摘要
从交通流状态的模糊特性出发,设计基于Sugeno型神经模糊系统的交通流状态预测算法.选择交通流状态的影响指标作为模糊推理系统的输入、交通流状态作为输出;据经验对输入、输出划分模糊子集,给出相应的隶属度函数并制定模糊规则;建立具有5层结构的神经模糊推理系统,利用神经网络优化调整模糊推理系统的隶属度函数和模糊规则.仿真实验表明,神经网络可直接优化模糊推理系统的隶属度函数,通过对连接权值的训练间接优化模糊规则,故Sugeno型神经模糊系统相比常规模糊系统具有更好的交通流状态预测性能.
According to the fuzziness of traffic flow states,a traffic flow state-forecasting algorithm based on Sugeno neural fuzzy system（NFS） is proposed.In this algorithm,a number of traffic parameters are chosen as inputs,and the traffic flow states are taken as output of a NFS.The fuzzy subsets of inputs and output are given empirically.In addition,the corresponding membership functions and fuzzy IF-THEN rules are also built up by experience.A 5-layer NFS is presented in the given algorithm;and a neural network is used to optimize the fuzzy inference system（FIS）.The experiment shows that neural network can optimize the membership functions directly and the fuzzy rules indirectly.Hence,the Sugeno NFS is more effective than the normal FIS in traffic flow state-forecasting.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2010年第12期1637-1640,共4页
Control Theory & Applications
基金
广东省科技计划资助项目(2009B010800052
2009B090300388)
广东省教育厅"育苗工程"资助项目(LYM08053)
关键词
神经模糊系统
交通流状态预测
动态交通管理
neural fuzzy system
traffic flow states forecasting
dynamic traffic management
