【作者】 刘佰龙；

【导师】 张汝波；

【作者基本信息】 哈尔滨工程大学 ， 计算机应用技术， 2009， 博士

【摘要】 群集智能是受自然界生物群体社会性行为启发而产生的概念,它是指简单个体通过局部交互所涌现出的不可预见的全局智能行为的特性。系统中的个体仅在局部感知能力下,通过同其他个体以及环境的交互就可以实现复杂的行为模式。群集智能利用群体的优势,在没有集中控制、不提供全局模型的前提下,为寻找复杂问题的解决方案提供了新的思路。显然,对群居生物集体行为的研究有助于理解群集智能系统的内在机理,及其在不同领域的应用。本文目的是研究群集智能的有关理论问题,给出一套用于分析群集智能系统及相关行为稳定性和复杂性的方法,并将研究成果应用到多机器人系统中,实现多机器人的协调协作。首先,证明了个体模型下群集智能系统自聚的稳定性。自聚的稳定性是指个体在移动过程中能够成群移动并最终聚集到目标位置周围的邻域内。微观上,通过将个体之间的相互作用转化为势场力,结合考虑个体的感知能力、动态环境以及噪声等因素,建立了群体聚集中的个体模型,运用Lyapunov稳定性理论证明了在上述情况下群体自聚的稳定性。其次,证明了群集智能系统中环形分类行为的稳定性。环形结构同群体聚集时产生的最终结构在空间上具有一定的相似性。将群体聚集个体模型进行扩展,增加不同类对象之间的相互作用,建立了环形分类中被分类对象的个体模型。通过Lyapunov稳定性理论给出了多类物体环形结构构建的稳定性分析。再次,研究了群体模型下的群体自组织特性。宏观上,分析群集智能系统个体之间以及个体和环境之间相互作用产生的效果,并考虑个体的行为规则,有助于建立系统的群体模型。将群体和环境看作一个开放系统,结合耗散结构的相关理论,基于平均场法建立了描述系统宏观状态的群体模型。通过模型的解析解、数值解以及Monte Carlo模拟,分析了系统状态随各个参数的变化规律,得出了群集智能系统多稳态性、模式涌现以及分叉现象产生的机制和条件。最后,将群集智能模型和理论分析结果应用到机器人领域中,以编队和觅食任务为背景,实现了多机器人的协调协作。从行为上看,群体聚集和多机器人编队在过程上具有一致性。机器人间的引力和斥力平衡时可以保证机器人满足一定的几何约束,而环境中的障碍物可以引入同目标相反的势函数模拟。通过引入机器人的动力学方程,结合群体自聚个体模型实现了多机器人的编队控制。以觅食任务为背景的多机器人协作行为,由机器人之间的交互方式入手,提出了基于信息素反应的觅食策略,有效地提高了多机器人觅食的效率,降低了机器人之间的冲突。更多还原

【Abstract】 Swarm Intelligence(SI),which is inspired by the cooperation of social animals,means the unpredictable macroscopical intelligence behavior emerges from the interaction of simple individuals.The individuals only have the capability of local perception.SI provides the new idea to find the solutions of complex problems without center controller and global information.It is helpful to understanding the internal mechanism of SI system,and to apply SI into different fields through researching the behavior of social animals.The aim of this thesis is to research the theoretic problems on Swarm Intelligence,and to provide the general approaches on analyzing the stability and complexity of the behaviors or algorithms in SI system.And then,the results obtained here are applied to multi-robot system.So the coordination and cooperation of multi-robots are implemented.Firstly,the stability of self-aggregation of SI system in complex and dynamics environment is proved by the individual model.The stability of self-aggregation means that the individual can move coherently to the target and can stay in finite neighbor field of the target finally.The interaction of individuals is transformed into potential force.So the individual model of aggregation is built by considering the sensing capability of the individual,noise in and dynamics of environment.And then the stability of self-aggregation is proved by Lyapunov stability theory.The analysis mentioned above makes the research on SI more factual.Secondly,the stability of annular structure in SI system is proved.The annular structure in social animals exhibits high cooperation.The spatial relationship in annular structure seems similar to that in aggregation mentioned above.So the object model in annular sorting can be built by increasing the interaction between different types of objects.And the stability of building annular structure is given by Lyapunov stability theory.Thirdly,the Self-organization of swarm model is researched.In macroscopical perspective,in order to build the swarm model of the system,the interactions between individuals and the environment need to be analyzed,and the rules of individual behavior should to be considered.In addition,the swarm and the environment can be considered to be an open system.From the point of view of Dissipation Structure Theory,the swarm model of SI system in macroscopical status can be described by the differential equations.The characters of stationary solution of the equations are obtained by analyzing the solutions and numerical results the equations,and Monte Carlo simulation.So the multistability,emergent patterns and bifurcation in Self-organization systems are also obtained.Finally,the models and their analysis are applied into robotic systems.So the coordination and cooperation of multiple robots are implemented in the background of formation control and foraging task.The aim of researching on SI theory is to apply it into factual system.There is something consistent between the formation control of multiple robots and aggregations mentioned above.The balance between the attractive and repellent force can satisfy the formation restriction.And the obstacles in formation can be represented by the potential field which is contrary to the target.So the formation control is implemented by combine the dynamic equation and the aggregation model.In addition,in order to improve the performance of foraging behavior in multiple robots system,the patterns of interaction are considered by the results analyzed in each model. Through adding different type of pheromone which can react with each other,the performance the system increases and the conflicts between the robots decrease efficiently.更多还原