Resource-constrained search

An optimal resource-constrained heuristic search algorithm is presented. Whereas finding an optimal solution is known to be NP-complete in a number of nodes, a polynomial-time near-optimal approximate solution is initially computed. It is argued that the algorithm is more efficient than other methods, and results of numerical experiments showing how close to the optimal solution one can get by using only phase one of the algorithm are reported. The algorithm is compared with infeasibility backtracking for three tasks: computing a feasible, approximate, and optimal solution and is shown to be better.<
>