Influence of Heuristic Functions on Real-Time Heuristic Search Methods

Abstract

In this paper we determine if heuristic function properties affect traditional and real-time heuristic search methods differently. The extent to which heuristic functions affect search method properties is measured by comparing traditional and real-time heuristic search methods. Problems of concern are observable, deterministic problems whose solution is a sequence of actions. Research is limited to influence of heuristic functions in real-time search domain only. It is shown that LSS-LRTA* produce higher wall times with a middle value heuristic and that solution cost is linear with heuristic value in IDA*. The same is not the case in LSS-LRTA*. Furthermore, the behaviour of LSS-LRTA* nodes expansion is similar to that of IDA*: a downward slope relative to heuristic value increase. The solution time decreases for LSS-LRTA* but is almost constant for IDA*.