Bio-Inspired Intelligence-Based Multiagent Navigation With Safety-Aware Considerations
Tingjun Lei, Chaomin Luo, Simon X. Yang, Daniel W. Carruth, Zhuming Bi
Abstract
Multiple autonomous vehicles (MAVs) enhance efficiency and task execution compared to a single vehicle. Real-world applications necessitate MAVs to safely navigate in dynamic formation along planned trajectories, while sensing, mapping, and avoiding obstacles. Addressing the need for trajectory adaptation amidst real-world scenarios, a safety-aware bioinspired framework is proposed in this paper. Our approach employs a chaotic gravitational search algorithm (CGSA) for global trajectory generation in a predefined formation. A quadtree-driven variable resolution (QVR) algorithm using monocular cameras provides occupancy grid maps (OGMs) at different resolutions. A formation control with target tracking minimizes a potential function for MAVs to follow the CGSA trajectory. Additionally, a bio-inspired neural network (BNN) local navigator coupled with dynamic moving windows (DMW) advances obstacle avoidance and refines safe trajectories using QVR and OGMs. Simulation and comparative studies validate the framework’s robustness and effectiveness for MAVs.