Research On Global Path Planning And Local Obstacle Avoidance Methods For Unmanned Vehicles

With the development of technology in recent years,unmanned vehicles play a very important role,whether it is automatic driving on urban roads,military supplies in complex road conditions,or storage and express logistics transportation.Path planning,as the core technology of unmanned vehicles,has become more of a focus of attention.Global path planning of unmanned vehicles is to plan a collision-free path from the starting point to the target area in the global environment.The unmanned vehicle takes this path as a reference path,and in the process of driving forward along it,the surrounding environment is not unchanging.At this time,the unmanned vehicle is required to use its own sensors to sense the local environmental changes in a certain range around it in real time,avoid various sudden obstacles,and complete the local obstacle avoidance task.In this paper,we propose a method combining global path planning and local obstacle avoidance to enable unmanned vehicles to efficiently and safely complete the driving tasks in such environments.The main research elements of this paper are as follows.First,the global path planning part of the unmanned vehicle.For the problems of blind random search and slow convergence speed in traditional algorithms.In this paper,we design a global path planning algorithm based on restricted sampling asymptotically optimal rapidly exploring random tree,and propose a node search improvement strategy based on K-dimension Tree to obtain a relatively better path on the basis of more efficient planning.For the defects of sharp corners in the path obtained by this type of path planning algorithm,a path post-processing method based on cubic B-spline is proposed to reduce the spikes and use the resulting smooth curve as the reference path of the local obstacle avoidance part.The improvement in path optimality and planning efficiency between the improved algorithm and the traditional algorithm in this paper is verified through simulation experiments,and the effectiveness of the algorithm is verified in a variety of classical scenarios.Second,the local obstacle avoidance part of the unmanned vehicle.A local obstacle avoidance trajectory planning algorithm based on the Frenet coordinate system is designed for the problem that it is difficult to describe the relative position relationship between the unmanned vehicle and the reference path under the Cartesian coordinate system.In this paper,the local obstacle avoidance problem is introduced into the Frenet coordinate system for decoupling,and a series of end-state parameters of the vehicle are sampled in the horizontal and vertical directions respectively,and the set of local obstacle avoidance trajectories is generated using a quintic polynomial.Finally,the trajectory merit part of local obstacle avoidance for unmanned vehicles.In this paper,a multi-conditional screening mechanism is proposed under the premise of intensive sampling to screen out some trajectories that exceed the constraints and improve the drawback of multiple useless calculations in the traditional method.Then the remaining trajectories are added to the set of alternative trajectories,and the optimal trajectories are output according to the cost function merit.To address the problem that the obstacle threat coefficient is difficult to determine in the traditional cost function and the algorithm outputs non-optimal trajectories under the influence of multiple factors,this paper proposes a collision detection algorithm based on triple safety enveloping circle,reconstructs the cost function,and uses the collision detection screening mechanism with the nature of iterative merit selection to make the final obstacle avoidance trajectory converge to the optimal solution.The improvement of this paper’s improved local obstacle avoidance method in terms of computational efficiency and path optimality is verified by simulation.