16-899: Topics in Robot Motion Planning

Spring 2008 Course Information

Announcements

4/06 : new due date for homework 3 (Mon, 4/14).
3/24 : homework 3 assigned.
2/22 : homework 2 assigned.
1/29 : homework 1 assigned.
1/28 : lecture 1 slides posted.
1/24 : mailing list created.

	Course Essentials Mondays and Wednesdays 12:00PM to 1:20PM NSH 3002
E-mail list: Used for announcements and discussion.
Course calendar:	Schedule of class topics, assignments, readings, and due dates

Instructors

Who	Email	Office Hours	Phone
James Kuffner		4:30-5:30 Thurs @ NSH4228A (and by appointment)	(412)268-8818
Rosen Diankov (TA)	rdiankov _at_ cs	5:00-6:00 Tues/Thurs @ NSH 4208

Course Description

This course will cover an assortment of topics related to robot motion planning and high-dimensional search of configuration spaces. Students should have a general background in AI and discrete search algorithms (Dijkstra's, A*, etc), as well as probability and statistics, and in interest in robot motion planning. The course will consist of programming assignments as well as a final course project. Two industrial ABB manipulator arms, as well as a possible small humanoid robot platform will be made available to students for implementing and testing their planning algorithms.

Syllabus

course intro and background
search-based AI
+ three primary applications of planning in robotics:
1) navigation, 2) manipulation, 3) scheduling
(geometric planning concerns the first two)
discrete vs. continuous representations
+ cell-decompositions and roadmaps
+ potential fields and navigation functions
dimensionality issues and sources of problem difficulty
+ DOFs
+ kinematic and dynamic constraints
+ multiple robots
+ sensing and uncertainty
tradeoffs of solution quality/optimality/feasibility
sampling-based planning
+ randomized / probabilistic techniques
+ deterministic
+ importance sampling (statistical priors)
problem classes
+ single robot, static environment, known start and goal
+ dynamic environments
+ multiple agents
+ replanning loops and safety issues
issues concerning:
+ what can be precomputed a priori
+ what must be searched online
connections to other research topics
+ path/trajectory smoothing / optimization
+ statistical models (connections to machine-learning)
+ estimation of unknowns (Kalmann /particle filtering)
+ planning under partial info/noisy sensors (SLAM)
+ POMDPs
+ information spaces
applications
+ vehicles with dynamics
+ mobile manipulation
+ biped locomotion / humanoids
+ multi-robot coordination