Abstract:
In the context of a Command, Control, Communication and Intelligence (C31) system we have built a CBR module for plan recognition. It is based on knowledge representation structures called XPlans, inspired in part by Schank’s Explanation Patterns. The uncertainty inherent to an uncontrolled flow of input and the presence of lacunary data make difficult the retrieval of cases. This led us to develop an algorithm for partial and progressive matching of the target case onto some of the source cases. This matching amounts in practice to a credit assignment mechanism, included in the algorithm associated with each XPlan. This method has been designed to meet the requirements of a DRET l project to build a decision support module for a C3I system--the MATIS project. Its tasks is to interpret and complete the results of an intelligent "pattern-recognitionand- data-fusion" module in order to make the intentions underlying the recognized situation explicit to the decisionmaker. This advice is given as a causal explanation of an agent’s behavior from low-level information.