| Unmanned driving technology is the product of intelligent science development,and has played an extremely important role in many industries.This paper studies the key technologies of environment perception,vehicle positioning,decision planning and so on.Based on the idea of modularization,the autonomous vehicle that can run stably on structured road is designed from the perspectives of environment perception,central decision and bottom execution.Perception and positioning is the basic premise for the operation of unmanned vehicles.Without accurate perception of the surrounding environment and accurate estimation of its own position and posture,the decision control system will not be able to make judgments and unmanned vehicles will not be able to operate.In this paper,the laser odometer based on NDT(Normal Distributions Transform)algorithm is applied to realize front-end matching,and the sliding window mode of scan to map is used to improve the efficiency of cloud matching.At the back end,a factor graph optimization method based on multi-sensor fusion was adopted to realize the constraint and algorithm optimization at the back end.GNSS(Global Navigation Satellite System)information and loopback detection factor were added to correct the time accumulation error between laser odometer and IMU(Inertia Measurement Unit).The LIO/GNSS/RTK fusion vehicle body positioning was realized based on global pose image optimization.Path planning is to plan an effective driving path from the starting position to the target point according to certain requirements on the basis of a certain environment model.Based on the characteristics of the research in this paper,the path planning problem of unmanned vehicles is explored respectively by using Dijkstra algorithm based on graph search and RRT(Rapidly Exploring Random Tree)algorithm based on sampling.In the unmanned environment with known key road point files,the global route matrix is small and the time and space complexity is low,so Dijkstra algorithm can efficiently traverse nodes in the way of directed graph to obtain the global optimal path.In the unmanned case without waypoint files,the search pressure of Dijkstra algorithm increases rapidly and its efficiency is low,so a more efficient RRT algorithm is adopted to realize the global path planning and design,and the cubic B-spline curve method is adopted to smooth the planned path.Finally,KITTI open source data set and data collected from Caotang campus were used to verify the algorithm and evaluate its effectiveness.Based on the overall architecture of the unmanned vehicle and the design requirements of the software and hardware platform,the unmanned experimental vehicle was built,and the unmanned experiment and analysis of the structured campus road were carried out.The experimental results show that the algorithm and the designed unmanned vehicle can meet the established requirements and realize the autonomous driving function. |
