Midterm Exam: Thursday, Oct. 19, in class.

HELP!: I made a numerical error on HW1 for the point assignment for the implementation problem. If you bring your HW back to class, you may get more points.

Office Hours: 2-3pm, TuTh.

Motivation: Planning algorithms are impacting technical disciplines and industries around the world, including robotics, computer-aided design, manufacturing, computer graphics, aerospace applications, drug design, and protein folding. This course is intended for computer scientists and engineers with interests in robotics, artificial intelligence, robotics, control theory, and the connections between them. The course focuses mainly on the modeling, algorithmic, and computational issues that arise when designing autonomous decision makers.

Course Calendar:

Class #	Date	Lecture Topic	Reading	Assignments
Week 1
1	8/24	Course overview	Ch. 1
Week 2
2	8/29	Discrete feasible planning; search algorithms	Sec. 2.1, 2.2
3	8/31	Discrete optimal planning; value iteration	Sec. 2.3
Week 3
4	9/5	Plans of unspecified length; logic-based formulations	Sec. 2.3-2.5	HW 1 Assigned
5	9/7	Brief motion planning background	Ch. 3,4
Week 4
6	9/12	Sampling-based motion planning	Intro text of Ch. 5; Sec. 5.1.1, 5.2.1-5.2.3
7	9/14	Sampling-based motion planning	Sec. 5.4.1-5.4.2, 5.5, 5.6.1
Week 5
8	9/19	Combinatorial motion planning	Sec. 6.1-6.2
9	9/21	Combinatorial motion planning	Sec. 6.1-6.2	HW 1 Due
Week 6
10	9/26	Feedback planning	Sec. 8.1-8.2
11	9/28	Games against nature	Sec. 9.1-9.2
Week 7
12	10/3	Zero-sum games; nonzero-sum games	Sec. 9.3-9.4	HW 2 Assigned
13	10/5	Nonzero-sum games; criticisms of decision theory	Sec. 9.4-9.5
Week 8
14	10/10	Intro. to sequential decision making	Sec. 10.1
15	10/12	Forward projections; backprojections; plan and execution	Sec. 10.1	HW 2 Due
Week 9
16	10/17	Value iteration	Sec. 10.2
17	10/19	Midterm (in class)
Week 10
18	10/24	Policy iteration; Dijkstra-like algorithms	Sec. 10.2
19	10/26	Infinite horizon problems; reinforcement learning	Sec. 10.3-10.4
Week 11
20	10/31	Sensors; history information space	Sec. 11.1
21	11/2	Information mappings; derived I-spaces	Sec. 11.2.1
Week 12
22	11/7	Nondeterministic I-spaces	Sec. 11.2.2
23	11/9	Probabilistic I-spaces	Sec. 11.2.3
Week 13
24	11/14	Derived I-space examples; approximations	Sec. 11.3-11.4
25	11/16	Comparing the power of robots	Jason O'Kane
Week 14
26	11/28	More sensors; sensorless planning; localization	Sec. 11.5
27	11/30	Localization	Sec. 12.1-12.2
Week 15
28	12/5	Environment uncertainty and mapping	Sec. 11.5
29	12/7	Pursuit-evasion, summary	Sec. 12.1-12.2

Final Exam: 7pm Thursday, Dec. 14, 1111 SC

Homework Assignments

• HW1: Assigned: Sep. 5, Due: Sept. 21.
  Solve problems 2.4 (Note: changed from 2.3a), 2.5 (argue the correctness of your answer), 2.10, 2.13, and 2.23. For the last problem, you may use whatever programming language you like. You may want to design your code nicely so that extensions can be made in later assignments. Allow grid-based planning problems, such as the one shown in Figure 2.2, and compare Dijkstra to forward value iteration on a couple of interesting grid-based planning problems. Turn in written (or printed) copies of your work in class.
• HW2: Assigned: Oct. 3, Due: Oct. 12.
  Solve problems 5.6, 5.18, 6.4, 9.5, 9.14, 10.1. Other problems, which might be helpful for exam preparation (don't submit): 5.5, 6.5, 9.6, 9.13, 9.15, 10.2, 10.6. For problem 5.5, you need to express the average dispersion as a function of n. In other words, what is the average dispersion for the first n points of the van der Corput sequence?
• HW3: Assigned: Oct. 31, Due: Nov. 14.
  Solve problems 10.13, 10.18, 11.7, 11.12, 11.16, 10.23. For the last problem, indicate running times, number of iterations, and various other output that help to understand the execution. Illustrate the algorithm on a simple graph example and also for grid problems such as the one shown in Figure 10.1.
• Additional suggested problems: 12.2, 12.3, 12.6, 12.11, 12.13, 12.18

Tentative Topics
(with estimated number of lectures):

• Introduction, motivation (1)
• Discrete planning (basic search algorithms, optimal planning) (2)
• Continuous planning, sampling-based methods (2)
• Continuous planning, combinatorial methods (2)
• Basic decision theory (games against nature, matrix games, criticisms) (3)
• Sequential decision theory (feedback plans, infinite-horizon problems, reinforcement learning) (5)
• Sensors and information spaces (general concepts, sensor models, minimal information requirements, examples) (5)
• Planning in probabilistic information spaces (POMDPs, belief spaces) (2)
• Planning in nondeterminsitic information spaces (2)
• Planning under sensing uncertainy in robotics (localization, map building, pursuit-evasion) (5)

Textbook: Planning Algorithms, S. M. LaValle, Cambridge University Press, 2006. Also available for free download at http://planning.cs.uiuc.edu/. Most of the material will come from Chapters 2 and 9-12.

Course Mechanics

• Midterm (30%), Final Exam (40%), Homeworks (30%)
• An independent project is required for those registered for four credit hours.
• Homeworks will be a combination of written and programming assignments. There are no requirements to use a particular language for programming assignments. There will be four or five homework assignments during the semester.