Fault Tolerant Control for Unmanned Aerial Vehicle Formation Based on Distributed Nonlinear Model Predictive Control

This article proposes a new fault-tolerant obstacle avoidance and trajectory tracking strategy for unmanned aerial vehicle (UAV) formations. The proposed method integrates deep neural networks (DNNs) with distributed nonlinear model predictive control (DNMPC) to handle actuator faults and collision risks in complex environments. The DNNs serve as the primary component for fault-dynamics modeling and real-time compensation, which ensures the system's stability through adaptive weight updating. The DNMPC framework utilizes state information from neighbors to predict future states and compute optimal control inputs. This approach reduces computational and communication burdens and enables real-time operation. A terminal cost function and terminal set are designed to guarantee the system's stability. Comparative simulations and experimental results demonstrate that the proposed methodology improves the trajectory–tracking accuracy and formation-maintenance performance. The formation retains full tracking capability under actuator faults, which validates the approach's effectiveness and practicality in challenging environments.