CSE525
| Course | CSE525 |
|---|---|
| Title | Robotics |
| Credits | 3 credits |
| Course Coordinator |
Dimitris Samaras |
| Description |
Building a robot that operates robustly in an unstructured environment has been a long-standing goal of artificial intelligence. This course introduces the fundamental concepts in robotics, covering the core computational tools for robot perception, learning, control, and planning. The material covered in this course ranges from classic foundational tools in discrete search, robot kinematics, geometric perception, motion planning, and motion control, to modern machine learning based techniques including reinforcement learning and large-scale supervised learning of robot action policies. Evaluations will comprise a mix of programming assignments with simulated robots and in-class written exams. 3 credits, grading ABCFB |
| Course Outcomes | |
| Textbook |
No required textbook; Recommended texts:
|
| Major Topics Covered in Course |
Week 1: discrete planning — BFS, DFS, Dijkstra, A* Week 2: Kinematics pt 1 — coordinate transforms, 3D rotations Week 3: Kinematics pt 2 — forward/inverse kinematics Week 4: Robot vision — cameras, point cloud registration, partial views, outliers Week 5: Motion planning — C-space, graph-based search, sampling-based search Week 6: Motion control — velocity control in joint space and in task space Week 7: intro to deep learning (FCNNs, CNNs, transformers) and imitation learning Week 8: MDPs, dynamic programming, policy gradients Week 9: Actor-critic, sim2real RL, reward design for robotics Week 10: Robot vision with deep learning — object recognition and segmentation, perceptual representations for manipulation Week 11: task planning — STRIPS, PDDL Week 12: integrated task and motion planning — SeSamE, learned models for planning Week 13: hierarchical planning with LLMs |
| Laboratory | |
| Course Webpage |
|