service latency described in step-1.
NOTE 2: It is assumed that the UE-A and UE-B have the same computation capacity, i.e. the time used for computing the certain AlexNet model layers are the same for UE-A and UE-B. Otherwise, the data rate on Uu and Sidelink may be changed accordingly.
4. The UE-A sends the intermediate data (data after calculating layer 1-4) to UE-B via sidelink and let UE-B make further processing then transmit the intermediate data (data after calculating layer 5-15) to application server via Uu. The specific model layers being computed by UE-A and UE-B are shown in the right(b) in figure 5.1-2.
5. UE-A continues to perform image recognition by leveraging sidelink and UE-B’s computation capacity while the source and destination IP address and the E2E service latency for the image recognition service is unchanged. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.1.4 Post-conditions | Thanks to UE-B’s help, the proximity-based work task offloading is performed. By doing so,
- it decreased the UE-A’s work task by letting UE-A to compute fewer layers of AlexNet model, which helps to meet the low battery condition happened to UE-A;
- the UE-B computes the rest of layers which is originally from the UE-A’s work task;
- the mobile network does not need to increase the QoS parameters such as guaranteed data rate because the intermediate data rate transmitted by UE-B is unchanged. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.1.5 Existing features partly or fully covering the use case functionality | In TS 22.261 clause 6.9, the description about the direct network connection mode and the indirect network connection mode as well as the service continuity for switching between the two modes have been described. They are summarized as below:
The UE (remote UE) can connect to the network directly (direct network connection), connect using another UE as a relay UE (indirect network connection), or connect using both direct and indirect connections.
The 5G system shall support different traffic flows of a remote UE to be relayed via different indirect network connection paths.
The 5G system shall be able to maintain service continuity of indirect network connection for a remote UE when the communication path to the network changes (i.e. change of one or more of the relay UEs, change of the gNB).
However, there is no proximity-based work task offloading which means that the “relay UE” not only performs the indirect network communication but also performs task computation for the “remote UE”. This may impact the current discovery mechanism, QoS determination on Uu and PC5, and charging aspect. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.1.6 Potential New Requirements needed to support the use case | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.1.6.1 Potential Functionality Requirements | [P.R.5.1.6-001] The 5G system shall be able to support the means to modify the communication QoS ensuring the end-to-end latency can be satisfied when a relay UE is involved for a proximity-based work task offloading.
NOTE 1: Due to the proximity-based work task offloading, the data size transmitted via sidelink and Uu of the indirect network connection is different
[P.R.5.1.6-002] The 5G system shall be able to collect charging information for proximity-based work task offloading.
[PR.5.1.6-003] The 5G system shall support service continuity when a UE communication path changes between a direct network connection and an indirect network connection, including the case when the data size transmitted over the two connection is different (e.g. for a proximity-based work task offloading). |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.1.6.2 Potential KPI Requirements | Considering the widely-used AlexNet and VGG-16 model for proximity-based work task offloading, the following KPIs need to be supported:
Table 5.1-2 KPI requirements for proximity-based work task offloading
UL data size
(for sidelink)
UL data rate
(for sidelink)
Intermediate data uploading latency (including sidelink+Uu)
Image recognition latency
AlexNet model with 30FPS (NOTE 1)
0.15 - 0.02 Mbyte for each frame
4.8 – 65 Mbit/s
- 2ms for Remote driving, AR displaying/gaming, and remote-controlled robotics;
- 10ms for video recognition;
- 100ms for One-shot object recognition, Person identification, or photo enhancement in smart phone
1s
VGG-16 model with 30FPS
0.1 - 1.5 Mbyte for each frame
24 - 720 Mbit/s
1s
NOTE 1: FPS stands for Frame Per Second |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.2 Local AI/ML model split on factory robots | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.2.1 Description | In a modern factory, a team on a workstation comprises two human operators, two mobile robots and a fixed robot. Everyone has his own pre-defined task. Robots assist the human operators by accomplishing painful tasks in a fluid and precise manner; they also monitor that the workstation environment remains safe for the human operators. The mobile robots must not interfere with humans and between them.
The robot control is not executed on a distant server in the cloud because reliability and confidentiality were not ensured at a sufficient level. Furthermore, as stated in [14], the overall end-to-end latency is not always guaranteed which can cause production loss. Communications between robots can rely on private wireless networks in the factory that enable expected QoS (reliability, throughput, and latency) as well as confidentiality.
The new robots are autonomous robots that can react to human voices or learn in real-time what operators do. They can perceive their environment and transmit information to other robots. They can communicate, learn from each other, assist each other and do self-monitoring.
The autonomous robot’s skills rely on several AI/ML models running on the robot itself which has the inconvenience that the mobile robot’s battery drain quicker. To overcome this issue, when the battery level reaches a certain value, a part of the AI/ML model can be transferred to a service hosting environment and / or to another robot by splitting the AI/ML model as defined in [20]. The split model approach of [20] is applicable to a UE-to-UE (or robot-to-robot) architecture. Thus, the AI/ML model M is split and shared between (e.g.) 2 robots, say an assisted robot and an assistant robot. Intermediate data generated by the assisted robot are transferred to the assisting robot which finalizes the inference and transmits the results back to the assisted robot. This intermediate data transfer must be extremely efficient in terms of latency and throughput. When many models are at stake, the split model method is an additional challenge for the 5GS in terms of throughput, latency and synchronization.
Because they are more autonomous, mobile and smart, the industry 4.0 robots embed a large variety of sensors that generate huge amount of data to process. Table 5.2.1-1 reflects that variety.
Each type of sensing data requires a different AI/ML model. Each of these models produce predictions in a certain delay and with a certain accuracy.
