Parallel distributed processing: explorations in the microstructure of cognition, vol. 1
Creating a Memory of Causal Relationships: An Integration of Empirical and Explanation-Based Learning Methods
Knowledge Structures
| Hi-index | 0.01 |
Recent advances in autonomous robotic control may be applied to the problem of designing intelligent, mobile agents for cyberspace. This paper examines the field of behavior-based reactive systems (BBRS) as used in robotics that may be applicable in constructing intelligent agents. The problem of search in an unstructured and dynamic, distributed environment has been explored in a physical robot. Two major aspects -- the use of chaos to generate stochastic path selection, and a unique associative learning mechanism that bounds the search space on subsequent runs -- have been shown to be an efficient strategy for the robot in exploring a nonstationary environment. The role of chaos is to produce novel, yet constrained paths. The robot learns from experience that certain associations will increase its chances of finding mission-critical events, while others will inhibit it. On subsequent forays into the environment, the robot uses this knowledge to improve its discovery process. This paper describes the application of these mechanisms to an intelligent agent in search of information resources in a distributed, dynamic, and unstructured computer environment. We present a search for keywords, where the agent learns the node/directory/filename cues that will increase its chances of finding documents containing those keywords.