Research On Intelligent Vehicle Obstacle Avoidance Algorithm

Nowadays,with the continuous progress and development of national science and technology,more and more intelligent vehicle terminals have been designed and developed,and have been put into various fields of national enterprises.Mature vehicle radar technology has become a necessary condition for realizing automatic driving,and the basis for improving vehicle safety performance.Therefore,the research on vehicle radar obstacle detection has important engineering application value.Taking the obstacle avoidance and path planning technology of ground autonomous vehicle as the core,this thesis discusses the key technologies of ground autonomous vehicle,such as environmental perception unit and path planning algorithm.The correctness of the research results in this thesis is proved by practice.The main research contents and contributions of this thesis are as follows:1.Aiming at the problem that the angle deviation is more likely to occur in the process of transmitting and receiving signals from target obstacles by millimeter wave radar,resulting in the inaccurate distance between the measured object and the sensor,a joint calibration method is designed to install and calibrate the lidar sensor and millimeter wave radar sensor to reduce the scanning error as much as possible.2.An obstacle detection algorithm based on lidar is designed,and the common robustness and real-time problems of other similar methods are improved.The proposed algorithm maps the three-dimensional point cloud in polar coordinates,classifies the point cloud through a variety of constraints,finally clusters and segments the remaining point cloud to extract obstacles,and optimizes the isolated points in the final stage.Thus,the overall obstacle detection accuracy and real-time are improved,and the obstacle avoidance operation is completed combined with the obstacle tracking algorithm in the previous chapter.3.Aiming at the problem that it is difficult to realize obstacle avoidance function in complex environment and the robustness of the overall algorithm is poor,an obstacle tracking algorithm based on direction fitting is designed,and the ground information is extracted according to the fitted linear equation.Combined with the boundary information of obstacles,the direction information is estimated,and then the tracking behavior of intelligent vehicles for obstacles is guided.The translation equation is used to guide obstacle avoidance when adjacent to obstacles.Finally,the feasibility of obstacle direction prediction and tracking is verified,and it is found that the algorithm can reduce the amount of calculation and meet the requirements of real-time.4.Aiming at the problem that the environment map of feature points can not be used for the path planning of intelligent vehicle obstacle avoidance,a new method is designed to locate the intelligent vehicle in real time and construct the environment map at the same time.This method can not only obtain the real-time state of the intelligent vehicle,but also construct a two-dimensional high-precision map.The data processing and analysis are realized based on the library in tensorflow software.Finally,the reliability and accuracy of the proposed algorithm are verified based on the experimental results.This thesis contains 34 figures and 5 tables.