Thus, as an offloading strategy, we can imagine that a model is split between 2 robots because it has been established that for a particular AI/ML model, the latency with the sidelink communication was better (smaller) than with the regular 5G communication path, as stated in [12] and [13].
At the same time, other AI/ML models are split between the robot and the service hosting environment because from an energy standpoint this configuration is the best.
Table 5.2.1-2 is an example of this offloading strategy where four AI/ML models are split between a robot and the service hosting environment and four other AI/ML models are split between two robots.
Sidelink and 5G communication paths are complementary from an AI/ML model split policy standpoint.
Table 5.2.1-1 shows some typical and diverse AI/ML models that can be used on robots. For each model, all the split point candidates have been considered and only the split points that generate the minimum and the maximum amount of intermediate data have been noted.
Table 5.2.1-1: Intermediate AI/ML data size per AI/ML model
Model Name
Model type
Intermediate data size (MB)
8 bits data format
32 bits data format
Min
Max
Min
Max
AlexNet [21]
Image recognition
0.02
0.06
0.08
0.27
ResNet50 [22]
Image recognition
0.002
1.6
0.008
6.4
SoundNet [11]
Sound recognition
0.0017
0.22
0.0068
0.88
PointNet [15]
Point Cloud
0.262
1.04
0.0068
4.19
VGGFace [19]
Face recognition
0.000016
0.8
0.000064
3.2
Inception resnet
Face recognition
0.0017
0.37
0.0068
1.51
In Table 5.2.1-2 the AI/ML models are distributed between the service hosting environment and the proximity robot. The way the models are distributed is out of scope of this use case and depends on various criteria as said previously. Therefore, the next table is an example that illustrates the distribution. The intermediate data size is presented with the range [Min – Max], where Min and Max are respectively the sum of the Min values and the sum of the Max values of the selected models as defined in Table 5.2.1-1 (figures in bold).
Table 5.2.1-2: Example of models distribution and data rate for intermediate AI/ML data
Model Name
Offloading target
Intermediate data size (MB)
Transfer time (ms)
Data rate (Gb/s)
AlexNet [21]
Proximity robot or Service Hosting Environment
[0.000016 – 1.6]
(8 bits data format)
10
[0.128 - 1.28]
ResNet50 [22]
VGGFace [19]
SoundNet [11]
[0.000064 – 6.4]
(32 bits data format)
10
[0.512 – 5.12]
PointNet [15]
Inception resnet
As previously said latency is a critical requirement. Figure 5.2.1-3 summarizes the latency cost in three scenarios:
(A) The inference of model M is done locally. Latency is denoted LLI.
(B) The inference process is fully offloaded on a second device (Robot/UE). Latency is denoted LFO.
(C) The inference process is partially offloaded on a second device (Robot/UE). Latency is denoted LPO.
Figure 5.2.1‑3 Latency summary
The current Use Case promotes the scenario (C) where a model M is split in two sub-models Ma and Mb. If both robots (UEs) have a similar computing power, the assumption is that the latency due to the inference of model M is almost equal to the latency of model Ma plus the latency of model Mb.
Hence, once the split model is deployed on the two robots (UEs), the aim is to minimize the E2E latency and to be as close as possible to the non-split case. This is done with a transfer delay of both the intermediate data and the inference results as small as possible. We can note that if the computing power on the assistant robot is more important, then scenario (C) would be the preferred scenario.
In scenario (B), the inference process is fully offloaded on the assistant robot (UE). The major inconvenience is the strong and negative impact on latency of the raw data transfer towards the assistant robot.
5.2.2 Pre-conditions
Two human operators are working.
Two mobile AI-driven robots (Arobot and Brobot) and one static AI-driven robot (Crobot) assist them.
The three robots (Arobot, Brobot and Crobot) belong to the same service area, embed the same two powerful AI/ML models M1 and M2, sensors (e.g. LIDAR, microphone) and cameras (e.g., 8K videos stream).
Arobot and Brobot are powered with a battery and Crobot with fixed ground power.
The three robots (Arobot, Brobot and Crobot) are connected, e.g., to the AF, 5GC, or to each other using D2D technologies (Prose, BT, WiFi, etc.).
The workstation is equipped with camera and sensors.
The service area is 30 m x 30 m and the robot speed is at maximum 10 km/h.
The service area is covered by a small cell and a service hosting environment is connected and can support AI/ML processes.
5.2.3 Service Flows
Figure 5.2.3‑1 Factory service flow
a) Brobot battery level is rather low but it can still work for a while if a part of its machine learning process is offloaded.
b) Brobot broadcasts a request message to get assistance. Crobot and the service hosting environment responds positively.
c) Brobot negotiates with Crobot and the service hosting environment what parts of M1 and M2 respectively for the inference process they are in charge of, knowing that the quality of the prediction must not be under a certain level and that the end-to-end latency must not be above a certain value. M1 model is split between Brobot and the service hosting environment. M2 model is split between Brobot and Crobot.
d) Brobot, Crobot and the service hosting environment agree on split points for both M1 and M2 models and Brobot starts sending the intermediate data to Crobot and the service hosting environment.
e) Crobot infers and transmits using unicast mode with a very short delay the predictions back to Brobot. The service hosting environment infers and transmits using unicast mode the predictions back to Brobot in a very short delay.
f) In the meanwhile, Arobot is carrying a load to the operator Aoperator.
g) Aoperator bends down to pick up a screw that has fallen on the floor. At the same time Boperator is passing between Aoperator and Arobot. Arobot can’t see Aoperator anymore.
h) Brobot is busy with another task, but it can observe the scene. It reports the scene as intermediate data to Crobot and the service hosting environment.
i) Crobot and the service hosting environment amend the ML model based on the new training data.
j) Crobot and the service hosting environment infer and then transmit in unicast the prediction back to Brobot. The safety application on the service hosting environment collects the inference results.
