Path deviation control for multi-UAV cooperative task execution based on laser vision guidance and trajectory prediction

Abstract

This paper proposes a path deviation control method that integrates laser vision guidance and trajectory prediction to address deviations caused by limited environmental perception in multi-drone cooperative tasks. First, a local 3D point cloud map is constructed using Light Detection and Ranging (LiDAR). Global environment modeling is achieved through Oriented FAST and Rotated BRIEF Simultaneous Localization and Mapping 3 (ORB-SLAM3) and a multi-agent cooperative mapping mechanism. Second, multidimensional constraints, including turning radius, velocity, and safety distance of the unmanned aerial vehicles (UAVs), are incorporated into an improved A* algorithm to enhance the accuracy and feasibility of trajectory prediction. Finally, real-time path correction is realized using a fuzzy proportional-integral-derivative (PID) controller, forming a closed “perception-planning-control” loop. Experimental results show that the proposed method achieves a path deviation of only 1.0 m, reduces task time to 10 min, and maintains a collision rate as low as 2 %. It remains stable under complex conditions such as Global Positioning System (GPS) denial and strong winds. The method significantly improves path accuracy, stability, and obstacle avoidance capability in multi-drone cooperative tasks and can be widely applied in fields such as logistics distribution, disaster monitoring, and cooperative inspection.