Coordinated Control Of Distributed Multi-agent Systems And Applications

Due to its strong tolerance, robustness as well as scalability, multiple agent system (such like mobile sensors, unmanned aerial vehicles), considered as one of the most im-portant subsystem in distributed Artificial Intelligence field, is widely used in military, industrial and agricultural production, medical, traffic, service industry and so on. Multiple agent system is composed of a couple of agents with computing and mobility capability, where each agent is either physical or abstract. The agents could communicate and cooper-ate with each other. This thesis provides some cooperative and control algorithms based on different applications. The main work and results of the thesis can be cataloged as follows:Firstly, the search problems with a single agent and a group of agents are addressed, respectively. In the case of a single agent searching, we give two efficient simulated-annealing like algorithms to solve the problem in partially known environment and un-known environment. The results show that under both proposed control strategies, the agent will eventually converge to a globally optimal segment with probability1. Compared to the existing searching strategies, our algorithm avoids being trapped into locally optimal solu-tions, and also gives a significant relaxation on memory requirements. In the multiple agent case, we use multi-agent searching simultaneously to reduce the computation complexity and accelerate the convergence of the algorithms we have given for a single agent. By ex-ploiting graph partition, a gossip consensus method based scheme is presented to update the key parameter—radius of the graph, ensuring that the agents spend much less time finding a globally optimal segment.Secondly, we analysis the globally optimal coverage problem for non-convex grid environments. A distributed algorithm is proposed, which is locally implementable on in-dividual agent. That is, each agent only requires to update its current position to an adja-cent location with certain probability relying on its coverage costs at current location and adjacent locations. It shows that under the proposed algorithm, a network of agents will eventually converge to a globally optimal configuration minimizing the coverage cost with probability1. Compared to the existing coverage algorithms, Our algorithm only needs local information of the neighbor agents, but solves globally optimal coverage solution in non-convex environment. Moreover, our algorithm could deal with the obstacle in the environment as well.Thirdly, the optimal transmit beamforming problem based on the feedback informa-tion is addressed. According to different wireless networks system, we design two dis-tributed algorithms running on each agent (each carrying an antenna) such that the signals from all agents combine constructively at the receiver end and thus, the received power is increased. Both algorithms are iterative whereby each sensor adjusts its phase offset with-out directly communicating with each other. In each iteration, either one sensor or a subset of sensors are selected and then transmit the signals that are carefully crafted based on the SNR feedback from the receiver end. We have proved that guided by either algorithm, the signals from all sensor nodes converge to perfect phase coherence asymptotically. Our scheme brings a new idea to solve the feedback-based distributed beamforming problem, and has better convergence speed than the older strategies. Our simulation results have shown that the convergence steps of both algorithms are linear with respect to the number of agents in the group. In addition, the noise has little effect on our algorithms.Finally, we discuss the optimal beamforming problem based on a mobile assister node without direct feedback information, especially at the case that the receiver initially is out of the communication range of the transmitters. By assuming that the distance to the re-ceiver and the direction-of-destination (DoD) can be estimated, a stop-and-go strategy is proposed for a mobile assister node such that it moves gradually along the direction to-wards the receiver and provides SNR feedback information for transmitters. By receiving the SNR feedback information at each step, we utilize the feedback based optimal beam-forming algorithm above for the transmitting sensors to accelerate updating their phases to maximize the SNR at the assister end. We then show that when the assister node reaches a threshold distance, the maximum SNR at the assister node implies a large enough SNR at the receiver end such that the receiver is within the communication region. Our algorithm can handle the distributed beamforming without requiring the links between the destina-tion and the sensor array, which rarely exists before. In addition, the algorithm has a rapid convergence rate to ensure the communication requirement at the receiver compared to the existing non-feedback beamforming. At last, our algorithm has a certain DoD error tolerance.For the proposed analysis and control design methods, rigorous theoretical proofs are presented. Several Matlab simulations are also given in each chapter to illustrate our results.