Globally optimal scheduling over flexible intel, resources, and goals
Benjamin Hartman · OCEANS 2017 – Anchorage · 2017
Often in the various stages of mine countermeasure operations we are confronted with a situation in which we have knowledge informing us of the number and spatial distribution of mines, but we are burdened by limited time or resources. Conventional planning methods rely upon track spacing algorithms which attempt to reduce risk to naval assets. We present algorithms which determine the global optimal configuration of tracks which meet the goals defined by various objective functions. We investigate a variety of new approaches, principally exploring the merits of Adaptive Simulated Annealing (ASA) and the DIRECT algorithm in solving this higher dimensional problem. Adaptive simulated annealing acts as a multi dimensional hill climbing algorithm: as sampling progresses, the simulation will transition from favoring a search for global optima to favoring local optima while adjusting the manner in which points are sampled along each dimension. While typically our problem is solved by approximating the objective function or by finding locally or incrementally optimal solutions, we focus on mitigating any errors caused by sampling points by choosing a proper objective function and simulation parameters.