Abstract:

We propose an optimization model and its corresponding parallel algorithm for optimized signal-planning. Based on a local enumeration method, the algorithm adds a virtual guiding penalty, which is derived from the traffic weight. This makes the algorithm has a global spatial and temporal optimized property. In the situation of saturated or over-saturated traffic status, this algorithm outperforms the traditional single-point intersection control method.

Key words: neural network, signal timing optimization, dynamictraffic weight, parallel algorithm