论文部分内容阅读
为信号控制的城市道路交叉口定义一个Agent结构模型,在分析相邻交叉口交通流关联关系的基础上,利用记忆因子δ、学习概率α、交叉口交通流变化概率βi等参数阐述了交叉口Agent间的多遇协调流程。交叉口Agent多遇协调采用部分获利协作策略,其交互策略更多地考虑在获利少于对方时候如何以更加协作的态度进行协调。利用记忆因子δ构建了交叉口Agent多遇历史学习协调算法。以交叉口Agent集合达到协调平衡模式需要的交互次数为性能指标,以数个交叉口相连接的主干道为例分析了δ、α、βi等参数对此策略和算法的协调性能的影响,结果表明交叉口Agent集合达到协调平衡模式需要的交互次数随着α的减少、βi的增加、δ的减少而增加,具有一定的动态环境适应能力和协调能力。
An agent structure model is defined for signalized urban road intersections. Based on the analysis of the traffic flow relationships at adjacent intersections, the intersection parameters of the memory factor δ, the learning probability α, the probability βi of the traffic flow at the intersections, Agent between the process of coordination. Intersection Agent frequently adopts partial profit coordination strategy, and its interaction strategy considers more about how to coordinate more cooperatively when profit is less than the other. The memory factor δ is used to construct the coordination algorithm of historical multi-agent learning at intersection. Taking the number of interaction required to reach coordination and coordination mode of intersection agent as the performance index and the main road connected by several intersections as an example, the influence of δ, α, βi and other parameters on the coordination performance of this strategy and algorithm is analyzed. It is shown that the interaction times needed to reach the coordination and equilibrium mode by the intersection agent cluster increase with the decrease of α, increase of βi and decrease of δ, and have some dynamic environment adaptability and coordination ability.