United Game Under The Framework Of Multi-agent Reinforcement Learning Algorithm

MAS is currently a hot research field of artificial intelligence ,which is a complex, dynamic environment. Problem solving system is huge. One of the main features of intelligent system is able to adapt to unknown environments, where learning is the key technology of intelligent systems.According to the feedback of different learning technologies, machine learning can be divided into supervised learning (Supervised learning), non-supervised learning (Unsupervised learning) and reinforcement learning (Reinforcement learning). However, there are learning difficulties Agent: Agent is only part of the perception of the environment, a huge search space, learning efficiency is not high, in fact, most of the multi-Agent is a learning algorithm for a single Agent's, any of them can not effectively solve all of the above problem, therefore, a variety of learning algorithms in the consolidated basis, the paper made the following tasks:On the basis of the analysis of the structure and principle of autonomy negotiation model, the Causes of Negotiation impasse, and Effectiveness of consultation,Q-learning has been used to eliminate the deadlock of multi-issue bilateral negotiation process. The learning process in Multi Agent System has been Supported. Experimental results showed that the model was feasible and effective.On the basis of the analysis of the structure and principle of self-government negotiation model, the Causes of Negotiation impasse, and Effectiveness of consultation. Q learning has been used to eliminate the deadlock of multi-issue bilateral negotiation process. the learning process in Multi Agent System has been Supported. Experimental results showed that the model was feasible and effective.