Research On Path Planning Algorithm For Unmanned Ground Vehicle In The Complex Environment

Unmanned ground vehicle(UGV)is a kind of intelligent vehicle which can travel independently in various ground environments.It is an important branch in the field of robot research,which involves mechanical,optical,electronic information,computer science and technology,artificial intelligence and automation.Its development has great influence on national defense,society,economy and science and technology.The basic environmental elements for UGV include the surface condition(such as covering medium,vegetation,ride comfort,obstacle and so on),weather and illumination conditions,electromagnetic,traffic signal and sound and so on.Complex environments mean that environmental information may be partially or completely unknown,and may be chaotic and dynamic.If coupled with the limitations of sensor technology and environment perception algorithms,UGV will need to make planning decisions in an environment model which is incomplete,inaccurate,or even concontradictory.Therefore it has high requirement for the flexible of the path planning which is the one of the core technology of UGV.In this thesis,the development status of UGV in worldwise is firstly introduced,and then the common path planning and algorithms are summarized;based on that,the research work is carried out in the following three aspects in combination with the actual project demand of UGV:(1)Static obstacle avoidance algorithm in the condition of partial environment information is unknown;(2)Dynamic obstacle avoidance algorithm without the road constraint;(3)Path planning algorithm for the unstructured road without the road model.The main innovative work of this paper is as follows:(1)Considering the nonholonomic kinematics model of UGV,an improved Morphin algorithm based on fuzzy Q learning is proposed.By constructing a multi-layer Morphin search tree,the search direction of UGV in the unknown area is expanded,and its ability to avoid obstacles is improved.Three evaluation functions,such as through rate,safety and target tendency,are constructed to evaluate the search tree.Fuzzy Q learning is used to learn the weighting factors of the evaluation function,so that UGV can have dynamic behavior.Experiments show that the algorithm can effectively improve the local obstacle avoidance capability of UGV.(2)A collision detection model,which name is collision detection circle,is established,and a dynamic obstacle avoidance algorithm based on collision time histogram is proposed.The collision detection model is used to calculate the collision time for UGV in all drivable directions,then the collision time histogram is constructed.On this basis,the behavior planning module and the speed planning module are designed to get the execution angle and speed for UGV.The simulation results show that the algorithm proposed is effective in complex dynamic environments.(3)Considering that the obstacles on both sides of the unstructured road usually contain the road boundaries information,a local path planning algorithm based on grid map and support vector machine is proposed.The algorithm uses nonlinear SVM to extract the security path from the grid map.Then the continuous multi-frame security paths are projected into the same local coordinate system and the RANSAC algorithm is adopted to estimate the road model.Finally,the final planning path is obtained by optimizing the RANSAC path combined with the real-time state of UGV.The algorithm can effectively extract the road from the local grid map to make up for the condition when the vision based road detection algorithms are interfered by the bad weather and light.