Traffic Disturbance Mining and Feedforward Neural Network to Enhance the Immune Network Control Performance

Traffic disturbance in urban cities challenges the most advanced traffic signal control systems (TSCS). The challenge is mainly related to the capability of TSCS to ensure a quick detection and to suggest suitable decisions. Neural network has shown great potential in predicting traffic disturbance. In addition, smart clustering could be beneficial to ensure fast disturbance reaction while TSCS are providing control decisions. Moreover, the immune network approach has succeeded in controlling interrupted intersections. Motivated by these assumptions, we propose in this paper a disturbance mining approach based on the occurrence of traffic disturbances to ensure optimal signals control that minimizes traffic delay. Initially, the queue delay is calculated based on mutual information of different traffic scenarios. At that point, within the maximum traffic delay constraint, the feedforward neural network is considered to find the optimal traffic delay and maximize traffic fluidity. As a result, disturbances and related control decisions are clustered based on the calculated traffic delay. Our approach helped the immune network control system (INCS) by prompting it with faster reaction and lower traffic delay compared to its classical version.