Gaussian belief propagation for dynamic obstacle avoidance and formation control in second-order multi-agent systems

Abstract

This paper proposes a control strategy that combines Gaussian Belief Propagation (GBP) with the Artificial Potential Field (APF) method, enabling Multi-Agent Systems (MASs) to achieve global consensus in formation control while flexibly responding to dynamic obstacles within the GBP-based framework. Existing APF-based methods are difficult to cope with fast-moving obstacles exceeding the speed threshold, while the adaptive formation control and stochastic dynamic obstacle avoidance methods proposed in this paper effectively address this challenge. By utilizing the control method proposed in this paper, the MASs are able to accurately predict the future position of such obstacles and pre-plan the obstacle avoidance path. In addition, they are also able to seamlessly return to the desired formation trajectory after effectively solving the collision avoidance challenge, which proves the generalizability of our newly proposed method in various dynamic scenarios. This research offers novel insights and approaches for adaptive control and dynamic obstacle avoidance in MASs.