5.2.4 Post-conditions
Intermediate data can be exchanged between two robots (UEs) and / or service hosting environment, and the robot with a low battery level can continue working for a while.
All the robots in the group receive the alert message and react:
a) They all stop working; or
b) Arobot changes its trajectory.
Aoperator and Boperator can work safely.
The huge amount of data that is required for inferring is kept local in the factory.
5.2.5 Existing features partly or fully covering the use case functionality
The Use Case can rely on the Proximity Service (ProSe) services as defined in 3GPP TS 23.303 [17].
Cyber-Physical Control Applications, see 3GPP TS 22.104 [18], already proposes to rely on a ProSe communication path. The proposed requirements are limited in terms of data transfer as shown in Table 5.2-1, where the message size does not exceed a few hundred of Bytes (250 kB at maximum).
3GPP TS 22.261 [16] clause 6.40 provides requirements for AI/ML model transfer in 5GS. The requirements in this clause does not consider requirements for direct device connection.
In 3GPP TS 22.261 [16] Table 7.6.1-1, the max. end-to-end latency is 10 ms, the maximum data rate is [1] Gbits/s and reliability is 99.99% for Gaming or Interactive Data Exchanging.
Table 7.6.1-1 KPI Table for additional high data rate and low latency service
Use Cases
Characteristic parameter (KPI)
Influence quantity
Max allowed end-to-end latency
Service bit rate: user-experienced data rate
Reliability
# of UEs
UE Speed
Service Area
(note 2)
Gaming or Interactive Data Exchanging
(note 3)
10ms (note 4)
0,1 to [1] Gbit/s supporting visual content (e.g. VR based or high definition video) with 4K, 8K resolution and up to120 frames per second content.
99,99 % (note 4)
≤ [10]
Stationary or Pedestrian
20 m x 10 m; in one vehicle (up to 120 km/h) and in one train (up to 500 km/h)
NOTE 1: Unless otherwise specified, all communication via wireless link is between UEs and network node (UE to network node and/or network node to UE) rather than direct wireless links (UE to UE).
NOTE 2: Length x width (x height).
NOTE 3: Communication includes direct wireless links (UE to UE).
NOTE 4: Latency and reliability KPIs can vary based on specific use case/architecture, e.g. for cloud/edge/split rendering, and can be represented by a range of values.
NOTE 5: The decoding capability in the VR headset and the encoding/decoding complexity/time of the stream will set the required bit rate and latency over the direct wireless link between the tethered VR headset and its connected UE, bit rate from 100 Mbit/s to [10] Gbit/s and latency from 5 ms to 10 ms.
NOTE 6: The performance requirement is valid for the direct wireless link between the tethered VR headset and its connected UE.
These requirements partially cover the current Use Case needs.
5.2.6 Potential New Requirements needed to support the use case |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.2.6.1 Potential Functionality Requirements | [P.R.5.2.6-001] Subject to user consent and operator policy, the 5G system shall support the transfer of AI/ML model intermediate data from UE to UE via the direct device connection.
[P.R.5.2.6-002] Subject to user consent and operator policy, the 5G system shall be able to provide means to predict and expose network condition changes (i.e. bitrate, latency, reliability) and receive user preferences on usage of the direct device connection or the direct network connection in order to meet the user experienced data rate and latency.
[P.R.5.2.6-003] Subject to user consent and operator policy, the 5G system shall be able to dynamically select the intermediate device that is capable to perform the needed functionalities, e.g., AIML splitting.
[P.R.5.2.6-004] Subject to user consent and operator policy, the 5G system shall be able to maintain the QoS (latency, reliability, data rate as defined in the Table 5.2.6.2-1 below) of the communication path of the direct device connection.
[P.R.5.2.6-005] Subject to user consent and operator policy, the 5G system shall be able to have the means to modify the QoS of the communication path of the direct device connection.
NOTE: The split point selection is dynamic. In consequence, the amount of intermediate data will vary. To maintain the QoS, the bandwidth is adjusted. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 5.2.6.2 Potential KPI Requirements | Based on Table 5.2.1-2, the potential KPI requirement is as below:
Table 5.2.6.2-1 KPI for intermediate AI/ML data transmission for model split based robot control
Model Name
Payload size (Intermediate data size)
Max allowed end-to-end latency
Experienced data rate
Service area dimension
Communication service availability
Reliability
AlexNet [21]
0.000016 – 1.6 MByte
(8 bits data format)
10 ms
0.128 - 1.28 Gbps
900 m2
(30 m x 30 m)
99.999 %
99.999 %
ResNet50 [22]
VGGFace [19]
SoundNet [11]
PointNet [15]
0.000064 – 6.4 Mbyte
(32 bits data format)
10 ms
0.512 - 5.12 Gbps
Inception resnet |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6 AI/ML model/data distribution and sharing by leveraging direct device connection | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.1 AI Model Transfer Management through Direct Device Connection | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.1.1 Description | Based on the earlier study in phase one 3GPP TR 22.874 [2], operators can provide services to help manage and distribute the AI/ML models especially in the edge server so that the UE can acquire a proper model immediately. However, when a lot of UEs requesting for the same model at the same time or the UE is blocked by barriers with poor connection with the base station, the model transfer process will become longer than expected.
To overcome this difficulty, as shown in Fig.1, a volunteer UE which is well connected to the base station can help relaying AI/ML models or receive and store AI/ML models first. Then, the other UEs can download AI/ML models from the volunteer UE through direct device connection. In this way, all UE can have a stable and reliable model transfer process while the radio resource of the base station can be saved. Besides, the volunteer UE can transfer the stored models to other volunteer UEs under operator’s control.
