生物启发的多无人机协同四维航迹规划方法研究

发布时间：2018-01-26 05:41

本文关键词： 生物启发广义tau理论多无人机四维航迹规划多智能体Q学习矢量场制导协同目标跟踪　出处：《浙江大学》2016年博士论文　论文类型：学位论文

【摘要】：多无人机在协同执行同时攻击、编队形成等飞行任务时,通常需要严格按照期望时间飞抵指定的目标点,因而多机协同四维航迹(三维位置与时间)规划技术日益受到青睐。该技术能够有效降低航迹的不确定性,提高任务执行的安全性和可靠性；并可对任务空域进行有效利用,获得更优化的协同航迹。针对多机协同四维航迹规划问题进行研究具有良好的理论意义和广阔的应用前景。广义tau理论是一种生物启发的四维运动规划理论,采用当前与目标运动状态之间的间距闭合的时间信息进行制导。基于广义tau理论的tau制导策略能够真正进行时空的同步规划,且形式简洁,适合应用于多机协同四维航迹规划中。本文以生物启发的多无人机协同四维航迹规划方法为研究内容,将tau制导策略应用于多机协同四维航迹规划中,重点研究了符合无人机需求的tau制导策略,并在此基础上对集中式、分布式和基于多智能体Q学习算法的协同四维航迹规划方法进行研究；进而,利用tau制导策略的时空协同性质,构建了四维tau制导矢量场,并用其解决多机协同standoff跟踪问题。本文的主要研究工作和贡献如下：1、系统地综述了多无人机协同四维航迹规划的研究现状,回顾了生物启发的广义tau理论及其制导策略的研究历程和进展,并总结了已有tau制导策略的优缺点和应用范围。2、针对已有tau制导策略存在的缺陷和四维航迹规划的应用需求,对制导策略进行创新。首先,针对已有策略航迹空间调整范围小、初始加速度非零的缺陷,提出了基于简谐运动的本征tau-harmonic制导策略,提升了航迹的可飞性和空间形状调整能力。其次,针对始末速度非零的四维运动规划问题,通过在本征tau-gravity (tau-G)制导策略的虚拟制导运动中加入初速度,提出了改进tau-G制导策略,能够满足编队形成、目标的持续跟踪等任务的应用需求。3、基于tau-harmonic制导策略,提出了一种集中式多机协同四维航迹规划方法。该方法采用tau-harmonic策略描述四维航迹,设计了集中式四维航迹规划问题,并用粒子群优化获取最佳航迹参数。航迹之间的无碰需求通过冲突检测和消解方法加以保证。对多机同时到达任务的大量仿真结果表明,该集中式四维航迹规划方法具有较好的收敛性和更强的冲突消解能力,能够提供更可飞和更安全的四维航迹。4、利用改进tau-G制导策略,提出了一种分布式多机协同四维航迹规划方法。该方法将集中式航迹优化问题分解为局部优化问题,每架飞机根据本地和邻居信息对局部优化问题进行求解。为了应对环境和通信拓扑的突变,采用时间采样和冲突事件双重驱动的分布式滚动优化方法不断对航迹进行优化和改进。与分布式预测控制对照的综合飞行任务仿真结果表明,该分布式协同航迹规划方法能够严格按照期望时间导引无人机以期望速度通过目标点,且能够更高效地提供费用小、安全性高韵四维航迹。5、为了提升多无人机航迹规划的自主性和灵活性,提出了一种基于多智能体Q学习算法的协同四维航迹规划方法。各无人机采用连续状态-连续动作的Wire Fitting Neural Network Q学习算法对基于改进tau-G策略的航迹规划任务进行学习训练,多无人机系统的学习用Win or Learn Fast-Policy Hill Climbing算法进行组织；为提升算法对不同任务环境的适应性,在有效通信的机群中引入学习经验交流机制。对综合飞行任务的动态仿真结果表明,该方法能够在分布式航迹规划任务中更加高效地提供满足可飞性、安全性和任务需求的四维航迹。6、利用tau-G制导策略将时间信息引入矢量场中,构建了四维tau制导矢量场,可严格按照期望时间将无人机的位置导引至目标圆上；进而利用tau矢量场解决有时间约束的多机协同standoff跟踪问题。对典型协同standoff跟踪任务的仿真结果表明,tau矢量场提供的制导策略具有计算负荷低、跟踪偏差小、可飞性好、安全性高等优势,能够更好地满足多机协同任务的四维制导需求。
[Abstract]:Multi UAV attack in collaborative execution at the same time, the formation of the formation of the mission, usually need to specified targets in strict accordance with the expected time, and multi Aircraft Cooperative four-dimensional trajectory (3D position and time) planning technology has been increasingly popular. This technology can effectively reduce the uncertainty of the track, improve the safety and reliability of task execution and on the task; the effective use of airspace, get more cooperative route optimization. For multi Aircraft Cooperative four-dimensional path planning problem research has good theoretical significance and wide application prospect. The generalized tau theory is four-dimensional motion planning theory of a bio inspired, using the time information between the current and target state space closure the guidance planning guidance strategy tau. Synchronization of the generalized tau theory truly based on space time, and concise form, suitable for multi machine Four cooperative path planning. Multi UAV cooperative four-dimensional path planning method as the research content in this paper, bio inspired, the application of tau guidance strategy for multi Aircraft Cooperative four-dimensional path planning, focusing on the UAV guidance strategy in accordance with tau requirements, and on the basis of centralized, distributed and collaborative D route planning method agent based on Q learning algorithm is studied; then, the nature of collaborative use of space-time tau guidance strategy, establishes four-dimensional tau guidance vector field, and used to solve the multi machine collaborative standoff tracking problem. In this paper, the main research work and contributions are as follows: 1, summarized the research status of collaborative multiple D route planning the aircraft, reviews the history and research progress of bio inspired generalized tau theory and its guidance strategy, and summarizes the advantages and disadvantages of the existing tau guidance strategy and application scope of.2, according to The