An Architectural Framework for Integrated Multiagent Planning, Reacting, and Learning
ATAL '00 Proceedings of the 7th International Workshop on Intelligent Agents VII. Agent Theories Architectures and Languages
| Hi-index | 0.00 |
WE DESCRIBE AN ARCHITECTURE OF A DISTRIBUTED DIAGNOSIS MODULE (DDM) DESIGNED TO WORK IN CONJUNCTION WITH A DISTRIBUTED PROBLEM-SOLVING SYSTEM TO DIAGNOSE INAPPROPRIATE SYSTEM BEHAVIOR. THE PROBLEM-SOLVING SYSTEM BEING DIAGNOSED IS THE DISTRIBUTED VEHICLE MONITORING TESTBED (DVMT), AN INTERPRE TATION SYSTEM BASED ON THE HEARSAY-II ARCHIECTURE. THE WORK DESCRIBED HERE IS THE CONTINUATION OF OUR WORK ON A CENTRALIZED DIAGNOSIS MODULE (DM), WHICH HAS BEEN IMPLEMENTED AND IS USED TO IDENTIFY FAULTS IN THE DVMT, SUCH AS SENSOR FAILURES OR INAPPROPRIATE CONTROL PARAMETER SETTINGS. IN THE CEN TRALIZED CONFIGURATION, WE ASSUME THAT ONE DIAGNOSTIC NODE IS RESPONSIBLE FOR DIAGNOSING THE SYMPTOMS AT ALL PROBLEM-SOLVING NODES. WE BRIEFLY DES- CRIBE THE ARCHITECTURE OF THE CENTRALIZED DM AND ILLUSTRATE ITS DIAGNOSIS VIA CAUSAL PATHWAY CONSTRUCTION. WE THEN FOCUS ON THE ISSUES IN DISTRIBUT- ING THE CENTRALIZED DM AND ON THE CHANGES NECESSARY TO MAKE THE DM FUNCTION IN A DISTRIBUTED ENVIRONMENT; PRIMARILY MORE FLEXIBLE LOCAL CONTROL. WE DISCUSS CASES WHERE THE DIAGNOSIS RESULTS IN TWO OR MORE FAULT CANDIDATES AND EVIDENTIAL REASONING, THROUGH INTERACTION WITH OTHER DIAGNOSIS MODULES, IS NECESSARY TO DISAMBIGUATE THESE. THE EVIDENCE NECESSARY FOR THIS COMES FROM REDUNDANCY IN THE DVMT SYSTEM. WE PROPOSE AN ARCHITECTURE WHERE EACH PROBLEM-SOLVING NODE HAS ITS OWN DIAGNOSIS MODULE AND THESE MODULES THEN INTERACT IN CONSTRUCTING THE CAUSAL PATHWAYS NECESSARY TO EXPLAIN THE INAPP