# Understanding Robot Kinematics

In this section, we'll build intuition for robot kinematics through a concrete, worked example. Kinematics describes the mathematical relationship between joint angles and end-effector positions - given the joint configuration, where does the robot's hand end up? We'll explore this fundamental concept by walking through a simplified but representative case.

Let's examine how traditional robotics approaches robot control using a specific example you can follow step by step.

## From Complex to Simple: The SO-100 Example

We'll start with a familiar robot platform and systematically simplify it to isolate the core kinematic principles without unnecessary complexity.

The SO-100 arm simplified to a 2D planar manipulator by constraining some joints.

The SO-100 is a 6-degree-of-freedom (6-DOF) robot arm. To understand the principles, let's simplify it to a **2-DOF planar manipulator** by constraining some joints.

## The Simplified Robot

To keep the math clear, our simplified robot has:
- **Two joints** with angles θ₁ and θ₂  
- **Two links** of equal length *l*
- **Configuration** q = [θ₁, θ₂] ∈ [-π, +π]²

## Forward Kinematics: From Joints to Position

**Question:** Given joint angles θ₁ and θ₂, where is the end-effector?

**Answer:** We can calculate the end-effector position mathematically:

$$p(q) = \begin{pmatrix} l \cos(\theta_1) + l \cos(\theta_1 + \theta_2) \\ l \sin(\theta_1) + l \sin(\theta_1 + \theta_2) \end{pmatrix}$$

This is called **Forward Kinematics (FK)** - mapping from joint space to task space.

> [!TIP]
> **Understanding the Math:** This equation comes from basic trigonometry! 
> - First link: endpoint at $(l \cos(\theta_1), l \sin(\theta_1))$
> - Second link: starts from first link's end, rotated by $\theta_1 + \theta_2$
> - Final position: sum of both link contributions
>
> **Why it matters:** For a simple robot, FK is relatively easy - given joint angles, we can always compute where the robot's hand is. More complex robots - for instance, dexterous hands, are more challenging to model.

## Inverse Kinematics: From Position to Joints

Now let’s try to invert the mapping: given a desired hand position, what joint configuration achieves it?

**Question:** Given a desired end-effector position p*, what joint angles should we use?

**Answer:** This is **Inverse Kinematics (IK)** - typically more intricate to solve, as it deals with the Jacobian matrix of the forward kinematics function!

We need to solve: $p(q) = p^*$

In general, this becomes an optimization problem:
$$\min_{q \in \mathcal{Q}} \|p(q) - p^*\|_2^2$$

> [!WARNING]
> **Why IK is Hard:**
> - **Multiple solutions** - Same end position can be reached with different joint angles
>
> - **Nonlinear equations** - Some functions make analytical solutions difficult
> - **Constraints** - Joint limits and obstacles further complicate the problem
>
> This is why robotics engineers spend so much time on IK algorithms!

## The Challenge of Constraints

*Free to move*

*Constrained by floor*

*Multiple obstacles*

Real robots face **constraints**:
- **Physical limits** - Can't move through the floor
- **Obstacles** - Must avoid collisions  
- **Joint limits** - Finite range of motion

These constraints make the feasible configuration space $\mathcal{Q}$ much more complex!

## Key Insight

Even for this **simple 2-DOF robot**, solving IK with constraints is non-trivial. Real robots have:
- **6+ degrees of freedom**
- **Complex geometries**  
- **Dynamic environments**
- **Uncertain models**

Traditional approaches require **extensive mathematical modeling** and **expert knowledge** for each specific case.

> [!TIP]
> Mental model: FK is a direct calculator (q → p) and is usually easy; IK is a search (p → q) and becomes hard as soon as you add workspace limits, obstacles, or joint constraints. When IK gets brittle, we'll switch to differential reasoning (velocities) in the next step.

## References

For a full list of references, check out the [tutorial](https://huggingface.co/spaces/lerobot/robot-learning-tutorial).

- **Modern Robotics: Mechanics, Planning, and Control** (2017)  
  Kevin M. Lynch and Frank C. Park  
  Chapter 4 provides an in-depth treatment of forward kinematics with extensive examples, while Chapter 6 covers inverse kinematics with both analytical and numerical approaches.  
  [Book Website](http://hades.mech.northwestern.edu/index.php/Modern_Robotics)

- **Introduction to Robotics: Mechanics and Control** (2005)  
  John J. Craig  
  A classic textbook with detailed coverage of robot kinematics, including the Denavit-Hartenberg notation and systematic approaches to solving forward and inverse kinematics problems.  
  [Publisher Link](https://www.pearson.com/en-us/subject-catalog/p/introduction-to-robotics-mechanics-and-control/P200000003519)