Authors: [tex2html_wrap4184]J. Zelek, M. D. Levine
Investigator username: levine
Subcategory: active perception
Biological creatures apparently execute many tasks in the world by using a combination of routine skills, without doing any extensive reasoning. Brooks (MIT) has used such behaviors in his subsumption architecture as a building block for developing intelligent robots. This was in sharp contrast to the traditional robotics approach in the 1970's when robot processing was functionally decomposed into sequential processes of sensing, modelling, planning, and acting. The major problems with Brooks' architecture are its scalability to more difficult tasks which may include reasoning, and the limitations imposed by not having an internal model. The subsumption architecture did not have any internal model and thus was prone to possible inescapable cyclic behavior. Other researchers (e.g. Arkin:90) have incorporated a global planner. However, in this case, the role of the behavioral architecture has been reduced to a purely reactive one while executing a sequence of linear piecewise path segments.
The intention of this study is to design an architecture that allows the behavioral control strategy to be more flexible, generalizable, and extendable. The component dedicated to behavioral activities should be able to attempt tasks with or without a reasoning module. We are investigating 2D navigational tasks for a mobile robot possessing sonar sensors and a controllable TV camera mounted on a pan-tilt head. The major aspects of our proposed behavioral architecture are as follows: (1) A natural language lexicon is used to represent spatial information and for defining task commands. The lexicon is used as a language for internal communications and user-specified commands. The task is to go to a location in space, either known or determined by locating a specific object. (2) An extension of a formalism, referred to as teleo-reactive (T-R) programs (Nilsson:94), is used for specifying behavioral control. The extensions of this approach involve dealing with real-time resource limitations and constraints. The formalism permits the "personality" of the robot (i.e., the set of behaviors) to be programmed in a similar fashion as conventional programming (i.e. with parameter passing and binding, hierarchy, and recursion). (3) Biologically plausible potential fields are used to represent a model of the world and the current navigational and camera movements. The potential field encodes the task goals and is continually updated with sensed features from the real world. The potential field representation is used to drive the actuators. Experimentation with an implementation of this system will help determine the tradeoffs and limitations of the architecture in various contextual settings.