|
| Implementing Behavioral Models in Robots: What a Learning Robot Can Teach Us |
| Friday, May 27, 2005 |
| 6:00 PM–9:00 PM |
| 4K (4th floor) |
| Area: EAB; Domain: Basic Research |
| CE Instructor: William R. Hutchison, Ph.D. |
| WILLIAM R. HUTCHISON (Behavior Systems), BETSY J. CONSTANTINE (Context Systems) |
| Description: The workshop will introduce participants to robot technology and its uses in behavioral research. Participants will be given an overview of robot technology, including sensor types, sensor preprocessing, motors and actuators, and alternative approaches to controlling robot motors. Instructors will present an existing quantitative/computational model of operant learning?the Seventh Generation system?that is currently being used to control robots capable of operant conditioning. By studying the design and operation of the operant model interacting directly with the real world, participants will have an opportunity to reexamine some basic behavior analytic principles, such as primary reinforcement, conditioned reinforcement, punishment, stimulus control, transfer of stimulus control, etc. Participants will explore these principles by working with an actual robot whose behavior is learned and controlled by the Seventh Generation behavioral model. After observing demonstrations of a variety of teaching procedures and the effect each has on the robots learning behavior, participants will analyze those processes at a level of detail unachievable with living subjects. Participants will have a hands-on opportunity to develop a simple teaching procedure with the Seventh Generation operant learning system and use it to teach a small mobile robot?the Garcia robot from Acroname, Inc.?to perform a simple behavior. |
| Learning Objectives: At the conclusion of the workshop, the participants will be able to: - Describe basic learning processes at a level appropriate for control of a robot. - Describe key issues and problems in implementing behavior analytic models in robots. - Discuss parallels and differences between teaching robots and teaching living organisms. - Describe several areas of current research in robotics and machine learning relevant to behavior analysis. - Develop a simple example of computerized training and an example of live training for a Garcia robot with the Seventh Generation operant learning system. |
| Activities: Classroom presentation and discussion. The class will break into 2 or 3 lab groups after each topic, each with a robot and instructor to demonstrate the topic. By the last stage of the workshop, the groups will be able to demonstrate a simple training procedure with the robot. |
| Audience: Psychologists and behavior analysts interested in computational models of learning and how working with robots can improve our teaching procedures. Behavior analysts who are interested in developing computational behavioral models and robotics are particularly welcome. |
| Content Area: Theory |
| Instruction Level: Advanced |