Coordinated Obstacle Avoidance Of Multi-agent System With Partially-aware Interaction

In the biological world, collective movement refers to that a large number of biological organisms move together. The realization of group cooperation and state synchronization through self-organized behavior of each individual, such as predator avoidance, foraging and so on. A prominent feature of collective movement is that the achievement of global movement coordination depends on the simple local rules of individual. It is distributed, self-organizing and coordinating. The characteristic which the group performance is greater than the sum of individuals attracts many researchers. The dynamic model of each agent is supposed to be first order dynamics or second order dynamics. The appropriate control protocol of coordinated movement in multi-agent system is designed accordingly. The designing of control protocol for each agent is based on the behavior of its adjacent neighbours. Since the real-time states of agent group are not pre-set, the group can emerge some intelligent behaviour such as mutual coordination. But the coordinated movement of multi-agent system may have constraints when there are obstacles in the moving environment. The control protocol needs to be re-designed considering the distribution, shape and movements of obstacles. If each agent keeps coordination with other agents in group, the individual which only has simple detecting sensor or does not have sensor can avoid obstacles as well,when only a small part of agents in the group has obstacle detecting ability.In this article, the collective coordinate control protocols that make every agent in a group trend to consensus state is called consensus control. If each agent maintains an equal distance from its adjacent agent while their velocities trend to consensus, it is called flocking control. And also, the fixed topology in this paper indicates that the topology of multi-agent system is determined at the initial, as the switched topology means that each agent receives information from neighbouring agents in real time. This article studies on the cooperative control of multi-agent system avoiding obstacles in various conditions. The detailed contents are as follows:(1) A coordinated control strategies of multi-agent system that make agents group avoid smooth convex obstacles and trend to consensus with navigation item when only a fraction of agents are informed of obstacles in the environment is proposed. The navigational item is a feedback that drives dispersed agents regathering at a location.The unknown agents can also move to the rendezvous place if they maintain an adjacent path with informed agents. We prove the convergence property, consensus property and collision avoidance in the system, and demonstrate the simulation results.(2) The coordinated control of multi-agent system that make agents group flocking while avoiding non-convex obstacles is proposed, when only a fraction of agents are informed of the global information of obstacles. We generates mass virtual obstacle agents according to the crystal structure. These virtual obstacle agents give repulsion force to physical agents and guide them avoiding obstacles. Also, physical agents can calculate the geometry relationship of obstacle agents, judge and fill the non-convex region, thus ensuring there is no informed agent stays in the place with local minimum potential. The uninformed agents which keep linkage with informed ones can leave the non-convex region and flocking at the target location. The convergency property,velocities consensus and obstacle avoidance property are proved and demonstrated with simulation.(3) The coordinated control of multi-agent system bypass an unique obstacle and trend to consensus is proposed, when only a small fraction of agents are informed of obstacle information. A circular matrix is used to construct the navigational direction in control input, so that the individual agent can move around the obstacle. We give out the control strategy in the situation of fixed topology and switched topology for agents with first order dynamics and second order dynamics respectively, to avoid obstacles with varying velocity or varying acceleration. We proves the convergency property,velocities consensus and obstacle avoidance property and demonstrate the theorems.(4) A control strategy with preserved interaction is proposed for multi-agent system bypassing an irregular obstacle and trending to flocking. The adjacent agents can maintain the connectivity with the hysteresis switching rule and upper limit setting of potential function. The connectivity between each agent and obstacle is preserved as well. In the control strategy, only one agent needs to have obstacle detecting capability.And we take the nearest point on the surface of obstacle as the virtual obstacle agent in some time interval. The virtual obstacle agent is the distinct reference point for multiagent system. Since the unique informed agent is driven to bypass the obstacle agent at each time interval, other agents can follow the informed agent and bypass obstacles as well. We also proves the convergency property and obstacle avoidance property for the proposed strategy, and demonstrate the theorems.