面向伙伴选择的模糊Markov博弈控制及仿真研究

Study on Fuzzy Markov Game Controller for Partner Selection Management Simulation

针对不确定条件下的伙伴选择决策问题,把自适应模糊控制系统理论及神经网络理论引入到Markov博弈中,提出一种基于多智能体的伙伴选择模糊控制模型。该模型引入基于ANFIS和神经网络的模糊神经网络,实现了一种全新的进行值函数逼近的梯度下降Q学习的算法。并应用该模型对伙伴选择问题进行研究,对多影响因素进行FNN学习,将输出量作为标准Markov博弈模型的输入量,得到影响的策略,最后研究了一个应用实例,利用具体历史数据对建模方法和模型进行了验证和分析。

According to partner selection under uncertain conditions, a multi-agent fuzzy Markov game controller was proposed based on adaptive neuron-fuzzy inference system （ANFIS）, neural network and Markov game. Fuzzy neural network was used as value function approximators. In this model, FNN was used to train the factors which influenced the partner selection and the results of FNN was taken as the input for the standard Markov game while the finial policy was taken as the output. A case was studied and the simulation model was validated by historic data.