Traffic Congestion Pattern Detection Using An Improved Mcmaster Algorithm

Traffic congestion has been an important problem all over the world. Advanced transportation management system (ATMS) that provides information for traffic control and management addresses this problem. McMaster Algorithm, one of the most classical congestion detection algorithms, has been widely used in practice. However, it still has some limitations such as difficulty of determining its parameters which are lower bound of uncongested data (LUD), critical occupancy (Ocrit) and critical volume (Vcrit). This paper transforms this problem to be an optimization problem and presents a new method that integrates gradient descent algorithm and particle swarm optimization algorithm (GPSO), with which parameters of McMaster Algorithm are extracted from the occupancy and volume of the detector. Thus it can determine the parameters of McMaster Algorithm more quickly and precisely. The parameters of McMaster algorithm are validated with real data from Yuwu Expressway in Chongqing. The results show that GPSO can help operators find the best parameters more effectively and make McMaster Algorithm yield a higher accuracy rate and recall rate.