Autonomous UUV control via tunably decentralized algorithms

Abstract

We apply previous studied control algorithms to the cooperative target observation (CTO) task for multiple UUVs. The algorithms are based on k-means clustering and hill climbing, and each are scalable in the degree of decentralization. In the underwater formulation of the CTO problem, k-means is not sensitive to the degree of decentralization, while the hill climber is sensitive. Unlike in previous work, K-means outperformed hill-climbing across all environmental parameters.