Efficient Path Planning and Formation Control in Multi-Robot Systems: A Neural Fields and Auto-Switching Mechanism Approach

Current implementations of leader-follower systems in multi-robot environments typically separate path planning and formation control, often leading to inefficiencies in execution speed, path quality, and system stability. These systems rely on complex control strategies that struggle to adapt to dynamic and challenging environments, resulting in sub-optimal performance and potential collisions. This paper presents a novel approach to enhancing the path planning and formation control of multi-robot leader-follower systems by integrating neural fields and novel potential field modeling. The primary focus is on developing a unified platform that addresses the challenges of maintaining formation integrity, optimizing path quality, and ensuring real-time responsiveness. Neural Fields, which has gained significant traction in the past few years, utilizes fully connected neural networks to effectively encode continuous signals across various dimensions and resolutions. This growing interest in neural fields has highlighted their potential to provide solutions that are not only more precise and high-fidelity but also highly expressive and efficient in terms of memory usage. Thus, this work proposes leveraging neural fields to achieve high-quality path planning in significantly reduced time. By utilizing the expressiveness and memory efficiency of neural fields, the suggested approach aims to generate optimized paths quickly, making it well-suited for real-time applications where both accuracy and speed are critical. On the other hand, the follower robots are equipped with an auto-switching potential method, which intelligently toggles between attractive forces guiding the followers toward the leader and repulsive forces preventing collisions based on a mathematical model. The effectiveness of this approach is validated through experiments, demonstrating significant improvements in execution speed, path smoothness, and overall system stability compared to competitive methods, including well-known techniques such as A*, Probabilistic RoadMap (PRM), Rapidly-exploring Random Tree Star (RRT*), and also against recent optimization techniques including Grey Wolf Optimization (GWO) and Whale Optimization Algorithm (WOA).