Research On Low NO_x Emission Path Planning For Diesel Freight Cars

With the growth of motor vehicle ownership,the exhaust emissions generated by motor vehicles have become an issue that cannot be ignored.Therefore,how to reduce the exhaust emissions generated by vehicles from their departure to their destination has become an important issue.In existing research,in addition to applying technological means to reduce vehicle emissions,emissions are also reduced by guiding vehicles to travel,such as selecting the possible path with the least amount of pollutants emitted by the vehicle during the driving process through navigation guidance.At present,navigation aims to reduce exhaust emissions during vehicle travel by minimizing the time,distance,and number of traffic lights.However,due to the complex and variable road conditions and not directly targeting emissions,the route may not necessarily be the path with the least emissions.Therefore,this article focuses on two aspects: dynamic road environment and minimizing exhaust emissions,in order to achieve path planning for reducing vehicle emissions in dynamic road environment.Firstly,an overview of the "green" path,vehicle emissions,and path planning models is conducted to determine the route of "modeling before planning".By combining qualitative and quantitative analysis of the factors that affect emissions,correlation and regression analysis are conducted between emission data and influencing factors,and an emission model is developed and validated with actual driving data.The research shows that the predicted results of the fitted emission model are slightly higher than the actual data,and lower than the predicted results of the CMEM(Comprehensive Modal Emission Model)model,indicating that the demand for prediction is met.Secondly,DDQN(Double Deep Q-learning Network)is suitable for dynamic problems and has self-learning characteristics.At the same time,it is rarely used in the field of map navigation,so this method is selected for path search.The components of DDQN are designed separately to be suitable for the scenario in this article,where the action set is the type of turn,the environment is the state of the road network,the state set is the information obtained by the agent at the node,and the reward function is designed according to the greedy strategy.Based on this,an improved A * algorithm in existing research is introduced to ensure that DDQN can converge directionally,and the emissions at intersections are estimated by referring to existing literature,During the interaction with the environment,update the road condition data obtained by Baidu API and update the emission weight values in the state set according to the constructed emission model.Finally,the algorithm provided planning paths for both dynamic and static environments after 5000 training sessions.In a static environment,DDQN converges on the same route as A* and Dijkstra,reducing the theoretical emissions by 0.343 grams compared to the actual path and shortening the distance by 4.131 kilometers.In a dynamic environment,the total NOx emissions of the path planned by the DDQN algorithm are the least,at 12.29 grams,which is5.64 kilometers shorter than the actual path distance.The emissions are reduced by 0.249 grams,1.101 grams lower than the shortest path emissions,and the distance is increased by3.431 kilometers.This article provides a reference for the study of dynamic environmental green emissions.