The selection of volunteer UE can be realized by local network policies and strategies. And it also can be exposed as a capability to the 3rd party company when the company wants to choose one or a few certain UEs to be volunteer UEs in an activity. For example, a travel company may assign the tour guides’ Augmented Reality (AR) headsets as volunteer UEs in a carnival through the operator’s network exposure. The travel company may sign a higher quality plan for tourist guides’ devices to provide better user experience for following tourists. Meanwhile, operator can benefit from the alternative open service based on AI/ML model management capabilities and may avoid low Quality of Service due to crowding direct connections to base stations during the carnival.
Fig.1 AI/ML Model management through direct device connection |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.1.2 Pre-conditions | The operator’s MEC near the Jurassic Park stores a variety of AI/ML models according to the the park company’s requirements. And it is capable to transfer the stored model to the device such as AR headset.
The operator rolls out a new high-quality plan which can allow the user customizes own Service Level Agreement (SLA) for specific network address access and data (e.g. AI/ML Models) download. As a trade-off, the user’s device will help transfer the same data through direct device connection to nearby devices sharing common aspiration.
The AR headset can transfer the stored AI/ML model to the other AR headsets. However, the AR headset cannot store all models for different scenarios due to limited storage. Indeed, a model needs to be downloaded when or a few seconds before the UE first appears in the certain area.
Alice and Bob are tour guide hired by Jurassic Park and their real-time positions can be acquired when they are in the park based on signed agreements.
All of AR headsets should in the coverage of the base stations serving for the Jurassic Park. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.1.3 Service Flows | 1. Jurassic Park provides panorama AR tour guide services in a commercial area and a tropical rainforest area. AR headsets need to download Model A and B (both are VGG-16, 552MByte) respectively.
2. To provide high quality user experience, Jurassic Park company indicates to the operator that AR headsets require to download model A in area A and model B in area B.
3. The Jurassic Park company signs a high-quality plan for tour guide Alice and Bob’s AR Headsets for providing better service to the tour group using direct device connection.
4. When Bob and his tour group enter area A, their headsets request for the Model A. The operator network finds they requested the same model and Bob is a signed volunteer UE, then triggers to establish a QoS acceleration for Bob’s model downloading timely within 1 second. Meanwhile, Jurassic Park requests the operator network to inform Bob to help transfer the model to all other UE near Bob. Also, the operator network informs all other UE near Bob that Bob can provide the model as well. The UE which is a little far from Bob (e.g. out of Bob’s coverage) will still download the model through the base station directly.
5. Alice and her group are 10 meters far from Bob and also move towards to area A. Jurassic Park predicts their desire model based on their movement and finds Bob has already downloaded it based on the model transferring records. Jurassic Park requests operator network to inform Alice that she can request model from Bob. Meanwhile, the operator network indicates all other UE near Alice to download the model from Alice.
6. For Alice and Bob, they can see the status of all direct device connections to themselves through network exposures providing by operators (e.g. monitored bandwidth and latency of each direct device connection)
7. When Alice and Bob notice that their groups have a poor QoS of model transfer through direct device connection, they can send a request to the park company for promoting the performance of their direct device connections and the park company will send a similar message to the operator through network exposure to active a temporary acceleration of these direct device connections (e.g. expand the bandwidth of each direct device connection). |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.1.4 Post-conditions | 1. The tourists can enjoy the continuous AR services with smooth model switchover when their location and responding models change.
2. Tour group’s AR headsets provides user experience of the panorama AR tour guide services that can help retrain and improve AI/ML models in operator’s MEC by Jurassic Park company (e.g. Federated/Distributed Learning).
3. the operator network performs analytics, based on network statistics and quality of experience reported by Jurassic Park company, to improve and optimized the model transfer process (e.g. setting constraints for maximum direct device connection for one volunteer UE and choose a temporary volunteer UE for sharing model transfer task). |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.1.5 Existing features partly or fully covering the use case functionality | In 3GPP TS 22.261 [8] clause 6.27.2 "Requirements"
The 5G system shall be able to make the position-related data available to an application or to an application server existing within the PLMN, external to the PLMN, or in the User Equipment.
In 3GPP TS 22.261 [8] clause 6.9.2.4 "Relay UE Selection"
The 3GPP system shall support selection and reselection of relay UEs based on a combination of different criteria e.g.
- the characteristics of the traffic that is intended to be relayed (e.g. expected message frequency and required QoS),
- the subscriptions of relay UEs and remote UE,
- the capabilities/capacity/coverage when using the relay UE,
- the QoS that is achievable by selecting the relay UE,
- the power consumption required by relay UE and remote UE,
- the pre-paired relay UE,
- the 3GPP or non-3GPP access the relay UE uses to connect to the network,
- the 3GPP network the relay UE connects to (either directly or indirectly),
- the overall optimization of the power consumption/performance of the 3GPP system, or
- battery capabilities and battery lifetime of the relay UE and the remote UE.
NOTE: Reselection may be triggered by any dynamic change in the selection criteria, e.g. by the battery of a relay UE getting depleted, a new relay capable UE getting in range, a remote UEs requesting additional resources or higher QoS, etc.
In 3GPP TS 22.261 [8] v18.6.0 clause 6.40.2
Based on operator policy, the 5G system shall be able to provide an indication about a planned change of bitrate, latency, or reliability for a QoS flow to an authorized 3rd party so that the 3rd party AI/ML application is able to adjust the application layer behaviour if time allows. The indication shall provide the anticipated time and location of the change, as well as the target QoS parameters.
Subject to user consent, operator policy and regulatory constraints, the 5G system shall be able to support a mechanism to expose monitoring and status information of an AI-ML session to a 3rd party AI/ML application.
NOTE: Such mechanism is needed for AI/ML application to determine an in-time transfer of AI/ML model.
