Near-optimal search-and-rescue path planning for a moving target.

Discrete search and rescue path planning for a moving target problem is computationally hard. Despite heuristic methods proposed so far, for which efficiency is further plagued by random episodic target kinematic behavior, qualifying real solution goodness and specifying optimality gap still remain elusive. In this paper a new alternate formulation is proposed to solve the search path planning problem for a moving target. Rather than tackling explicitly the original problem model, a remarkable property is exploited to solve a conjugate mixed-integer linear programming problem. The approach lies on a compact open-loop search path planning problem model with anticipated feedback to efficiently minimize probability of detection failure as opposed to traditionally maximize cumulative probability of success. Preserving path solution optimality, solving the conjugate problem proves very valuable in significantly reducing computational complexity. This is mainly achievable by virtue of the resulting problem objective property which concisely defines the function to be minimized in terms of terminal belief contributions only. For the first time optimal or upper bound solution computation are made possible in reasonable time for practical size problems. A network representation is further utilized to simplify modeling and facilitate constraint specification. Computational results show the strength of the approach for one and two searching agent problem instances. [ABSTRACT FROM AUTHOR]