Epistemic Logic Planning: Case-based Planning Adaptation, Using Epistemic Logicrevision for Robot's Decision Making

Artificial Intelligence algorithms can be dividedinto two groups according to the type of problemsthey solve. Knowledge-intensive domains containexplicit knowledge, whereas knowledge-poor domainscontain implicit knowledge. Rules, decision trees andlogical methods are more suitable for the first type. Neural networks and case-based reasoning (CBR) aremore suitable for the second type. This project combines the inferencing power ofepistemic logic (type 1) in the adaptation of CBRwith the performance of case-based planning (type 2). This method is proved to be more efficient then usingplanning algorithms alone. Here a simulated robot is assigned to deliverparts in a factory. The robot needs to plan the paththat it should choose to achieve its goals and usesepistemic logic to solve problems. Planningalgorithms are computationally expensive. CBR, usingKNN is used to make the process faster. A STRIPSplanner creates plans. The manager defines theproblem, KNN extracts a plan and a logic sub-systemadapts it according to belief revision theorems. Thenew plan is retained for future use.