Subject to user consent, operator policy and regulatory requirements, the 5G system shall be able to expose information (e.g. candidate UEs) to an authorized 3rd party to assist the 3rd party to determine member(s) of a group of UEs (e.g. UEs of a FL group). |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.1.6 Potential New Requirements needed to support the use case | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.1.6.1 Potential Functionality Requirements | [P.R.6.1-001] Subject to user consent, operator policies and regional or national regulatory requirements, the 5G system shall be able to support means to monitor a direct device connection and expose corresponding monitoring information (e.g. experienced data rate, latency) to an authorized 3rd party.
NOTE: The monitoring information in [P.R.6.1-001] doesn’t include any user position-related data.
[P.R.6.1-002] Subject to user consent and operator policies, the 5G system shall be able to provide means for an authorized third-party to authorize a group of UEs to exchanging data with each other via direct device connection.
[P.R.6.1-003] The 5G system shall support a mechanism for an authorized third-party to negotiate a suitable QoS of direct device connections for a group of UEs to exchange data with each other.
[P.R.6.1-004] Subject to user consent, operator policies and regulatory requirements, the 5G system shall support means to monitor, characteristics of traffic relayed by a UE participating in the communication and expose to 3rd party. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.1.6.2 Potential KPI Requirements | [P.R.6.1-005] The 5G system shall support to use direct device communication to transmit the AI/ML model of image recognition and 3D object recognition with the following KPIs.
Table 6.1.6.2-1: KPIs for image recognition and 3D object recognition using direct device connection
Model Type
Max allowed DL end-to-end latency
Experienced data rate
in PC5
Model size
Communication service availability
AlexNet
1s
1.92 Gbit/s
240 MByte
99.9 %
ResNet-152
1s
1.92 Gbit/s
240 Mbyte
99.9 %
ResNet-50
1s
0.8 Gbit/s
100 Mbyte
99.9 %
GoogleNet
1s
0.218 Gbit/s
27.2 Mbyte
99.9 %
Inception-V3
1s
0.736 Gbit/s
92 Mbyte
99.9 %
PV-RCNN
1s
0.4 Gbit/s
50 Mbyte
99.9 %
PointPillar
1s
0.14 Gbit/s
18 Mbyte
99.9 %
SECOND
1s
0.16 Gbit/s
20 Mbyte
99.9 %
For the size of image recognition model, it refers to table 6.1.1-1 in TR22.874 [2], for the size of 3D object recognition model, see [24].
Reliability is assumed to be [99.9 – 99.999]% |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.2 5GS assisted transfer learning for trajectory prediction | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.2.1 Description | AIML model transfer learning is beneficial for lowing cost and raising effective when training a model using a target UE based on a pre-training model. The principle of transfer learning is to use the knowledge from the source domain to train a model in the target domain to achieve more expedient and higher accuracy efficiency [25].
Figure 6.2-1 AI/ML model transfer learning from source UE to target UE [26]
Since the AI model is a kind of knowledge, when the centralized application server acquires enough number of AIML model used by UEs, it may perform a backward inference/inversion attacks [27] to derive the feature of UE’s local data set, which means a privacy risk exists. In order to resolve the privacy concern for transfer learning, the model transfer via direct device connection is a better to be used so that the network node (e.g. application server) cannot acquire the AIML model used by UE and no way to do backward inference. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.2.2 Pre-conditions | Alice is a customer of intelligent-driving service provided by company-A. She lives in Chaoyang district in Beijing and driving to her office building in CBD every working day. By using the intelligent driving service, Alice’s car can predict the trajectory of neighbouring vehicles (as figure 6.2.2 shows), so as to pre-alert Alice of some potential collision and Alice can decide whether to steer, accelerate, or any other driving operation.
Figure 6.2-2: Qualitative results using model of trajectory prediction: the orange trajectory represents the observed 2s. Red represents ground truth for the next 3 seconds and green represents the multiple forecasted trajectories for those 3s [24].
An AIML model can be for the object recognition and prediction, the model is offered by company-A and customers of company-A have signed “smart driving project” (an agreement for AIML model sharing and improvement). |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.2.3 Service Flows | 1. Bob bought a car equipped with intelligent driving functionality and he would like to use auto-driving for his daily driving, so he applies to company-A to offer the intelligent-driving service.
2. Company-A needs to install certain AIML model to Bob’s car while use Bob’s local data to train the model. The company-A identified Alice’s model to be shared to Bob’s car.
In order to minimize privacy issue, the “smart driving project” signed by customer only allows the model to be transferred among users directly instead of letting application server to acquire and forward it.
3. Company-A requests 5G system to transmit the AIML model for intelligent driving from Alice’s car to Bob’s car via direct device connection at a proper time (e.g. when the direct device connection can be established)
4. When acquiring the AI model from Alice’s car, Bob’s car performs “fine-tuning” operation of transfer learning based on the local data to tune the model to be better used for its own intelligent-driving service. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.2.4 Post-conditions | Thanks to 5GS assisted AIML model transfer via direct device connection, Bob’s car efficiently gets an ideal AIML model for intelligent-driving by means of transfer learning. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.2.5 Existing features partly or fully covering the use case functionality | In 3GPP TS 22.261 [8] v18.6.1 clause 6.9
The 5G system shall support different traffic flows of a remote UE to be relayed via different indirect network connection paths.
The connection between a remote UE and a relay UE shall be able to use 3GPP RAT or non-3GPP RAT and use licensed or unlicensed band.
The connection between a remote UE and a relay UE shall be able to use fixed broadband technology.
The 5G system shall be able to provide indication to a remote UE (alternatively, an authorized user) on the quality of currently available indirect network connection paths.
The 5G system shall be able to maintain service continuity of indirect network connection for a remote UE when the communication path to the network changes (i.e. change of one or more of the relay UEs, change of the gNB).
