Research On Path Optimization Of Guiding Signs Based On Deep Learning Algorithm

As an important traffic control facility,guide signs are the essential constituent of urban traffic system.It plays the role of conveying road information to drivers.To optimize the existing method of the guiding path planning,the information overload problem of guide signs and the guiding path planning were studied in this paper.The propose of the guiding planning is to meet the demand of drivers’ trips and reduce the management cost of guide signs.Thus,the objective function was built by the drivers’ trips cost and the guide signs cost.The drivers’ trips cost was defined by using three variables: path length,road traffic flow and vehicle turning frequency,which were the normalized factors that affect drivers’ trips between the origin and destination of the guiding path.The guide signs cost was related to the number of guide signs on the path.Then,the objective function was obtained by combining the two costs.After the objective function of the model was established,the guidance path planning model was built based on A* algorithm.In the modeling process of the guiding path planning,the process of drivers’ visual cognition of guide signs in the real driving scenarios was not considered,which may mislead to biased guiding path planning results and the information overload of guide signs at traffic nodes.Therefore,for improving the practicability of the guiding path planning model,a static visual recognition experiment was carried out by using E-Prime software to analyze the quantitative relationship between the information content of guide signs and the visual recognition time in different scenarios.And a penalty function was designed based on the result.After the penalty function was added to the objective function of the model,if the information content of the guide sign at a node searched by the algorithm exceeds the set threshold value,a reasonable penalty value will be given to the current guiding path cost,which enabled the model to objectively judge whether it is necessary to avoid searching for this node.There are the demand planning guiding paths between multiple starting and ending points in the actual road network.In order to enhance the application value of the model proposed in the paper,a multi-guiding path planning model was proposed based on DQN algorithm.This model can solve the multi-guiding paths planning in two scenarios which were from single starting point to multiple POI,and from multiple starting points to single POI.Finally,the advantages and disadvantages of the guiding path planning scheme obtained by the guiding path planning model and the multiguiding path planning model were compared through a case study.It was concluded that the guiding path plans calculated by the multi-guiding path planning model had lower total cost.