Abstract
Motion planning for multiquadrotor in cluttered environments presents a significant challenge, particularly due to the need for precise navigation through static obstacles, dynamic obstacles, and the simultaneous interaction with other agents. In this article, a decentralized and asynchronous trajectory planning method is proposed to generate collision-free trajectories. We utilize a guided time-optimal model predictive control (GTO-MPC) algorithm to achieve trajectory planning. First, a dynamic programming (DP)-based controller is employed to design boundary state-constrained primitives (BSCPs), which are then utilized to generate a motion primitive library. Then, the algorithm integrates a neural network (NN) with gradient-free differential evolution (DE) to perform rolling optimization, coupled with a collision detection mechanism for local replanning in response to environmental changes. We conducted multiple simulations and experiments to compare performance in obstacle-free and cluttered environments. The results demonstrate that our distributed method effectively generates safe, smooth, and deadlock-free motion trajectories for multiquadrotor systems, significantly enhancing their motion planning capabilities.
Funding Information
  • National Natural Science Foundation of China (62503149)
  • National Natural Science Foundation of China (62403183)
  • National Natural Science Foundation of China (U24A20271)
  • Natural Science Foundation of Hebei Province (F2024208002)
  • Natural Science Foundation of Hebei Province (F2022208007)
  • Science and Technology Project of Hebei Education Department (QN2024205)
  • Key Research Projects in Fundamental Sciences at Higher Education Institutions in Hebei Province (241791007A)
  • Major science and technology support project of Hebei Province (242G1601Z)
  • Major science and technology support project of Hebei Province (2510102107A)