A Personalized Lane-Changing Decision-Making And Trajectory-Planning Method Based On Safety Field For Intelligent Vehicles

Intelligent vehicle has become one of the most promising development tendencies of automobile,and it is expected to provide more safe,economical,comfortable and convenient travel plans in the future.As an important part of the upper framework of intelligent vehicle,there has been considerable mount of work done on decision-making and trajectory planning algorithm,but little takes personalized needs of driver or passengers into account.In reality,there are obvious differences among different drivers in decision-making and trajectory planning behavior.Take car-following and lane-changing conditions for example: users of radical style prefer to keep a close gap between cars and make a lane change swiftly,while users of cautious style are inclined to keep a loose following distance and make the course of lane change gentle and smooth.Neglecting these differences in decision-making and trajectory planning will lead to deviation from driver or passengers' expectation,thus reduce the ride experience.Supported by the National Key R&D Program of China(2016YFB0100904),National Natural Science Foundation of China(51775235),and Jilin Province Science and Technology Development Plan Projects(20170101138JC),this paper conducts research on safety-field-based personalized lane-changing decision-making and trajectory planning method on the basis of drivers' field test data.This paper has carried out the following tasks,(1)Driving data acquisition testIn order to obtain drivers' data under car-following and lane-changing conditions in real environment,a driving data acquisition platform based on real vehicle is built with highprecision inertial navigation system and data acquisition system.Then,according to the demand of data acquisition,five typical working conditions including following,lanechanging and overtaking are designed,and driver samples are selected from multiple dimensions to conduct the data acquisition test.Finally,diversity of drivers in decision-making and trajectory planning behavior is verified by statistical analysis on parameters like following time headway,triggering time of braking and lane changing,braking deceleration,and lanechanging duration.(2)Personalized lane-changing decision-making method based on safety fieldBased on the analysis of braking process,a safety field with relative speed and relative distance between vehicles as input and critical deceleration needed as output is established to characterize the driving safety.Then,given the difference of drivers' operating habits and sensitivity to danger,different decision-making parameter thresholds are set up to personally make lane-changing intention and feasibility judgement.Furthermore,the decision-making method is made adaptive to multi-vehicle environment by adding the consideration of contralateral rear vehicle and speed regulation before the lane change.Finally,Car Sim and Simulink/Stateflow are used to carry out off-line joint simulations under various conditions,which show that the lane-changing decision-making method can adapt to different drivers' lane-changing habits while ensuring safety and comfort.(3)Trajectory planning algorithm for structured road based on safety fieldThis section makes research about planning tasks for structured road from two aspects: macro path planning and micro lane-changing trajectory planning.For the former,Dijkstra algorithm and A* algorithm are used for path planning,and traffic environment constraints are added to make the searching module suitable for dynamic driving environment.For the micro lane-changing trajectory planning,curves like polynomial curve,B-spline curve and trapezoidal acceleration curve are analyzed and compared,then a personalized lane-changing trajectory based on clothoid curve is designed.Finally,taking preview optimal curvature model as controller,this section conducts off-line simulations to verify the advantage of clothoid lane-changing trajectory in terms of tracebility and comfort.(4)Trajectory planning algorithm for unstructured road based on safety fieldOn the basis of RRT and RRT* algorithm,an improved RRT* algorithm adapted to unstructured road environment is established by adding target-biased strategy,bidirectional search strategy,and node expansion strategy,which is under angle constraint as well as safety field guidance.Next,the searching result of improved RRT* algorithm is processed by pruning and then smoothed with B-spline curve,so that the trajectory is shorter and smoother.Finally,RRT algorithm,RRT* algorithm and improved RRT* algorithm are compared by simulation,whose results show that the improved RRT* algorithm can generate smooth trajectory that meets the heading angle constraint,while the searching efficiency can be significantly improved.