Research On Heuristic Path Planning Method For Intelligent Vehicle Based On Field Theory

Path planning algorithm is the leading way to achieve the autonomy of intelligent vehicle decision-making,including global path planning and local path planning.This research belongs to intelligent vehicles’ local path planning,a key technology to realize autonomous driving.Existing local path planning algorithms mainly use field theory to make behavioral decisions for vehicles or use sampling and loss functions to solve the optimal path.The Driving Risk Field(DRF)makes decisions by quantifying the driving risk,but there are local minimum points with the approach.State Lattice Planners(SLP)selects the optimal driving path by using the cost function,which requires a large amount of calculation.Firstly,this paper combines the two methods to propose a new intelligent vehicle heuristic real-time path planning method DRF-SLP to solve these problems.Secondly,proposes a Mixed Perspective Risk Field(MRF)from the horizontal and top view perspectives for the driver’s blind spot on the road.Finally,applied MRF to the proposed heuristic path planning method,the main research contents are as follows:(1)Aiming at DRF and SLP problems,a heuristic planning algorithm for selecting low-risk sampling areas for path planning is proposed.The algorithm used batch optimization of the planning target as its operation mode and gave a planning target’s sampling plan according to the DRF’s mechanism.The simulation results show this method can significantly reduce the computing time,avoid the vehicle from falling into the local minimum point while retaining the traditional method’s advantages,and obtain a more reasonable and practical driving plan in the given scenario.(2)Through road driving scenes,the limitations of traditional field methods to quantify risk from the overhead perspective are demonstrated,and a method to quantify road risks from a horizontal perspective based on the blind area of the driver’s field of vision is proposed.The two perspectives are combined to calibrate driving risk jointly.A road simulation driving scene with blind areas of vision proves that this method can effectively calibrate the risk level.Compared with the traditional field method,the calibration risk from different perspectives can select the low-risk target sampling area more accurately,making the path planning result more anthropomorphic.(3)Starting from algorithm testing requirements,a highway scene simulation test environment was built,which can automatically generate vehicles to simulate a long-term autonomous driving process.This paper used this environment to test the two proposed algorithms.The statistical results show that both algorithms can maintain high driving efficiency in highway automatic driving scenarios.The planning method using mixed perspective to calibrate the risk has more stable road driving performance.