Multi-objective Optimization of Confidence-Based Localization in Large-Scale Underwater Robotic Swarms

Abstract

Localization in large-scale underwater swarm robotic systems has increasingly attracted research and industry communities’ attention. An optimized confidence-based localization algorithm is proposed for improving localization coverage and accuracy by promoting robots with high confidence of location estimates to references for their neighboring robots. Confidence update rules based on Bayes filters are proposed based on localization methods’ error characteristics where expected localization error is generated based on measurements such as operational depth and traveled distance. Parameters of the proposed algorithm are then optimized using the Evolutionary Multi-objective Optimization algorithm NSGA-II for localization error and trilateration utilization minimization while maximizing localization confidence and Ultra-Short Base Line utilization. Simulation studies show that a wide localization coverage can be achieved using a single Ultra-Short Base Line system and localization mean error can be reduced by over 45% when algorithm’s parameters are optimized in an underwater swarm of 100 robots.