Collision-Free Nonlinear Model Predictive Control for Quadrotors in Dynamic Outdoor Environments

Ensuring safe and efficient navigation of quadrotors in dynamic outdoor environments requires accurate prediction of obstacle motion and adaptive collision avoidance strategies. This paper proposes a Nonlinear Model Predictive Control (NMPC) framework that integrates an Interacting Multiple Model Kalman Filter (IMM-KF)-based motion prediction scheme to enhance dynamic obstacle avoidance. The IMM-KF concurrently evaluates multiple motion models and probabilistically merges their estimates, enabling robust and adaptive predictions of obstacle trajectories under varying dynamic conditions. To further enhance obstacle avoidance, an uncertainty-aware repulsive potential field term is incorporated into the NMPC cost function, dynamically adjusting the repulsion force based on prediction confidence. This formulation enables smoother navigation and reduces abrupt maneuvers, enhancing reliability. The proposed approach is evaluated in simulated outdoor environments with static and dynamic obstacles, showing improved trajectory tracking, obstacle avoidance, and robustness to sensor noise and motion uncertainty. These results demonstrate the framework’s effectiveness and real-time suitability for complex, uncertain environments.