Theoretical Research And Application On Multi-Agent System For Distributed Constraint Optimization

With the development of internet technology,there are some optimization problems which can not be solved faultlessly by traditional methods,named distributed optimization.The goal of distributed optimization by using multi-agent systems is: each agent only communicates with its neighbors to minimize(maximize)the global objective function.Each agent has a local objective function,and the global objective function is the sum of all the local objective functions.Due to the complexity of constrained optimization problems,most of the current research results are based on unconstrained distributed optimization.Yet,how to better solve more complex optimization problems with constraints is a difficult problem.Most of the literature mainly consider the multi-agent system how to solve the optimal solution,while the limitations in communication and other aspects are seldom considered.In order to deal with the distributed optimizations which have limitations in commnunication,we has completed the following work : Based on the general form and solving method of the distributed optimization problem,we study more complex distributed constraints optimization problems,and consider four communication problems in communication graph,such as,communication time delay,event-triggered communication,and quantized event-triggered communication.The following is the main content of this dissertation.According to the difference in dealing with constrained problems,the general models of constrained distributed optimization problems are summarized,and corresponding solving methods are introduced for different models.This dissertation presents two specific models named primal separable model and dual separable model,and introduces the solving methods for the two types of problems.And this dissertation analyzes how to reduce the complexity of communication network if there are too many agents.After considering the multi-region problem,the corresponding multi-agent system is designed.It is proved that the system can converge to an optimal solution of the optimization problem.Finally,the correctness and advantages of the proposed system are verified through simulation results in the economic dispatch of smart grid.The multi-agent system with communication delay is studied.First,how to reduce the communication information between agents is considered.Then,with take bound and equality constraints into consideration for distributed optimization,the multi-agent system with communication delay is considered.The inequalities and variable transformations techniques are used to get an upper bound of the communication time delay and the convergence conditions of the multi-agent system under this bound are estimated.The communication graph in the existing results is extended by using the proposed method.A multi-agent system with event-triggered communication is considered.The distributed optimization with intermittent communicating mechanism is further considered,and the discrete communication protocol of the continuous system is studied in the case of considering the communicating capability and energy consumption of the agents.It is proved that the multi-agent system with event-triggered communication designed in this dissertation can converge to an exact solution of the problem with constraints.And the application in economic dispatch for smart grid verifies the performance of this system.On the basis of event-triggered communicating mechanism,the problem of quantizing communication information in multi-agent system is studied to solve the distributed constrained optimization problem.Based on the study of reducing the communicating information and the frequency of communication,the problem of reducing channel occupation and reducing the total amount of communicating information is further considered.Based on event-triggered communicating mechanism,we quantify the communicating information before it was spread to reduce the transmission of total communicating information and the occupation of the channel.The distributed system of quantized communication can converge to an optimal solution of distributed constraint optimization under the time-varying quantizer.Finally,this dissertation summarizes the research on distributed constrained optimization,and looks forward to the future research directions.The research works in this dissertation provid theoretical and practical references for further discussion of distributed constraint optimization problems that are closer to practical applications.