Research On Path Planning For Autonomous 3D Reconstruction System Based On Robot

With the development of information technologies,such as internet of things,big data,mobile applications,and so on,the industrial revolution began to put on the agenda in the whole world.At the same time,industrial transformation began to enter the substantive stage.In China,with the strategies of intelligent manufacturing,manufacturing Chinese 2025,the industry also began to make snap reform.The digital factory is the foundation of the future intelligent factory,while the visualization technology is an important part of the digital factory.And the key technology of visualization technology is 3D reconstruction.However,the current 3D reconstruction technology is relatively backward.Focus on this issues,this paper puts forward the scheme of the autonomous 3D reconstruction system based on the robot.In this system,the main task of this paper is to design the robot path planning system.However,the robot's onboard hardware resources are limited.Therefore,this paper will divide the path planning system into three parts which are global path planning subsystem,local 3D dynamic mapping subsystem and local path planning subsystem to reduce the demand of hardware while improving the efficient of the system.In the global path planning subsystem,this paper proposed a 2D voronoi based 3D random trees path planning algorithm while considering the disadvantage of computational expensive of 2D graph based path planning methods and incompleteness of sampling based path planning algorithms in complicated 3D environments.At the same time,this paper designs a simplified sphere collision detection model and detection method to reduce the consumption of computing resources.Based on the existing line of sight algorithm,this paper proposed the reverse line of sight algorithm.To demonstrate the effectiveness,efficiency and robustness,the proposed method is examined and compared with RRT algorithm in various complex 3D environments.In the local dynamic map reconstruction subsystem,this paper created a 3D map according to the scrolling grid algorithm.At the same time,this paper do ray tracing and grid value updating with the help of bresenham algorithm for each point of the 3D point cloud which are received by real-time scanning sensor.After that,a 3D distance map is created and be updated in real time with the incremental brushfire algorithm.So that more environment information can be provided to the 3D local path planning system.This paper implemented some examinations based on the ground differential wheeled robot platform and analyzed the system property by comparing the local 3D map with the 2D and 3D static global map.In the local path planning subsystem,this paper made local plan by using the real-time matching,evaluating,optimizing and curve fitting strategy on offline trajectory set.In the evaluation phase,this paper designed an evaluation function based on the obstacle information in local map,global path information and local target information to select the optimal path from 29 local trajectories.In the optimization phase,this paper used the gradient attenuation method to adjust the local trajectory by combining the local distance map,so that it can avoid the dynamic obstacles better in the real environments.In the curve fitting stage,this paper adopts the polynomial fitting method.The seven order polynomial of time t is generated by considering the smoothness of the velocity,acceleration and kinematic constraints.Then,it is calculated by the derivative method that the velocity and acceleration commands for robot trajectory tracking.At the last,this paper examination the whole path planning system on the ground differential wheeled robot platform by combining the global path planning subsystem and local dynamic map reconstruction subsystem.