|
Purdue CS 59300 Robotics(Spring 2023)
This course covers topics in motion planning, estimation, and control to design algorithms for robots to safely interact with their environments and perform various challenging tasks under constraints. The first part of this course focuses on classical techniques such as search-based and sampling-based planning, PID control, Model Predictive Controller (MPC), and Bayesian filters. The second part covers modern deep learning and deep reinforcement learning techniques and their application to planning and decision-making in robotics.
Administrative Information
Instructor: Ahmed H. Qureshi (ahqureshi@purdue.edu) | Office Hours & Location (appointment only): Monday 1:00 - 2:00 pm at HAAS 152
TA: Hanwen Ren (ren221@purdue.edu) | Office Hours: Tuesday 3-4 pm at HAAS 115 or by appointment
Time & Location: TTH 9:00-10:15, LWSN 1106
Course Content
| Sr | Date | Topics (Tentative) | Readings |
| 1 | 1/10 | Introduction to Robotics | |
| 2 | 1/12 | Robot Manipulators (Forward/Inverse Kinematics) | |
| 3 | 1/17 | Motion Planning I: Sampling-based methods I (RRT, PRM) | [1] [2] |
| 4 | 1/19 | Motion Planning II: Sampling-based methods II (RRTConnect, RRT*, B-RRT*) | [1] [2] |
| 1/20 | Assignment 1 (Release) | |
| 5 | 1/24 | Motion Planning III: Advance Constraints (Kinodynamic, Manifolds, etc.) | [1] [2] |
| 6 | 1/26 | Optimal Control I: PID I | Ch6 |
| 7 | 1/31 | Optimal Control II: PID II | Ch6 |
| 8 | 2/2 | Project Discussions | |
| 2/6 | Assignment 1 (Due 11:59pm (EDT)) | |
| 2/7 | Assignment 2 (Release) | |
| 9 | 2/7 | Optimal Control III: MPC I | |
| 10 | 2/9 | Optimal Control IV: MPC II | [1] [2] |
| 11 | 2/14 | Estimation I: Probability Review, Bayes Filtering | |
| 12 | 2/16 | Estimation II: Kalman Filter + Combining Planning, Estimation & Control | [1] [2] |
| 2/18 | Project title and team members (Due 11:59pm (EDT)) | |
| 13 | 2/21 | ML for Motion Planning I: Neural Networks, Backpropagation, Dropout | [1] |
| 2/22 | Assignment 2 (Due 11:59pm (EDT)) | |
| 14 | 2/23 | ML for Motion Planning II: Convolutional Neural Networks, Recurrent Neural Networks (LSTMs) | |
| 2/23 | Assignment 3 (Release) | |
| 15 | 2/28 | ML for Motion Planning III: Variational Autoencoders (VAEs), Generative Adversarial Networks
| [1] [2] [3] |
| 16 | 3/2 | ML for Motion Planning IV:Informed Sampling, Path Generation | |
| 17 | 3/7 | Deep Reinforcement Learning I: MDPs, Value Function, Q-Function | [1] |
| 18 | 3/9 | Deep Reinforcement Learning II: Policy Gradients (Reinforce) | [1] |
| 3/10 | Assignment 3 (Due 11:59pm (EDT)) | |
| 19 | 3/14 | Spring Break | |
| 20 | 3/16 | Spring Break | |
| 3/20 | Project Milestone 1 (Due 11:59pm (EDT)) | |
| 3/21 | Assignment 4 (Release) | |
| 21 | 3/21 | Deep Reinforcement Learning III: Advance Policy Gradients Methods I (A3C, PPO, TRPO) | [1] [2] [3] |
| 22 | 3/23 | Deep Reinforcement Learning IV: Advance Policy Gradients Methods II (DDPG, TD3, SAC) | [1] [2] [3] |
| 23 | 3/28 | Imitation Learning | [1] [2] [3] |
| 24 | 3/30 | Inverse Reinforcement Learning I: Maximum Entropy(MaxEnt)-IRL | |
| 3/31 | Assignment 4 (Due 11:59pm (EDT)) | |
| 25 | 4/4 | Final Exam | |
| 26 | 4/6 | Inverse Reinforcement Learning II: GANs & IRL | |
| 27 | 4/11 | Deep Reinforcement Learning for Planning | |
| 28 | 4/13 | Closing Remarks, Discussions & Open Research Problems | |
| 4/17 | Project Milestone 2: project presentation files (Due 6:00pm (EDT)) | |
| 29 | 4/18 | Project Presentations I | |
| 30 | 4/20 | Project Presentations II | |
| 31 | 4/25 | Project Presentations III | |
| 32 | 4/27 | Project Presentations IV | |
| 5/2 | Project Milestone 3: Project Report, Demonstration Videos, Codes, Trained Models & Datasets(Due 11:59pm (EDT)) | |
|
|
Prerequisites
Familiarity with Data Structures & Algorithms, and a background in AI & Statistical Estimation is required. The interested undergraduate students should meet the following prerequisites:
Textbooks
There are no specific textbooks. However, some helpful reference books include:
Principles of Robot Motion (Theory, Algorithms, and Implementations) by Howie Choset, Kevin M. Lynch, Seth Hutchinson, George A. Kantor, Wolfram Burgard, Lydia E. Kavraki and Sebastian Thrun.
Reinforcement learning: An introduction by Richard S. Sutton, and Andrew G. Barto. MIT press, 2018.
Deep Learning by Goodfellow, Ian, Yoshua Bengio, and Aaron Courville. MIT press, 2016.
Grading Policy
There are no midterm and final exams.
| Assignments (4 X 10%) | 40% |
| Final Exam | 23% |
| Project | 35% |
| Class Participation | 2% |
|
|
| Project total | 35% |
| Milestone 1: Planning/Control component + Project proposal | 10% |
| Milestone 2: Learning component + Initial results + Project presentation | 10% |
| Milestone 3: Full Project Report + Demo videos + Code with documentation | 15% |
|
|
Project: Students can do solo project or as a team of two (max).
Project proposal and final report template:https://www.overleaf.com/latex/templates/ieee-journal-paper-template/jbbbdkztwxrd
Late Submission Policy: All deadlines are firm unless notified in advance. Late submissions can only be accepted within next 48hr (1second==48hr) of the deadline but will result in a straight 25% off.
Academic Integrity
This course defaults to Purdue standards on intellectual integrity and academic conduct. Therefore, students are responsible for reading the following pages and comply with them throughout this course.
|