An Improved McMaster Algorithm For Recurrent And Non-recurrent Congestion Detection In Freeway

Accurate detection of recurrent congestion and non-recurrent congestion is an important way to ensure traffic safty and improve traffic operation efficiency on freeway.McMaster Algorithm,one of the most classical congestion detection algorithms,is used in practice.McMaster algorithm uses traffic flow and occupancy data,combined with manual setting parameters,to establish a two-dimensional model.However,it ignores the speed,and the artificial preset parameters are difficult to be applied to different roads,so the accuracy of the model is low and the transplantation is poor.Therefore,based on the three parameters of traffic flow,occupancy rate and speed,it is of great significance to establish a three-dimensional McMaster model for recurrent and non-recurrent congestion detection on freeway.Based on the McMaster algorithm,this paper establishes the objective function between McMaster algorithm parameters and data classification accuracy rate,using particle swarm optimization algorithm to optimize the objective function and doing research on the self optimizing method of McMaster algorithm.At the same time,based on the optimization of the McMaster algorithm model,this paper considers the catastrophe of the speed and uses the cusp catastrophe theory to establish a three-dimensional optimization McMaster modelAiming at the poor portability of McMaster algorithm,this paper proposes an explicit discrete objective function between parameters of McMaster algorithm and data correct classification rate and chooses patical swarm optimization algorithm as the optimal algorithm.Considering the imbalance of traffic data,this paper improves partical swarm optimization algorithm and raises gradient based partical swarm optimization algorithm(GPSO).The experiment results show the parameters that are found by GPSO make McMaster algorithm perform high detection rate and meet portability requirements.Aiming at the low detection rate of McMaster algorithm,this paper proposes a speed – volume – occupancy three dimentional optimal McMaster algorithm model based on cusp catastrophe theory.Firstly,this paper analyzes the feasibility of establishing McMaster model based on cusp catastrophe theory from the qualitative point of view.Then,it is proved that the McMaster model can accurately reflect the traffic flow data from the quantitative point of view.The experiment results show that this model can improve detection rate by dynamicly reflecting the evolution trend of recurrent and non-recurrent congestion.Aming at the the existence of misjudgment of three dimentional optimal McMaster algorithm model,this paper deeply analyses the cause of misjudgment,proposes the establishment of ‘misjudgment area' and extract ‘catastrophe angle' as a new feature to detect recurrent congestion and non-recurrent congestion.Considering that the catastrophe angel of recurrent congestion is small while it is huge when non-recurrent congestion occurs,this paper utilizes outlier detection algorithm to solve this problem.The experiment results show that this algorithm,as a complementary algorithm of three-dimensional optimization McMaster algorithm,can play the role of second detection.In summary,this paper makes a series of improvements to the McMaster algorithm,which improves the portability of the McMaster algorithm,and enhances the detection accuracy of the McMaster algorithm.The experimental results show that the improved McMaster algorithm is feasible and effective to distinguish the recurrent and non-recurrent congestion.