Optimization Design Of Autonomous Vehicle Path Planning Algorithm Based On Localizability Estimation

As a kind of wheeled mobile robot,autonomous vehicle is an intelligent product in the fields of machinery,electronics,sensors,information and other disciplines.It mainly relies on the hardware and software system of the vehicle to realize autonomous driving.Major research institutes,enterprises,colleges and universities around the world are throwing their full weight to study the technology of autonomous vehicles,which symbolizes the country's scientific and technological strength.The arithmetic level of autonomous vehicle technology can be divided into environmental perception layer,decision-making planning layer and motion control layer.In this thesis,the environmental perception layer and decision-making planning layer are jointly studied.Existing path planning algorithms are often able to calculate excellent results,but there are areas with poor positioning performance in the actual environment,which creates the positioning error in the path tracking process of the autonomous vehicle or makes it inaccurate,so that the planned path cannot be put into operation or would even lead to an accident.In order to solve these problems,this thesis proposes an optimal design scheme of path planning algorithm based on localizability estimation results,so as to provide a better positioning path for autonomous vehicles.The main contents are as follows:1)The localizability estimation algorithm is studied.Based on map matching.According to the observation results of the robot in the map and the probabilistic grid map model,Fisher information matrix,which can take into account the changes of different pose parameters of the robot at the same time,is selected.which is deduced to the Cramér-Rao lower bound and analyzed,and the optimal distribution of the localizability estimation results is obtained.The information theory and the entropy rule are studied,sigmoid function is selected to integrate the above distribution results into the location information entropy,and a value which can describe the location is obtained,and it is calculated as the cost function of the path planning algorithm.2)Based on the classical A* algorithm,a new cost function is designed considering its localization.Thus,a path planning algorithm based on localizability is proposed,which takes into account path length,localizability performance index and so on.Different weight parameters are studied,and the effects of existing consumption,estimated consumption and localizability are analyzed,and the optimal combination of parameters is given.3)Design and implement the simulation and experiment based on pioneer-3DX wheeled robot and matlab: verify the localizability effect under different maps;compare the path planning effect with classical A* algorithm,prove the validity of the algorithm in this thesis.The results of simulation and experiment show that the path planning algorithm based on localizability estimation has certain practical application value,and provides a reference for further development of related algorithms.