Research On Economic Speed Optimization On Hilly Roads Based On Nested Monte Carlo Tree Search

The economical driving on hilly roads is to use the road information(such as the slope angle,the length of the hilly road,etc.)in front of the vehicle to optimize the speed when vehicle passes,it improves the fuel economy of the vehicle by allowing the vehicle to pass by the most energy-efficient driving style.However,for the research of economic speed optimization on hilly roads,the classic method such as dynamic programming(DP)are inverse optimization algorithm,although the optimal vehicle speed can be solved under the premise of knowing the slope of the front road.However,since the reverse characteristics for DP,each calculation is for the entire road length,and calculate from the end of the road to the beginning.So it has the following disadvantages: Complex model structure,large storage space,low computational efficiency and can not calculate in real time.The main work and achievements of this paper are summarized as follows:Firstly,exchanged the problem of economic speed optimization on hilly roads into markov decision process(MDP),and a forward intelligent search algorithm-Nested Monte Carlo Tree Search(NMCS)is used to establish the economic vehicle speed calculation model.Compared with the traditional DP algorithm,the NMCS algorithm has the advantages of high computational efficiency and small storage space,which is helpful for the engineering application of the vehicle speed optimization algorithm.Secondly,NMCS algorithm complete to optimize the economic vehicle speed by four steps as follows: sub-state search,sub-state simulation,simulation chain evaluation and optimal simulation chain update.After years of research and analysis on the vehicle operating conditions database,the research institute has obtained the driver's driving habit for different road environments.According to the driving information,this paper established a experience model that describes the driving behavior of drivers,which is introduced into the sub-state search and simulation of the NMCS algorithm.This two steps of the algorithm are changed,and the invalid search in the algorithm simulation experiment is reduced.Search efficiency provides a guarantee for real-time extraction and optimization of economic speed.Then,how to realize the real-time calculation under long-distance path is studied.Based on the NMCS algorithm and combined with the idea of rolling optimization from model predictive control algorithm(MPC),a rolling optimization algorithm based on nested Monte Carlo tree search,NMCS-R algorithm is proposed and designed.The NMCS-R algorithm is based on the forward estimation of NMCS and the idea of rolling optimization.The whole optimization interval is divided into several optimization domains and rolling domains,and the algorithm is continuously calculated by the rolling domain.In the calculation process,the algorithm does not need to output the optimization result of the entire domain,and only needs to calculate a part of the optimization domain to ensure that the algorithm can output the speed result in real time.Compared with the NMCS algorithm,the NMCS-R algorithm further reduces the amount of calculation,improves the computational efficiency,and realizes real-time calculation on long-distance roads.Finally,the NMCS and NMCS-R algorithms are simulated and verified systematically on Matlab,and the influence of algorithm parameters on the results is analyzed.On the one hand,NMCS algorithm is simulated under the design and real hilly roads,and the simulation result data is compared with the optimal solution of DP algorithm,which shows the feasibility and advantages of the NMCS.The results show that under the two type of hilly roads,the result of NMCS algorithm is approximately optimal,In terms of computational efficiency and storage efficiency,the NMCS algorithm has increased by 50.8% and 97.7%.On the other hand,the simulation experiment of NMCS-R algorithm is did under real hilly roads,which is compared with global dynamic programming and NMCS algorithm.The results show that the computational efficiency and storage efficiency of the NMCS-R algorithm are further improved by the rolling optimization form,and real-time calculation on engineering is realized.The establishment of the new algorithm provides new ideas for the development of energy-saving driving assistance products,and provides a new solution for the economical cruise problem.