Integrated tracking and control framework for robotic multistatic sonar networks with IMM-BP and distributed MCTS

Abstract

This paper investigates a novel application of robotic multistatic sonar networks, where multiple stationary acoustic sources collaborate with an autonomous underwater vehicle (AUV) equipped with a line array to track non-cooperative underwater maneuvering targets. To enhance tracking accuracy, an integrated framework that combines an improved IMM-BP tracking algorithm with a cooperative control strategy is proposed. The IMM-BP algorithm extends traditional particle-based belief propagation (BP) by incorporating the interactive multiple model (IMM) approach during the prediction phase, reducing computational complexity from quadratic to linear and improving scalability and efficiency. Leveraging the IMM-BP tracker, the receding horizon control method jointly optimizes the AUV’s heading angle and the source ping schedule. To address the large-scale, non-myopic tree search challenge inherent in this control strategy, a distributed Monte Carlo tree search algorithm is proposed. This algorithm partitions the search tree and distributes computation across multiple autonomous agents, significantly improving computational efficiency while maintaining effective parallel search performance with minimal communication overhead. Simulation results demonstrate that the proposed framework significantly improves tracking accuracy, cooperative control efficiency, and computational performance, underscoring its advantages in robotic multistatic sonar networks.