A Hmm-based Vehicle Trajectory Prediction Method Considering Space-Time Constraints

Vehicle trajectory prediction has become a hot research problem in many fields.For individual travelers,trajectory prediction can help optimize travel routes,make the best travel decisions,avoid traffic congestion,improve travel efficiency and reduce travel costs,and also improve travelers’ driving experience.For the traffic system,vehicle trajectory prediction can sense the traffic operation status and development trend in a period of time in advance,providing the basis for the optimization of the operation status of the traffic system,thus alleviating the congestion problem of the whole traffic system and promoting the efficient,healthy and sustainable development of the traffic system.At the same time,with the development of urban transportation in the direction of autonomy and intelligence and the maturity of automatic driving technology,it is especially important to sense the changes in the surrounding traffic environment and its future development.So,vehicle trajectory prediction has very important potential application value in future transportation.Most of the current vehicle trajectory prediction methods focus on the small spatial and temporal scales,and there are fewer studies on large spatial and temporal scales,and the prediction efficiency of the existing methods is unsatisfactory.Therefore,in this paper,we propose a vehicle trajectory prediction method considering spatiotemporal constraints on the basis of the traditional Hidden Markov Model,which optimizes the training data set of the prediction model by calculating the potential path areas of vehicles under the conditions of determined travel origination,destination and travel expected time,and discarding the historical trajectories of vehicles under special travel demand,so as to greatly reduce the scale of the Hidden Markov Vehicle Trajectory Prediction Model,while reducing the interference factors in the decision making process of the prediction model,and finally improving the prediction efficiency of the vehicle trajectory,while making a positive contribution to the prediction accuracy.In order to verify the effectiveness of the proposed method,we train,validate and test the model using 25 days of cab trajectory data under real road conditions in Chengdu city.The experimental results show that,when solving the vehicle trajectory prediction problem at large spatial and temporal scales,the proposed method improves the prediction efficiency by about 30% compared with the traditional Hidden Markov Model-based vehicle trajectory prediction method under the same conditions,and its prediction accuracy is also improved under under the constraint of 2 to 3 times of the minimum travel time.