【作者】 张国有；

【导师】 曾建潮；

【作者基本信息】 兰州理工大学 ， 控制理论与控制工程， 2013， 博士

【摘要】 群机器人系统是由数量众多,功能和结构相对简单的自主移动机器人所组成的系统,可通过有限的感知和交互,协作完成单个机器人所不能胜任的任务,具有鲁棒性、灵活性和规模可伸缩等特点。其协调控制策略受启发于生物群体的自组织行为,具有规则简单,分散控制,适应复杂环境等优点。本文围绕区域覆盖问题,受启发于黄蜂群响应阈值模型,进行群机器人控制策略及仿真研究。所进行的工作和取得的研究成果如下：(1)在综述机器人区域覆盖的研究现状的基础上,形式化描述了区域覆盖问题。建立固定响应阈值模型与区域覆盖问题之间的映射关系,提出了基于固定响应阈值模型的黄蜂群区域覆盖算法。通过将环境建模为无障碍的封闭栅格环境,机器人根据单元格的刺激量使用响应函数决定移动方向。在自由环境下的仿真实验,表明了该法的可行性。(2)对黄蜂群区域覆盖算法进行微观和宏观的数学分析,从而更有效地指导参数的选择。采用概率平均方法分析黄蜂群区域覆盖算法的覆盖率和重复覆盖次数。通过考察响应阈值、刺激量、刺激量修改参数等参数对覆盖性能的影响,结果表明,机器人群体的平均移动概率理论分析可在一定条件下适用。为该算法的研究和应用提供可靠的数学基础。(3)采用离散随机过程的分析方法对基于响应阈值模型的区域覆盖算法的覆盖性能进行分析和计算。结果表明该分析方法在一定条件下,不仅灵活而且有效。这种离散随机过程方法的优点是可以分析各参数的作用以及指导各参数的选取。(4)提出障碍环境下的基于固定响应阈值模型的黄蜂群区域覆盖移动策略,具体说明了机器人的建模,移动决策和避障等。在仿真实验中设置不同形状的障碍,分析和验证黄蜂群区域覆盖算法在障碍环境下的覆盖效率。说明在不同的障碍环境下算法参数对覆盖性能的影响。仿真实验的结果表明该移动策略的有效性。(5)在分析固定响应阈值模型的区域覆盖算法所存在的不足的基础上,提出改进的黄蜂群区域覆盖算法。改进的黄蜂群区域覆盖算法受启发于自强化模型,机器人在覆盖过程中根据外界信息和自身状态自主调节响应阈值,从而更灵活地响应外界任务,提高算法的性能,仿真结果表明,该方法可有效提高覆盖性能。(6)针对群机器人多目标搜索的任务分配问题,提出基于黄蜂群响应阈值模型的任务分配策略。群机器人的多目标搜索问题可分解为全局搜索和局部搜索,以及机器人的目标选取问题。针对目标选取问题,将机器人的状态分为漫游、搜索、等待、决策等状态,源于黄蜂群劳动分工的启发,提出基于黄蜂群的响应阈值模型的任务分配策略,用于目标选取。在仿真实验中,在局部搜索阶段采用扩展微粒群算法,在全局搜索阶段采用随机搜索,结合基于响应阈值的任务分配策略,实现了群机器人的多目标搜索,仿真结果表明了所提出的的任务分配方法能够有效地适用于群机器人多目标搜索。更多还原

【Abstract】 Swarm robotic system is composed of numerous of autonomous mobile robots, which functions and structure are relatively simple. Robots with limited sense and interaction can fulfill tasks cooperatively that a single robot can not be undertaken. Such a system has the characters of robust, flexibility and scalability, etc. The coordination strategy of the system, which is inspired of the self-organization behavior of biological society, has advantages of (a) simple rule,(b) decentralized control,(c) adapt to the complex environment, etc. In this thesis, it is mainly focus on the problem of terrain coverage on swarm robotics, prompted the control strategy of swarm robot and simulation research, which is inspired of response threshold model of wasp swarm. The detailed research work and results are as follows:(1) Formal description of terrain coverage problem is presented on the basis of the review on the state-of-art on terrain coverage on robotics. Due to the similarities and differences between wasp and robot, terrain coverarge can be mapped to response threshold model based on the common working mechanism. After the problem description of terrain coverage, the terrain coverage algorithm based on fixed response threshold model on swarm robotics is depicted. In this "algorithm, terrain is modeled as free closed grid environment, and robots decide their moving directions in term of stimuli of its neighbor cells.(2) In order to guide the selection of parameters in terrain coverage algorithm based on fixed response threshold, the performance of the algorithm should be anlysized. The performances of algorithm are evaluated by microscopic anlysis and probabilistic macroscopic anlysis. The macroscopic anlysis considers the effects of parameters, such as response threshold, stilumus, modified stimulus parameter etc. Verified by simulation results, the anlytic result is reliable under some conditions. This method provides us with a mathematical foundation of the research and application of the terrain coverage algorithm.(3) Another analytic method to evaluate the performance of this algorithm is stochastic discrete process. We set up this model to analyze the coverage performance of algorithm, such as mathematical expectation of average coverage rate. Results show that this analysis method is flexible and effective under certain conditions. This method based on probobalitic discrete process has the advantage of analyzing the effect of various parameters and guiding the selection of the parameters.(4) Obstacle avoidance is an important issue in automatic robotics. The moving strategy of swarm robots based on fixed response threshold under obstacle environment is put forward. In order to make the research more detailed, the modeling of the robot, and obstacle avoidance is presented. The coverage performance of the algorithm is tested through various shapes of obstacle in simulation experiments. Simulation results show the method is effective and feasible.(5) On the basis of the analysis of the advantage and disadvantage of terrain coverage algorithm based on fixed response threshold model, we introduces self-reinforcement model of response threshold and puts forward terrain coverage algorithm based on self-adjustable response threshold. The moving directions of robots are decided by the external stimuli and response threshold, which will be adjusted by robots based on the previous experience. The coverage performances of the algorithm are analyzed and tested in the simulation experiments which are conducted in free closed environments and obstacle environments. Final results show that this method can improve the coverage performance effectively.(6) Task allocation of multi-target search in swarm robotics is how to coordinate the tasks of searching targets among robots. The formal description of the problem of multi-target search and task allocation are presented^Inspired of division labor of wasp swarm, the strategy of task allocation of multi-target search based on the response threshold model in swarm robotics is prompted. The states of robots are divided into wander, decision, waiting, marking target and etc. Robots adopts random search in global search phase and particle swarm algorithm in local search in order to illustrate the effectiveness of the strategy. Simulation results show that the proposed task allocation method can be effectively applied to multi-target search.更多还原