Units: 3-1-8 (H-Level)
Prerequisites: 6.041 and permission of instructor.
Lecture: MW2.30-4 (1-371)
Lab: To be arranged
Description: Theory and application of probabilistic techniques for autonomous mobile robotics. Topics include probabilistic state estimation and decision making for mobile robots; stochastic representations of the environment; dynamic models and sensor models for mobile robots; algorithms for mapping and localization; planning and control in the presence of uncertainty; cooperative operation of multiple mobile robots; mobile sensor networks; application to autonomous marine (underwater and floating), ground, and air vehicles. J. J. Leonard
Links to class materials (restricted to enrolled students, MIT certifcates required):
Here's what I'd like to do: instead of lecture, I'd like to arrange one-on-one meetings with each of you during our usual lecture time, and then to ask you to join me in attending Larry's talk from 4-5pm. If you have a conflict for 4-5pm Monday, please let me know ASAP?
Here's a proposed schedule for the meetings, the goal is to nail down your project plans. If you planning to work work in a group, let me know your partner(s), we can iterate on the schedule.
2.166 one-on-one Meeting Schedule for Monday 10/27/08, room 32-232
In the 2nd half of the lecture, I'll give an overview of occupancy grid maps (read chapter 9 of Thrun, Burgard and Fox)
Here's some information for the schedule for the next month (see: https://web.mit.edu/2.166/www/handouts/future.html)
Next week: lectures on pose graph SLAM/scan matching and visual SLAM
Nov 3: guest lecture by Nick Roy (try to read chapters 14+15 before then?)
Then on November 5th, we start four weeks with 2 student presentations per class, with the schedule as follows:
(Hordur: I thought we had picked a day, but I didn't write that down..., add to one of the above days?)
I will email you individually soon to diccuss topics - Any questions, let me know.
The goal of the project is to try to produce an implementation of a probabilistic robotic algorithm that would be worthy (if all the stars fell into place, and you hit a "home run") of a conference paper (e.g. a paper for ICRA or IROS). That's probably too ambitious for a class project, but that's what I'd like you to shoot for. I would like you to write a conference paper as the project final report (e.g. see a recent ICRA paper such as Ed Olson's ICRA 2006 for a typical ICRA paper, e.g. 6-8 pages two-column format.
Next, some guidance, in the form a list of questions:
1. What algorithm do you wish to try to develop or implement?
2. What type(s) of robots?
3. What type(s) of sensors?
4. What type of software (e.g. C++ vs. matlab vs. python etc)?
5. Will you use simulation, and if so in what manner?
6. What is the current state-of-the-art? (can you give literature references for similar published work by others?)
7. What would be your "home run" result? What would be the ideal capability/how would you know for sure when you are "done"?
8 What would be a "fallback" result, if things work out to be more difficult than you thought, is there a minimal capability that you could iplement to at least demonstrate the concept?
9. What aspects of your implementation might provide a novel, useful, and/or un-obvious contribution over the state of the art?
10. Is this a solo effort or a team effort? if the latter, with who else are you working?
11. What aspects are out of scope that you would consider to do in the future, parts that you will not get to in the course project.
We have four of the Videre Erratic robots, each with Sick scanners, available as resources to share for the projects.
Please email us the following: