Research On Speed Control Method Of Connected Vehicles Based On Personalized Demands

In recent years,the economic,environmental,and safety issues caused by traffic congestion problems have become a growing concern.Vehicle speed control is considered to be an important way to relieve traffic pressure and increase the efficiency of road traffic.With the help of vehicle-to-vehicle and vehicle-to-everything technologies,connected vehicles can sense the surrounding environment in time,making real-time speed control possible.Investigation have shown that different groups have diversified preferences,including comfortability,safety,rapidity and so on.Therefore,it is important to provide personalized travel speed recommendations to enhance their travel experience and increase the efficiency of access.Therefore,it is significant to provide personalized speed recommendations according to the diverse travel needs of users to enhance their travel experience and improve traffic efficiency.For research work related to speed recommendations,the concentrate has less been on the variability of user travel and demand.To tackle the problem,this thesis intends to model the problem using a multi-objective Markov decision process with jerk,time to collision,and harmful gases emission as objectives,and use linear multi-objective optimization to approximate the Pareto optimal solution frontier for two challenges:information fallibility for feature extraction and network fusion,and exponential growth of the problem scale.Therefore,this work designs a holistic solution for personalized vehicle speed recommendation,including:1)An algorithm for foreground vehicle speed prediction based on spatio-temporal data is designed,which could infer the driving strategy of the vehicle preceding and reduce the impact of rapid acceleration and deceleration of the vehicle ahead on the vehicle itself;2)A Fusion-based guidance single-model multi-strategy vehicle speed recommendation method is proposed,which could reduce the time consumption and computing source,by fusing the explicit guidance of reward function with the hidden guidance of preference network to realize the mapping from single-model objective to region of strategy;3)A mobile speed recommendation system is designed and implemented based on Android platform,and it uses the Gaode Navigation API to assist users to travel.Experimental results show that the aforementioned solution can be adapted to the respective optimal policy based on user-defined target preferences,and reduces resource requirements at an exponential scale.