Optimizing sonobuoy placement using multiobjective machine learning

We present a new approach to finding optimal patterns for the placement of fields of sonobuoys in a complex undersea environment. The problem is modelled as a biobjective one, where the aim is to both minimize uncertainty over target localization and minimize sensor placement time. We develop a two-phase algorithm, where an offline multiobjective evolutionary phase finds initial Pareto-nondominated solutions to a static problem, and then an online multiobjective reinforcement learning phase finds improved solutions using updated information. Initial results show that our approach generates significant improvements over standard grid patterns.