The 5G system shall be able to support a UE using simultaneous indirect and direct network connection mode.
The 5G system shall enable the network operator to authorize a UE to use indirect network connection. The authorization shall be able to be restricted to using only relay UEs belonging to the same network operator. The authorization shall be able to be restricted to only relay UEs belonging to the same application layer group.
In 3GPP TS 22.261 [8] v18.6.1 6.40
Subject to user consent, operator policy and regulatory requirements, the 5G system shall be able to expose information (e.g. candidate UEs) to an authorized 3rd party to assist the 3rd party to determine member(s) of a group of UEs (e.g. UEs of a FL group). |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.2.6 Potential New Requirements needed to support the use case | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.2.6.1 Potential Functionality Requirements | [P.R.6.2-001] Based on user consent and 3rd party request, operator policy, the 5G system shall support a means to authorize specific UEs to transmit data (e.g. AI-ML model tansfer for a specific application) via direct device connection in a certain location and time.
[P.R.6.2-002] Subject to user consent and operator policy, the 5G system shall be able to expose information to an authorized 3rd party to assist the 3rd party to determine candidate UEs for data transmission via direct device connection (e.g. for AIML model transfer). |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 6.2.6.2 Potential KPI Requirements | [P.R.6.2-003] The 5G system shall be able to support transmitting an AI/ML model via direct device connection fulfilling the KPIs for transmission of typical AIML model for trajectory prediction and object recognition [24][28] in Table 6.2-1.
Table 6.2-1
Payload size
Latency for model transmission (NOTE 1)
Transmission Data rate
LaneGCN
15 MByte
3 seconds
5 MByte/s
ResNet-50
25 MByte
3 seconds
8.33 MByte/s
ResNet-152
60 MByte
3 seconds
20 Mbyte/s
PointPillar
18 MByte
3 seconds
6 MByte/s
SECOND
20 MByte
3 seconds
6.67 MByte/s
Part-A2-Free
226 MByte
3 seconds
75.33 MByte/s
Part-A2-Anchor
244 MByte
3 seconds
81.33 MByte/s
PV-RCNN
50 MByte
3 seconds
16.67 MByte/s
Voxel R-CNN (Car)
28 MByte
3 seconds
9.33 MByte/s
CaDDN (Mono)
774 MByte
3 seconds
248 MByte/s
NOTE 1: The transfer learning does not have a very high requirement for transmission latency since it is not a real-time inference service, hence it assumes the model transmission via direct device connection should be finished in 3 seconds. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7 Distributed/Federated Learning by leveraging direct device connection | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.1 Direct device connection assisted Federated Learning | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.1.1 Description | In many circumstances, an application server holding a Federated Learning (FL) task has a transmission delay requirement and limited FL coverage. An FL coverage means an area in which UEs the Application server can organize for federated learning.
An Application server has a transmission delay requirement for each FL member (UE). Some of UEs are actually holding valuable dataset but cannot fulfil transmission delay requirement, which leads to a decreasing of FL performance. However, if a UE’s direct network connection cannot fulfil the transmission delay requirement (i.e. an QoS on Uu), leveraging the devices with direct connections helps to involve more UEs holding valuable dataset for the FL task with the following case study:
A UE-A with bad transmission condition sends a UE’s training result to UE-B via direct device connection. In such case, a UE-B aggregates the training result locally and provides to UEs an update of training model for next round.
Some research e.g. in [6][7] have illustrated the increasing performance in subcase-B (we call it “decentralized averaging methods”). In order to include more devices to participate in FL and to reduce the devices’ reliance on the PS, the authors in [7] uses decentralized averaging methods to update the local ML model of each device. In particularly, using the decentralized averaging methods, each device only needs to transmit its local ML parameters to its neighboring devices. And the neighboring devices can use the acquired ML parameters to estimate the global ML model. Therefore, using the decentralized averaging methods can reduce the communication overhead of FL parameter transmission.
Figure 7.1-1 FL with decentralized averaging method outperforms the original FL
To show the performance of decentralized averaging method, the [6] implemented a preliminary simulation for a network that consists of one BS that is acted as an application server and six devices, as shown in Figure-1. In Figure-1, the green and purple lines respectively represent the local ML parameter transmission of original FL and the FL with decentralized averaging methods. Due to the transmission latency requirement, only 4 devices can participate in original FL. For the FL with the decentralized averaging update method, 6 devices can participate in the FL training process since the devices which are out of coverage can connect to their neighbouring devices (i.e. Device a and Device b) for model updating.
From Figure-1, we can see that the FL with decentralized averaging method outperforms the original FL in terms of identification accuracy. Specifically, the original FL (without using direct device connection) has an upper limit of identification accuracy to about 0.85, while using direct device connection for decentralized averaging method helps to increase the identification accuracy to about 0.88 which is actually a big optimization since the line already goes smoothly after 200 round of FL training.
Besides, the FL leveraging direct device connection can also reduce the energy consumption for some devices since it only needs to transmit its ML model parameters to device instead of the BS. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.1.2 Pre-conditions | Figure 7.1-2 two UEs performs decentralized FL using Direct Device connection
As depicted in Figure-2, there is an Application server for federated learning which needs to communicate with the UEs in a FL coverage for FL task.
To achieve an ideal performance (i.e. fast convergence and high model accuracy), there is a transmission latency requirement to each FL member UE’s data transmission.
Alice and Bob are FL members but their cell phones sometimes have bad signal condition which cannot transmit data to FL service directly. Meanwhile, Bob is willing to support the “decentralized averaging method” (as described in clause 7.1.1) service for its neighbouring cell phones.
