基于动态并行优先级解耦的大规模集群在线协同轨迹规划研究

As a potential game-changing technology, swarm intelligence has become a critical development trend for unmanned systems. Online trajectory planning of large-scale swarm is one of the bottleneck technologies for unmanned swarm intelligence. To ensure the completeness and enhance the efficiency of large-scale swarm trajectory planning, this project focuses on the research of decoupled cooperation and iterative optimization. The main research tasks are summarized as follows. (1) A dynamic-priority based decentralized parallel planning mechanism is proposed. On the basis of completeness analysis of the proposed method, optimization model of priorities and efficient optimization algorithm is studied to enhance the efficiency of decentralized swarm coordination. (2) Considering the communication constraints, a distributed prioritized cooperation mechanism within the receding planning frame is proposed to ensure the consensus and completeness under the centerless network. And the distributed combinatorial optimization method and coordination strategies are tailored to improve the distributed coordination efficiency. (3) An efficient iterative optimization method is proposed for single-agent trajectory planning under the decoupled cooperation framework. An adaptive trust region method is introduced to ensure the convergence of iteration. The sub-optimization computational cost is decreased via formulating low-complexity subproblems and developing sparse optimization algorithms, which is beneficial to improving online planning efficiency of swarms. This study can significantly improve the performance of large-scale swarm online trajectory planning, and provide advanced theories and methods for unmanned swarm intelligentization.

作为一种潜在的颠覆性技术，集群智能已成为无人系统发展的重要趋势。大规模集群在线轨迹规划是实现无人集群智能化的一项瓶颈技术。为了满足大规模集群轨迹规划的完备性和时效性要求，本项目围绕解耦协同机制和迭代优化算法展开研究。主要内容包括：①提出基于动态优先级的分散式并行规划机制，在分析其完备性的基础上，构建优先级优化模型和高效求解算法，提高集群分散式协调效率；②考虑集群通信约束，提出一种滚动的分布式优先级协同机制，在无中心条件下保证集群一致性和规划完备性，并定制高效的分布式组合优化算法与协调策略，提高分布式协调效率；③在优先级解耦协同架构下，研究个体轨迹的高效迭代优化算法，引入自适应信赖域约束以保证迭代收敛性，构建低复杂度子问题和定制稀疏优化算法以降低子优化成本，进而提高集群在线轨迹效率。本项目研究能够提高大规模集群在线轨迹规划能力，为实现集群智能化提供理论与方法支撑。

集群智能是无人系统的重要发展趋势之一，无人集群应用于军事、农业、物流等领域，能够提高任务效率和扩大任务范围。在线轨迹规划是实现大规模无人集群任务自主的瓶颈技术之一。本项目围绕基于动态优先级的集群解耦协调机制和基于迭代优化的个体轨迹规划算法展开研究，主要研究内容包括：①提出一种基于动态优先级的分散式并行规划机制，在集群信息互通的条件下改善了协同规划收敛性和规划效率，规划时间达到十秒级；在此基础上，引入滚动规划框架将问题分解为若干短时域规划问题，进一步提高了算法效率和扩展性，单次滚动规划耗时为秒级；②考虑集群通信约束，提出一种基于局部优先级的分布式滚动协同规划机制，在无中心条件下实现集群规划的协调一致；在此基础上，引入由粗到细的分层轨迹规划方案，提高分布式轨迹规划效率，分布式条件下滚动规划耗时达到秒级；③针对个体轨迹的迭代优化求解，提出一种基于自适应信赖域过滤的迭代凸优化方法和一种面向轨迹规划约束特征的定制内点法，集群规划效率提高到1秒左右。通过本项目研究，大规模集群在线轨迹规划效率提高了一个数量级以上，为实现集群智能化提供了理论与方法支撑。

内容获取失败，请点击重试