CS 536 - Situated Computing

Credits: 3

Course Description

This course covers the fundamental concepts involved in understanding and engineering a closed-loop, sensing, reasoning, and actuating agent. Biological models of sensing and actuation will be discussed and related to modern artificial counterparts. The course consists of three major topic areas: Vision, Brain, and Robotics. It will introduce students to the issues in computer and biological vision, to models of belief representation and modification, architectures for percept processing and reasoning, machine learning for vision, neural networks, path planning, intelligent localization based on visual cues, and to forward and inverse kinematics, intelligent grasping, and the integration of perception and action.

Prereqs: CS-463G or consent of instructor.

Needed Skills

Students should be familiar with topics in artificial intelligence problem solving methods, including logic, search, min-max game theory, and basic knowledge representation schemes. Students should show the mathematical maturity expected of first-year graduate students.

Learning Outcomes

Students will learn a broad base of concepts dealing with the design and implementation of a situated agent capable of sensing and acting intelligently on its world. The core concepts learned will be computer vision, models of intelligent computation in both machines and biological systems, and robotics. Students will be asked to implement a specific agent in the "Bugs World" simulator, based on Braitenburg vehicles. Student designed agents will use vision, reasoning, and control to compete for resources within the simulator, while avoiding predators. A variety of readings in conjunction with lectures will introduce students to the theoretical foundations of artificial, intelligent, agents. Students will understand the central issues in engineering an intelligent system as well as the different philosophical frameworks that have historically been used to address the problem of constructing a rational, situated agent.

CAC Categories

Topic	Core	Advanced
Math Fundamentals	15	5
Data Structures	10	0
Algorithms & Software Design	5	0
Computer Organization and Architecture	3	1
Concepts of Programming Languages	0	0
Social and ethical issues	2	0
Total	35	6

Math Fundamentals (20):

Core (15): Linear Algebra, Control Theory, Probability and Statistics

Advanced (5): Kinematics, Signal Processing

Data Structures (10): Core (10): Knowledge Engineering and Representation (Bayes Nets, Neural Networks)

Algorithms & Software (5): Core (5): Search and Learning algorithms, Multimodal design.

Computer Organization (4):

Core(3): Reactive architecture

Advanced(1): Subsumption and robotic design

Concepts of Programming Languages: none

Social and Ethical Issues (2): Core(2): Socially inspired multi-robot systems. Surviellance and Society.

Oral Communication (presentations)

none

Written Communication

none

Coverage

Theoretical Content: 65%

· Biologically inspired artifical intelligence: developmental psychology, vision and the human brain, motor control. Competing mechanical architectures.

· Computer Vision: Projective Geoemtry, Image formation process, multi-view systems, object recognition (challenges of), object tracking.

· Control and Kinematics: Linear control theory, path planning, multi-agent systems.

· Decision and Knowledge: representation, probability theory and control.

Problem Analysis: 15%

· Problem formulation related to vision, control, and robotics. Discuss and debate competing theories (nativist versus developmentalist, for example)

Solution Design: 17%

· Design and implementation of a robotic simulation agent. Agent involves practical implementation of computer vision, robotics, and planning systems embedded in architecture. Simulated robot then "competes" for resources in a graphics world.

Other: 3%

Course administration, social and current resesarch discussion, exams.

Student Evaluation and Feedback

Students are evaluated on their work (homeworks, exams). Students receive back their homework and exams. These papers are marked to indicate problems and they point out correct or better solutions. Problems that turn out to be especially difficult are discussed in class during lectures or during recitations.

NOTE: Since the course is open to both undergraduate and graduate students, different grading schemes are used for each group. Also, any grade normalization will be done against peer students, i.e. undergraduates will only be normalized with undergraduates, and graduates with graduates.

For the graduate students with score S on the scale from 0 to 100:

S ≥ 93	A
93 > S ≥ 82	B
82 > S ≥ 71	C
70 > S	E (fail)

For the undergraduate students with score S on the scale from 0 to 100:

S ≥ 90	A
90 > S ≥ 80	B
80 > S ≥ 70	C
70 > S ≥ 60	D
60 > S	E (fail)

Course Evaluation Questions

The course has helped me:

37. Gain a better understanding of the relationship between biological agents and those that might be engineered.

38. Understand the basics of Computer Vision.

39. Provided me with a good understanding of kinematics.

40. Introduced me to the concepts behind control and its relationship to sensing (i.e Vision, touch).

41. Allowed me to explore robotic design and learn from the experience.

42. Better understand the different approaches to developing an autonomous agent.

43. Improved my programming skills by providing me with a challenging problem with real time constraints, competing goals, and complex software design.

Possible Textbooks

There is no single textbook that covers the topic sufficiently. A collection of readings and course notes is the primary source of material for the course. Possible supplemental reading:

Active Robot Vision: Camera Heads, Model Based Navigation and Reactive Control,
H. I. Christensen (ed.), World Scientific Pub Co; 1993.

Vision, Brain, and Cooperative Computation.
Michael Arbib and Allen Hanson (eds). MIT Press, 1987.

The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents.,
Luc Steels and Rodney Brooks, Lawrence Erlbaum Assoc., 1995.