Alice, Bob are neighbouring to each other within a FL coverage. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.1.3 Service Flows | 1. Alice is a FL member and already acquires the global AI/ML model from the Application server for FL task, later on when Alice moves to a tunnel with bad signal condition, Alice cell phone’s with direct device connection with her neighboring cell phone cannot transmit model data to its Application server anymore.
2. In the tunnel, Alice discovers Bob, who is neighboring to Alice, a FL member and willing to activate the “decentralized averaging method” service. Thus, Alice requests Bob to establish a direct device connection so that Alice can transmit the AI/ML model training result to Bob.
3. Bob updates the AI/ML model based on Alice’s training result and Bob’s local training result. And Bob sends the updated AI/ML model to Alice for further training.
When Bob moves to a good coverage and is able to transmit the AIML training model (e.g. after several rounds of AIML model parameters exchange between Alice and Bob have been done), Bob transmits the training result to Application server to assist the Application server to perform a global model updating. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.1.4 Post-conditions | By leveraging direct device connection, Alice and Bob can keep the model training of a FL task even when they are under a bad network coverage. And the training result between Alice and Bob can be further uploaded to Application server for global model updating.
Thanks to leveraging direct device connection, it helps FL to be performed even when no communication availability to FL server. Such use case helps to optimize the FL performance. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.1.5 Existing features partly or fully covering the use case functionality | In 3GPP TS 22.261 [8] v18.6.1 clause 6.40.2
Based on operator policy, the 5G system shall be able to provide means to allow an authorized third-party to monitor the resource utilisation of the network service that is associated with the third-party.
NOTE 1: Resource utilization in the preceding requirement refers to measurements relevant to the UE’s performance such as the data throughput provided to the UE.
Based on operator policy, the 5G system shall be able to provide an indication about a planned change of bitrate, latency, or reliability for a QoS flow to an authorized 3rd party so that the 3rd party AI/ML application is able to adjust the application layer behaviour if time allows. The indication shall provide the anticipated time and location of the change, as well as the target QoS parameters.
Based on operator policy, 5G system shall be able to provide means to predict and expose predicted network condition changes (i.e. bitrate, latency, reliability) per UE, to an authorized third party.
Subject to user consent, operator policy and regulatory constraints, the 5G system shall be able to support a mechanism to expose monitoring and status information of an AI-ML session to a 3rd party AI/ML application.
NOTE 2: Such mechanism is needed for AI/ML application to determine an in-time transfer of AI/ML model.
Subject to user consent, operator policy and regulatory requirements, the 5G system shall be able to expose information (e.g. candidate UEs) to an authorized 3rd party to assist the 3rd party to determine member(s) of a group of UEs (e.g. UEs of a FL group). |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.1.6 Potential New Requirements needed to support the use case | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.1.6.1 Functional requirement | [P.R.7.1-001] Based on user consent and operator policies, the 5G system shall be able to configure a group of UEs who participate in the same service group (e.g. for the same AI-ML FL task) to establish communication with each other via direct device connection e.g. when direct network connection cannot fulfil the required QoS.
[P.R.7.1-002] Based on user consent, operator policies and the request from an authorized 3rd party, the 5G system shall be able to dynamically add or remove UEs to/from the same service (e.g. a AI-ML federated learning task) when communicating via direct device connection. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.1.6.2 KPI requirement for direct device communication | The 5G system shall be able to support the following KPI for direct device connection as defined in Table 7.1.6-1
NOTE: The table refers to a typical AI/ML model for image recognition i.e. a 7-bit CNN model VGG16_BN using 2242243 images as training data) [2].
Table 7.1.6-1: Latency and user experienced UL/DL data rates for uncompressed Federated Learning
Model size
(8 bit VGG-16 BN) (see NOTE 2)
Mini-batch size
(images)
Maximum latency for trained gradient uploading and global model distribution (see NOTE 1)
User experienced UL/DL data rate for trained gradient uploading and global model distribution (see NOTE 2)
132 Mbyte
64
3.24s
325Mbit/s
32
1.9s
55Mbit/s
16
1.3s
810Mbit/s
8
1.1s
960Mbit/s
4
1.04s
1.0Gbit/s
NOTE 1: Latency in this table is assumed 20 times the device GPU computation time for the given mini-batch size.
NOTE 2: Values provided in the table are calculative needs for an 8-bit VGG16 BN model with 132MByte size [2] |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.2 Asynchronous FL via direct device connection | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.2.1 Description | Federated Learning (FL) is an important machine learning service. Due to the Synchronous FL [8], Sync-FL, requires a strict communication quality for each UE in order to get all the intermediate results to FL server in time, the Synchronous FL is sometimes vulnerable to the unpredicted wireless condition and the divergence of UEs’ capabilities. Therefore, the Asynchronous FL [9], Async-FL, has been widely used in many circumstances. The main idea of Async-FL is to let UE report its result whenever it is ready and the FL server will refresh the model without waiting for all intermediates results are collected. The Sync-FL and Async-FL have pros and cons as the table 7.2-1 shows.
Table 7.2-1 Comparison of Sync-FL and Async-FL
Sync-FL
Async-FL
Total computation workload
Lower.
Higher. The UE will get a new model for training when it uploads the result without waiting for other UE’s result. So the computation work load in each UE can be increased.
Communication requirement
Higher. All UEs shall report its result before next FL round starts
Lower. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.2.2 Pre-conditions | The direct device connection can be used to realize the Async-FL. As The figure 7.2-1 shows, for some UEs who are in a bad coverage it can use the indirect network connection to communicate with Parameter Server (PS). The communication requirement via indirect network connection can be relaxed i.e. no need to transmitted all UEs training result with a restricted timing.
