|
| Implementing Behavior Analytic Models in Mobile Robots: Where the Rubber Meets the Road (Literally) |
| Sunday, May 29, 2005 |
| 3:00 PM–4:20 PM |
| Waldorf (3rd floor) |
| Area: TPC; Domain: Theory |
| Chair: Betsy J. Constantine (Context Systems) |
| CE Instructor: William R. Hutchison, Ph.D. |
| Abstract: Science fiction has presented many futuristic images of robots that behave and learn like animals and humans. Behavior analysis is the scientific discipline most able to engineer the processes that would be involved in such an artificial learning organism, but behavioral engineering has been applied to animals and humans rather than robots. The papers in this symposium will describe research by Hutchison and Constantine in which computer models of behavior analytic formulations have been implemented in robots. The three papers will describe the practical and conceptual challenges that must be solved to fill the gaps between the environment and behavioral processes; how the behavior of robots can illustrate basic behavior analytic principles such as primary reinforcement, conditioned reinforcement, punishment, stimulus control, transfer of stimulus control, etc.; the value of such work for behavior analysis and practical applications; and major theoretical and philosophical issues raised by this work. Live demonstrations of this work with a research robot will be shown in the second paper. |
| |
| Implementing a Behavior Analytic Model in a Mobile Robot |
| WILLIAM R. HUTCHISON (Behavior Systems) |
| Abstract: This paper will describe some of the issues that arise in implementing a quantitative behavior analytic model in a robot, especially sensory, motor, and consequence issues that do not arise when working with animals. Raw sensory data are far removed from the conceptual level at which behavior analysts typically describe “stimuli”, and the same is true of “responses” that in robots consist of motor movements. We must also look at the basic meaning of “reinforcement” to implement primary and secondary reinforcement in a mechanical being. Solutions to these issues will be described in a complete working computer model of operant learning?the Seventh Generation system?that is currently being used to control robots. |
| |
| Practical Demonstration of a Mobile Robot Controlled by a Behavior Analytic Model |
| BETSY J. CONSTANTINE (Context Systems), William R. Hutchison (Behavior Systems) |
| Abstract: This paper will describe how a robotic implementation of operant behavior can illustrate some of the basic behavioral processes that have been extensively studied with animals. The paper will present one or two cases with detailed descriptions of the sensory, motor, and reinforcement processes involved. The descriptions will be illustrated by live demonstrations with a research robot. |
| |
| Implications of Developing Behavior Analytic Models in Robots |
| WILLIAM R. HUTCHISON (Behavior Systems), David C. Palmer (Smith College), Betsy J. Constantine (Context Systems) |
| Abstract: The implications of implementing behavioral models in robots range from stimulating basic behavior analytic research to the thorny social issue of creating artificial creatures. Complete models of behavior in computer simulations are more useful than isolated equations as a medium for describing the complex set of relationships between organisms and environment. But only when implemented in robots can those detailed models be put to the ultimate tests of sufficiency and accuracy. By studying the design and operation of an operant model in a robot, researchers have an opportunity to reexamine some basic behavior analytic principles, such as primary reinforcement, conditioned reinforcement, punishment, stimulus control, transfer of stimulus control, etc. at a level of detail that is not possible with animal subjects. Developing robots that may become smarter, stronger, faster, etc., than humans raises issues that can no longer be dismissed as hypothetical. Behavior analysts are essential participants in this complex emerging social issue. |
|
| |