Teaching

A master's-level lab course on the algorithmic foundations of intelligent robotics, organised around the theme 'From Search to Learning.' Students build a full autonomy stack in MuJoCo using the course's TAMPanda library, treating navigation, manipulation, and task planning as search and learning problems — progressing from A* navigation through PDDL and sampling-based motion planning to a full task-and-motion-planning pipeline and deep reinforcement learning, then a final project that reimplements a recent ICAPS/RSS/CoRL/NeurIPS paper. Co-taught with Ulzhalgas Rakhman.

Topics · Search (A*) · PDDL/STRIPS Planning · Motion Planning (RRT/PRM) · Task and Motion Planning · Deep Reinforcement Learning (PPO, HER) · Goal-Conditioned RL · MuJoCo

Built on TAMPanda ↗

This lab course teaches the foundations of algorithmic planning techniques relevant to robotics applications, spanning motion planning, path planning, and task planning. Students work in small groups on practical exercises during the lecture period and propose their own integrated planning project for the final phase.

Topics · Kinematics & Coordinate Frames · Probabilistic Motion Planning (RRT) · Navigation Planning · STRIPS/PDDL Planning · Generalised Planning · ROS2 Integration · LLMs in Planning · Integrated TAMP

This course introduces students to the foundations and practical applications of Task and Motion Planning (TAMP) for robotics. Students gain hands-on experience in ROS2-based robot software development, study core TAMP approaches such as PDDLStream, and apply these concepts to implement and evaluate their own TAMP solutions in realistic scenarios.

Topics · Task and Motion Planning (TAMP) · ROS2 · Configuration Space · Motion Planning (RRT) · PDDL · PDDLStream · Long-horizon Problem Solving