Using an Assumption-Based Truth Maintenance System to Switch Contexts during dataFusion Processing

摘要

This report is organized into seven sections. The first deals with issues raisedby the effect of time on discrete event modeling, and is known as the frame problem. The second section describes a classical control theory approach to multiple target tracking, which relies exclusively upon Bayes theorem and a Kalman filter to track objects which change spatial position over time. The third section on justification-based truth maintenance systems introduces the concept of retraction to correct temporal inconsistencies which arise in situation databases when objects change spatial position. The fourth section on assumption-based truth maintenance system is reactionary to the third, in that it proposes to relieve the computational burden of retraction by pursuing several feasible hypotheses in parallel. The fifth section, the problem of uncertainty in data fusion and the implication for logic-seeking truth maintenance technology is discussed. The sixth section describes some Army work which can provide real-time collateral inputs to Reid's targe tracker, and also to the two styles of truth maintenance philosophy. The final section proposes a handshaking protocol to interface Reid's statistical system to a logic-driven truth maintenance system.