Robotics & State Estimation

A suite of simulated-robotics projects from AI for Robotics (CS 7638), all Python and NumPy from scratch: localize a robot that can't fully trust its sensors (particle filter), track a moving target through noise (Kalman filter), plan a route worth driving (A* with path smoothing), and then actually hold that line (PID).

The estimation half is the humbling half — a particle filter is just guesses that die when they disagree with the evidence, and watching a cloud of hypotheses collapse onto the robot's true position never stops being satisfying. The control half is the stubborn half: proportional gets you close, integral remembers the error you'd rather forget, and derivative keeps the whole thing from oscillating into the wall. Tuning gains by hand builds a respect for controls engineers that no lecture can.

Course runs Summer 2026; the suite grows as the semester does.

• PARTICLE-FILTER LOCALIZATION
• KALMAN-FILTER TRACKING
• A* WITH PATH SMOOTHING
• PID CONTROL LOOPS
• PYTHON + NUMPY, FROM SCRATCH