Research On Optimization Algorithm Of Unmanned Vehicle Path Planning

The path-planning technology of unmanned vehicles has always attracted the attention of the academic community.Technology often brings leapfrog innovation to society and leads to the faster development of other fields.The path planning of unmanned vehicles has important research value.In this paper,the environmental perception technology and route planning technology of unmanned vehicles are studied.The main research contents are as follows:In the study of environmental perception technology,three-dimensional laser radar is used to collect the point cloud data of the environmental maps.Firstly,the environment model map is established by the LIO-SAM algorithm.Then,in view of the advantages and disadvantages of the normal distribution transformation registration method(NDT),which has low accuracy,large error but high efficiency,and the iterative nearest point matching algorithm(ICP),which has high accuracy but low efficiency,a fusion NDT-ICP point cloud registration algorithm is proposed.The point cloud registration process is divided into NDT rough registration and ICP fine registration,filtering processing,Reduce the impact of noise points and ground point clouds on the registration effect;Finally,the voxel grid method was used to remove redundant data and preserve the integrity of the fused point cloud data.Through experiments,it can be seen that the NDT-ICP registration algorithm proposed in this paper can better complete the point cloud registration work and still has a stable and accurate effect in large scenes.In the research of path planning technology,aiming at the problems that traditional Q-Learning path planning algorithm has low learning efficiency,slow convergence speed,and poor path planning effect in the environment of dynamic obstacles,an improved GA-QL path planning algorithm for unmanned vehicles based on traditional Q-Learning algorithm is proposed.Firstly,the exploration factor ε is introduced to balance the exploration-utilization according to the mutation of probability to accelerate the learning efficiency.Secondly,deep learning factors are designed in the update function to ensure the exploration probability of the algorithm.Then the genetic algorithm is integrated to avoid falling into the local path optimization and to explore the optimal iteration step number according to the stage to reduce the dynamic map exploration repetition rate.Finally,the path inflection point of output is extracted and smoothed by the Bessel curve to ensure the smoothness and applicability of the path.Compared with the traditional algorithm,the improved GA-QL algorithm has greater optimization in the number of iterations,efficiency,and path,and can better realize the path planning under the dynamic map.Finally,through the autonomous navigation experiments of unmanned vehicles in indoor and outdoor real scenes,the practical applicability and effectiveness of the proposed GA-QL path planning algorithm are further verified.