Multi-Agent Theory And Research On Application In Elevator Group Control System

Nowadays, one of the hot points in distributed artificial intelligence research is multi-agent system, which is the further development in distributed problem solving. With the development of multi-agent theory and technology, its application is being expanded. But the theory and application is just starting soon, so there are many issues to be resolved. Among them, the structure of multi-agent is one of the future research directions.In this paper, we design multi-agent EGCS structure based on genetic algorithm and the function of each agent. It is not only used to ease centralized group controller with large amount of computing tasks and low efficiency problem of computing in the past, but also to take advantage of group control algorithm agent to achieve the elevator re-allocation of the multi-objective optimization process, and there is more self-learning and adaptive capacity, and also it can ensure real-time optimal allocation.Firstly, the thesis reviews the development and current status of EGCS, and points out the future research directions of multi-agent theory and application, the basic concepts and characteristics, agent structure and multi-agent system architecture, also the collaboration and coordination methods among some agents, communication as well as multi-agent system development languages and tools and so on.Secondly, we study and design EGCS structure based on multi-agent and multi-objective optimization respectively, and then conduct in-depth analysis. After those two structures are comprehensively improved, we design the multi-agent EGCS based on genetic algorithm. Each elevator is designed as an agent, so that each elevator agent can carry out their own local coordination planning, and calculate the comprehensive evaluation of their function. It's able to strengthen independence of each elevator agent.Finally, we introduce the structure of traffic flow module, traffic flow agent, coordination agent, group control algorithm agent and specific workflow of each agent and so on in the design of the multi-agent EGCS based on genetic algorithm. During the absence of new external calls, coordination agent starts the genetic algorithm in the group control algorithm agent. This is not only taking into account the four goals of EGCS synchronously, but also, completing the re-allocation process.On the other hand, when there're not new calls, the system uses classical scheduling algorithm to re-allocate the elevators by co-ordination and communication among agents, as well as to conduct a preliminary chromosome for initial population.This will improve the operating efficiency and self-adaptability.At last, we implement the simulation about multi-agent EGCS based on genetic algorithm, and compare it with the classical scheduling algorithm, the former one can get better integrated performance index.