existing tau guidance strategy defects and D route planning application needs, to innovate the guidance strategy. Firstly, aiming at the existing strategy adjustment range of track space is small, the initial acceleration of non zero defects, put forward the intrinsic tau-harmonic guidance strategy based on simple harmonic motion, improve the trajectory of flying and spatial shape adjustment ability. Secondly, according to the motion planning problem by four-dimensional non zero speed, the intrinsic tau-gravity (tau-G) virtual guidance guidance strategy is added in the initial movement speed, this paper proposes an improved tau-G guidance strategy can satisfy the formation, continuous tracking target application requirements.3 tau-harmonic guidance strategy based on a multi machine collaborative four-dimensional path planning method. This method adopts tau-harmonic strategy description D route design, route planning and centralized four, with grain Swarm optimization to obtain the optimal trajectory parameters. The collision free needs to be ensured by the conflict detection and resolution method between tracks. A large number of simulation results of multi machine task that arrived at the same time, the conflict of the centralized D route planning method has better convergence and stronger ability, can provide more and more secure 4D flight path.4 and using the improved tau-G guidance strategy, this paper proposes a distributed multi Aircraft Cooperative four-dimensional path planning method. This method will be centralized track optimization problem is decomposed into local optimization problems, according to each plane to solve the local optimization problem of local and neighbor information. In order to cope with environmental mutation and communication topologies, using time sampling and dual conflict drive the rolling optimization method of continuous distributed route was optimized and improved. And the distributed predictive control integrated flight control simulation The results show that the distributed cooperative path planning method can be in strict accordance with the expected time of the UAV to the desired speed through the target point, and can more efficiently provide low cost, high safety rhyme four-dimensional track.5, in order to enhance the autonomy and flexibility of multi UAV path planning, the paper presents a multi-agent Q learning algorithm of cooperative path planning method based on four. Each UAV using continuous state continuous action of the Wire Fitting Neural Network Q learning algorithm to learn the training task improved path planning based on tau-G strategy for multi UAV system using Win or Learn Fast-Policy Hill learning Climbing algorithm to enhance the adaptability of the algorithm; different task environment the introduction of learning experience exchange mechanism in the effective communication in the cluster. The dynamic simulation of the integrated flight task. The results show that this method can In the distributed path planning tasks more efficiently provide flying.6, four-dimensional track safety and mission requirements, the use of tau-G guidance strategy and time information is introduced in the vector field, establishes four-dimensional tau guidance vector field, can be in strict accordance with the expected time of the UAV's location is guided to a target circle; then using tau the vector field problem with time constraints for cooperative standoff tracking problem. Simulation of a typical collaborative standoff tracking task. The results show that the tau vector field provides a guidance strategy with a low computational load, tracking deviation is small, flying, high security advantages, can better meet the needs of four-dimensional guidance for cooperative tasks.

【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：V279

【相似文献】