Research On Motion Planning Method Of Snake-like Manipulator In Complex And Narrow Space

The Three Gorges Dam is the largest hydroelectric power station in the world and plays a crucial role in ensuring China’s annual electricity generation.The large-scale hydroelectric generator unit is the core power generation equipment of the Three Gorges Dam.Regular maintenance is necessary to ensure its operational efficiency and safety.Inspecting the internal condition of the stator end is an important maintenance project to ensure the safe operation of the generator,and the conventional inspection method is to use an industrial endoscope.However,the industrial endoscope lacks active freedom and cannot perform careful safety inspections of complex parts in the deeper areas of the generator stator end,which poses a safety hazard.To solve this problem,this study proposes using a highly flexible snake-like manipulator to assist human inspection of the generator stator end.However,due to the complex narrow environment and technical constraints of the snake-like manipulator itself,there are still some technical problems in motion planning,such as how to plan the motion trajectory of the snake-like manipulator in a complex narrow space and avoid collisions.This thesis focuses on the motion planning of snake-like manipulators in complex narrow spaces.Firstly,the kinematics of the snake-like manipulator are analyzed,and the kinematic model of the snake-like manipulator is established based on the constant curvature model.The mapping relationship between the joint space,driving space,and workspace of the snake-like manipulator is studied,and kinematic simulation analysis is performed in Matlab.Secondly,in view of the problems of slow convergence speed and failure to generate paths under specific conditions of the RRT*(Rapid-exploration Random Tree Star)path search algorithm,a modified path planning algorithm,GAPF-RRT*(Goal Bias and Artificial Potential Fields-RRT*),is proposed for complex narrow space applications.The algorithm combines the goal biasing approach and artificial potential field method with the RRT*algorithm to improve the efficiency of path planning while ensuring the feasibility of the planned paths.After conducting comparative experiments through Matlab simulations,the feasibility and effectiveness of the algorithm in path planning in complex and narrow environments were demonstrated.Thirdly,to solve the trajectory planning problem of snake-like manipulators in complex narrow environments with incomplete environmental information and inaccurate modeling,this thesis uses the Soft Actor-Critic algorithm based on deep reinforcement learning and trains a policy network model.To address the issue of sparse reward information during training,a reward function based on orientation is proposed,and experiments are conducted using the Stable-baselines 3 framework.The experimental results show that the function can provide positive reward feedback.Finally,a prototype system of a rope-driven snake-like manipulator is built based on modularization,and experiments are conducted on path passing in complex narrow spaces.The experimental results verify the feasibility and effectiveness of the proposed algorithm,which can meet the actual application requirements of snake-like manipulators in complex narrow spaces.