Fall 2017: ENGR-E599
-- Special Topics in Autonomous Robotics Planning and Learning
- Instructor: Prof. Lantao Liu (firstname.lastname@example.org)
- Time: Tu/Th 11:15am-12:30pm
- Location: Sycamore Hall (SY) 103
- Office hours: TBD
The syllabus can be found here
Note: this page and the syllabus may be updated from time to time, so check back frequently.
No particular course is a prerequisite. However, having taken courses of Algorithms, or Linear Algebra, or Probability Analysis will make your learning process faster. Familiarity with C/C++ or python and GNU/Linux environment will be beneficial. Enrollment is open to all graduate students and certain senior undergraduate students (undergraduate students will need the instructor's approval).
None. But there will be a term project.
This is a graduate level research seminar that covers a selected set of important and popular robotics planning and learning topics. Students need to read, present, and discuss selected articles, and do class projects on real robots. We will draw on current research in robotics, machine learning, and a broader sense of AI.
What you will learn in the course
- Understand the fundamental concepts such as space, actions, and classic control laws; Know popular sensors used on autonomous systems; Learn popular open-source libraries and simulators for processing sensing data and performing basic simulations.
- Learn state-of-the-art robotics techniques including data-driven planning, decision-making, state estimate (e.g., SLAM), deep and reinforcement learning, networked multi-robot coordination, etc.
- Hone skills of working on physical robotic systems, such as turtlebots, flying drones, and other choices of field robots. The robot operating system (ROS) in Linux environment will be used.
Longer version of course description
Autonomous robots can be seen everywhere now: self-driving cars are being tested on roads; unmanned drones can tape high quality videos and are under development for package delivery; marine robots have been deployed to explore and monitor vast oceans; service robots are used to vacuum houses and guide customers, etc. The robot industries are undergoing a “boom” period, with numerous start-ups established globally every year.
This course integrates a seminar-style survey of problems and methodologies to advanced autonomous robots with developing a term project to demonstrate them. We will focus on a selected set of popular planning and learning topics that have been successfully applied on autonomous systems. An important goal of this course is to first guide you to the state-of-the-art, and then help you do a research project of your interest, allowing you to understand the concepts in depth. Each student will write a paper style report of publishable quality. After some introductory lectures that aim at describing important concepts and approaches, most of the course will consist of presentation and discussion of recent research articles. Consequently, another important goal of this course is to advance your critical thinking and presentation skills.
Selected papers that need to be read and discussed will be listed in the course website. Students are expected to read the assigned readings before each class. In each class we will have one presenter (or two) to do the presentation and lead the discussion. The length of each presentation should be around 10~15 minutes (ideally 12 minutes), and the presentation should cover a few points such as
- What is the problem? Why it is important? (Introduction)
- How does the proposed method work? (Description)
- What are the strengths and limitations of the approach? Can we do better? (Discussion)
- What is the take-home message? (Conclusion)
For the presenter, prepare PowerPoint (or Keynote or Beamer) slides for your presentation. To design attractive and effective presentation slides, videos and any other interesting supporting material are welcome to be included. The presenter needs to arrive at class at least 10 minutes earlier to make sure that the slides work well on the projector in the classroom. Send me a copy of your slides at least two days before your presentation, and I will try to give you feedback as quickly as I can. It will also be nice to make paper copies of your slides (4-6 slides/page) to hand out to the class before your presentation.
For each paper to be discussed, each student (whether presenting or not), should bring to class their version of the paper (either printed hardcopy or electronic copy), including notes that they have made to themselves. They should have notes, including several questions and critiques which they can contribute to the discussion. The best way to have a good participation grade is to have serious, thoughtful, and insightful criticism of the paper. This can be reflected from how much time you have spent in reading and thinking about the paper.
In summary, for each class, you should bring:
- Noted printouts or electronic versions of the papers being discussed;
- A presentation if you are presenting/leading the discussion of a paper;
- Your personal thoughts and questions regarding the papers you read.
Each student is required to complete a term project. You may replicate one of the topics/methods we have discussed in the class, extend it and evaluate it with your own thoughts. Or you may investigate a new problem and use your learned techniques to solve it. You are encouraged to develop a completely new approach to tackle old or new challenging problems.
The chosen project topic requires instructor's approval, so each student will need to consult with me before making a decision on the topic. You are encouraged to select a topic that fits well with your background and research interests, so talk to your research advisor to narrow down the potential topics. Finally, every student will submit a final report in the research paper style/tone and give a final project presentation by the end of the course.
Teaming work (2-3 people in a team, depending on the “size” of the project) is absolutely acceptable. In fact collaboration is encouraged. Specifically, during the course students are encouraged to work with each other, exchanging thoughts/understanding of the papers, explaining your own ideas and making suggestions to others. Many good ideas are inspired from discussions.
- (5%) Class attendance
- (15%) Participation in discussions
- (30%) Class presentation
- (50%) Final project, presentation, and report
Hardcopies of textbooks are not required. Most books listed below have electronic versions and can be found either online or in the university library.
- Stuart Russell and Peter Norvig. Artificial Intelligence: A Modern Approach, Third edition. Prentice-Hall, 2010.
- Probabilistic Robotics. Sebastian Thrun, Wolfram Burgard, and Dieter Fox. 2005.
- Richard Sutton and Andrew Barto, Reinforcement Learning: An Introduction, MIT Press, 1998.
- Christopher Bishop. Pattern Recognition and Machine Learning, Second edition. Springer, 2006.
- Gaussian Processes for Machine Learning, Carl Edward Rasmussen & Christopher Williams. 2005.