Figure 7.2-1 Group based Async-FL
For each member UE, it can send its training result to the PS via either direct network connection or indirect network connection, and the PS will send a new model to the member UE without waiting for other Aggregate UEs’ result (i.e. Async-FL). |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.2.3 Service Flows | 1) The Parameter Server (PS) distributes the global model to FL member UEs via direct network connection or indirect network connection; For the UEs in bad coverage, it can use the indirect network connection to perform a Async-FL with PS.
2) When receiving the training result from member UE, the relay UE sends it to the Parameter server immediately to get a new model for the member UE. Due to the relay UE has a limited QoS for its own network connection (PDU session), the relay UE needs to determine the QoS for indirect network connection for each of member UEs based on an aggregated QoS (QoS upper limit) for the group of members served by the relay UE.
3) The Async-FL will be performed until the model accuracy reached a certain threshold. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.2.4 Post-conditions | Thanks to the indirect network connection, the FL server can still use the valuable data stored in UEs who are out of coverage with the method of Async-FL. The model training is finally finished with expected model performance.
The charging for an Remote UE using an Indirect 3GPP Communication will be done.
7.2.5 Existing features partly or fully covering the use case functionality
In TS 22.261 (v19.2.0) clause 3.1
aggregated QoS: QoS requirement(s) that apply to the traffic of a group of UEs.
In TS 22.261 (v19.2.0) clause 6.9
The 5G system shall support different traffic flows of a remote UE to be relayed via different indirect network connection paths.
The connection between a remote UE and a relay UE shall be able to use 3GPP RAT or non-3GPP RAT and use licensed or unlicensed band.
The connection between a remote UE and a relay UE shall be able to use fixed broadband technology.
The 5G system shall be able to provide indication to a remote UE (alternatively, an authorized user) on the quality of currently available indirect network connection paths.
The 5G system shall be able to maintain service continuity of indirect network connection for a remote UE when the communication path to the network changes (i.e. change of one or more of the relay UEs, change of the gNB).
The 5G system shall be able to support a UE using simultaneous indirect and direct network connection mode.
The 5G system shall enable the network operator to authorize a UE to use indirect network connection. The authorization shall be able to be restricted to using only relay UEs belonging to the same network operator. The authorization shall be able to be restricted to only relay UEs belonging to the same application layer group.
In TS 22.261 (v19.2.0) clause 6.40,
Subject to user consent, operator policy and regulatory requirements, the 5G system shall be able to expose information (e.g. candidate UEs) to an authorized 3rd party to assist the 3rd party to determine member(s) of a group of UEs (e.g. UEs of a FL group).
In TS 22.115 (V18.0.0) Clause 4.8 on "Charging Requirements for Indirect 3GPP Communication"
This section describes the requirements to enable operator collection of charging data for an Evolved ProSe Remote UE and Relay UE using an Indirect 3GPP Communication. The requirements also apply in the roaming case.
The 3GPP core network shall be able to collect charging data for an Evolved ProSe Remote UE which accesses the 3GPP core network through an Indirect 3GPP Communication. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.2.6 Potential New Requirements needed to support the use case | [P.R. 7.2-001] 5GS shall be able to support an aggregated QoS for a group UEs served by a relay UE.
[P.R. 7.2-002] 5GS shall be able to provision an aggregated QoS to a relay UE for a group-based service.
[P.E. 7.2-003] Based on 3rd party request and user consent, the 5G system shall be able to expose information (e.g. candidate UEs) to an authorized 3rd party to assist the 3rd party for UE member(s) selection in a group of UEs (e.g. UEs of a FL group), for UEs using direct or indirect network connection. E.g. the 3rd party request may include the expected QoS as a criteria for UE member selection. |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.3 5GS assisted distributed joint inference for 3D object detection | |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.3.1 Description | Distributed joint inference is to leverage multiple nodes (e.g. UEs) to provide inference results so that an aggregation of those inference results can lead to a better performance.
When a 3rd party vehicle wants to obtain relevant information of a certain vehicle 1 (e.g. position, width, length, height, profile, orientation), the data collected by the 3rd party vehicle itself is limited. For example, as shown in figure 7.3.1-1, the 3rd party vehicle, which is directly behind vehicle 1, can only obtain relevant data on the tail of vehicle 1 through sensors, and can identify the width and height of the vehicle 1 through the inference of the local 3D object detection model, but there is no way to know the length of the vehicle 1, or even a more precise vehicle profile, orientation, etc. In addition, although the location of UE1 can be known through equipment such as the radar of the 3rd party vehicle, limited by the singleness of the data, the positioning accuracy based on the information obtained by a single vehicle is limited.
Figure-7.3.1-1: Joint inference among multiple vehicles for 3D object detection
All of the above problems need to be solved through multi-vehicle joint inference. The performance to use the joint inference is shown in the Figure-7.3.1-2. Its clear shows that despite the green vehicle generating false orientations and location by its local model, the global map (i.e., the red box) can correct the orientation and location error for the green vehicle based on the aggregated results of three vehicles (i.e. blue box, green box and yellow box) [23].
Figure-7.3.1-2: distributed joint learning leads to a better inference performance |
3b008e4f4eb4734158412e812ecd3c39 | 22.876 | 7.3.2 Pre-conditions | As shown in figure-7.3.2-1, when a vehicle accident occurs somewhere and the road is congested. Alice's auto-driving vehicle wants to know the complete situation of the accident (i.e. the exact location and shape of the accident vehicle including the length, width and height of the vehicle), so as to use the inference result for auto-driving decision in real time. Alice’s vehicle needs to find and establish connections with vehicles located in different position to the accident vehicle, and collect inference results to perform the accurate 3D object detection of the accident vehicle.
Though the accident vehicle cannot move due to the collision to a barricade afront, the electronic device can still work as normal.
Figure-7.3.2-1: Joint inference among multiple vehicles for accident vehicle detection |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.