Week 03 | SKILL-023
Week 3 Summary: Mastering Inverse Kinematics
Published: April 13, 2026 | Author: Smartotics Learning Journey | Reading Time: 8 min
This week we tackled one of the most important and challenging topics in robotics: Inverse Kinematics (IK). We went from “what is IK?” to building complete solvers and understanding how they power real robots. Here’s everything we covered.
The Week at a Glance
| Day | Topic | Key Concept |
|---|---|---|
| D01 | IK Series Start | The inverse problem, workspace, multiple solutions |
| D02 | Analytical Solutions | Closed-form IK for 2-DOF and 3-DOF arms |
| D03 | Jacobian Method | Velocity-based IK with pseudoinverse |
| D04 | Numerical Methods | Newton-Raphson, CCD, and convergence |
| D05 | Redundancy & Optimization | Null space, joint limits, obstacle avoidance |
| D06 | Real-World Applications | Industrial arms, humanoids, surgical robots |
Core Concepts Recap
The IK Problem
Forward kinematics asks: “Given joint angles, where is the end-effector?” We learned this in Week 1.
Inverse kinematics asks the reverse: “Given a desired end-effector position, what joint angles achieve it?” This is harder because:
- The mapping may not be invertible (singularities)
- Multiple solutions may exist (elbow-up vs. elbow-down)
- No solution may exist (target out of reach)
Three Families of Solutions
-
Analytical (closed-form): Exact, fast, but only works for specific geometries. Uses trigonometry and the law of cosines. Best for: 2-DOF and 3-DOF planar arms, robots with spherical wrists.
-
Jacobian-based (numerical): Iterative, general-purpose. Uses the relationship between joint velocities and end-effector velocities. The pseudoinverse handles non-square Jacobians. Damped least squares adds singularity robustness.
-
Heuristic (CCD, FABRIK): Simple, robust, great for real-time. CCD optimizes one joint at a time. FABRIK works with positions instead of angles. Best for: animation and games.
Redundancy as a Superpower
When a robot has more DOF than the task requires, the extra freedom lives in the null space. We can project secondary optimization objectives (joint limit avoidance, obstacle avoidance, singularity avoidance) into this space without affecting the primary task.
This is why 7-DOF arms are popular — that one extra DOF provides enormous practical value.
Code We Built This Week
# Our IK toolkit (summary of all implementations)
# 1. Analytical IK for 2-DOF planar arm
ik_2dof(px, py, L1, L2, elbow='up')
# 2. Analytical IK for 3-DOF with decoupling
ik_3dof(px, py, phi, L1, L2, L3, elbow='up')
# 3. Jacobian pseudoinverse IK
ik_jacobian(target, q_init, L1, L2, alpha=0.5)
# 4. Damped Least Squares IK
ik_damped_ls(target, q_init, L1, L2, lam=0.1)
# 5. Newton-Raphson with line search
ik_newton_raphson(target, q_init, lengths, q_min, q_max)
# 6. Cyclic Coordinate Descent
ik_ccd(target, q_init, lengths)
# 7. Redundancy-aware IK with null space optimization
ik_redundant(target, q_init, lengths, q_min, q_max, obstacle_pos)
These seven functions form a complete IK toolbox. In practice, you’d use libraries like:
- Robotics Toolbox (Peter Corke): MATLAB/Python, excellent for learning
- Pinocchio: C++ with Python bindings, used in research
- KDL (Orocos): Production-grade, used in ROS
- MuJoCo / Drake: Physics simulators with built-in IK
- MoveIt!: ROS motion planning framework
What Makes IK Hard in Practice
Understanding the math is one thing. Deploying IK in real robots adds challenges:
- Real-time constraints: Industrial controllers solve IK at 1kHz — every millisecond counts
- Joint limits and collisions: Every solution must be physically feasible and safe
- Dynamic obstacles: The environment changes while the robot is moving
- Multi-task coordination: A humanoid must balance while reaching
- Precision vs. speed: Higher precision needs more iterations, but you have a time budget
Looking Ahead: Week 4 Preview
Next week we’ll explore Dynamics and Control — the physics of robot motion. Topics include:
- D01: Introduction to Robot Dynamics — forces, torques, and the equations of motion
- D02: Lagrangian Mechanics for Robots — energy-based formulation of dynamics
- D03: Newton-Euler Recursive Algorithm — computing forces from base to tip
- D04: Trajectory Planning — smooth paths between waypoints
- D05: PID and Computed Torque Control — making robots move accurately
- D06: Force Control and Impedance — robots that feel and respond to contact
- D07: Week 4 Summary + Robotics Learning Roadmap
If Week 1 was “where is the robot?” and Week 3 was “how do I get there?”, Week 4 answers “how do I get there smoothly and accurately while dealing with physics?”
Your IK Journey Checklist
Before moving on, make sure you can:
- Explain the difference between forward and inverse kinematics
- Derive the analytical IK solution for a 2-DOF planar arm
- Describe what the Jacobian represents and why its pseudoinverse solves IK
- Explain why damping is needed near singularities
- Compare Newton-Raphson and CCD — when would you use each?
- Explain the null space and why redundancy is useful
- Describe at least two real-world applications of IK
If you can check all these boxes, you have a solid foundation in inverse kinematics. Congratulations — this is one of the most mathematically dense topics in introductory robotics, and you’ve conquered it.
See you in Week 4! 🤖