未知混动态环境下多无人机轨迹规划  

Trajectory planning for multi-drone in unknow mixed dynamic environments

作　　者：胡克 孙洪飞 HU Ke;SUN Hongfei(School of Aerospace Engineering,Xiamen University,Xiamen 361102,China)

机构地区：[1]厦门大学航空航天学院,福建厦门361102

出　　处：《智能系统学报》2025年第2期445-456,共12页CAAI Transactions on Intelligent Systems

基　　金：航空科学基金项目(20220058068001).

摘　　要：实现多无人机在野外未知混动态环境下的快速在线重规划具有较大挑战。本文提出一种分布式的动力学规划方案,用于自主无人机集群在具有静态障碍和动态障碍的混动态环境中快速重规划动态可行轨迹。首先,提出一种改进的动力学路径搜索方法,利用最优相互避碰算法弥补动力学路径搜索难以处理动态障碍和搜索效率低下的不足,获取一条安全的参考路径。然后,根据参考路径拟合出一条初始轨迹并通过基于梯度的优化方法进行优化。为提高优化效率,提出了一种适配动力学规划方案的避障梯度构建方法,它充分利用已知信息快速构建避障梯度,使得轨迹优化可以在几毫秒以内完成。最后,通过与其他规划方案相比较,验证了本方案的可行性与快速性。The realization of fast online replanning for multi-drone in an unknown mixed dynamic environment poses a significant challenge.This paper proposes a decentralized kinodynamic planning scheme to rapidly replan dynamically feasible trajectories for autonomous drones in mixed dynamic environments with static and dynamic obstacles.Firstly,we introduce an improved kinodynamic path search method that addresses the limitations of dealing with dynamic obstacles and low search efficiency.This is achieved by incorporating the optimal reciprocal collision avoidance algorithm,resulting in a safe reference path.Then,an initial trajectory is fitted according to the reference path and optimized using a gradient-based optimization method.To enhance optimization efficiency,a gradient construction method is proposed,adapting kinodynamic planning schemes.This method efficiently utilizes known information to rapidly construct obstacle avoidance gradients,enabling trajectory optimization to be completed within a few milliseconds.Finally,the feasibility and efficiency of the proposed approach were validated through comparison with other planning schemes.

关 键 词：自主智能体 实时系统 运动规划 在线搜索 碰撞避免 优化 梯度方法 分布式控制 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]