Adaptive Swarm Navigation via Motion Salience in Cluttered Environments

Achieving adaptive swarm navigation for quadrotors in cluttered environments presents significant challenges, stemming from limited individual sensing capabilities and delayed collective responses. To address these issues, this paper proposes a swarm control approach that integrates a motion-salience-based coordination rule (MS-CR) with obstacle avoidance guided by a boundary-point-based rotational potential field (BP-RPF), thereby enabling safe and adaptive swarm navigation in cluttered environments. For coordination, a motion salience (MS) metric is introduced, enabling each quadrotor to dynamically weight its neighbors based on local perception. This allows local stimuli to propagate efficiently throughout the swarm, enhancing collective responsiveness. For obstacle avoidance, the BP-RPF-based approach relies solely on detected obstacle boundary points, eliminating the need for full obstacle reconstruction. A rotational potential field (RPF) is introduced to guide quadrotors in executing stable avoidance and surface-following maneuvers. These components are unified within a multi-rule swarm navigation framework that combines coordination, obstacle avoidance, and task-oriented behaviors. Extensive six-degree-of-freedom (6-DoF) simulations in cluttered environments demonstrate that the proposed method outperforms existing baselines in maintaining inter-agent spacing, enhancing obstacle avoidance success rates, and improving task completion efficiency. Real-world experiments with a quadrotor swarm further confirm the method’s robustness and practical feasibility. These results validate the effectiveness of the proposed approach and highlight its potential for real-world deployment in complex and constrained environments.