Motion Planning System Of Mobile Manipulator In Unknown Environment

In recent years,with the rapid growth of logistics business brought about by the rapid development of e-commerce,intelligent logistics equipment has shown irreplaceable advantages in improving efficiency.Intelligent transport equipment has become an indispensable tool for the logistics industry.With the introduction of robots in the logistics industry,we have discovered that the biggest difference between robots and humans lies in the perception of the environment.Even if artificial intelligence continues to develop in recent years,the robot's perception of the environment remains weak.Once the environment mutates,it is very likely that the robot will cause widespread paralysis and even cause a factory safety accident.Uncertainty and diversity of storage scenarios require that mobile robotic arms as handling equipment have better environmental awareness and obstacle avoidance motion planning capabilities.For this reason,this paper focuses on the task of recognizing and grasping an object in an unknown environment and placing them accurately at a given position.This paper studies the construction of discretization model based on three-dimensional vision sensors and the planning of obstacle avoidance motion based on the construction model.Based on RRT*,a fast convergence RRT* improved algorithm was proposed.Implement real-time modeling of obstacle environments and robot-free collision trajectory planning.The main contents are as follows:1.To build a mobile robotic hardware platform,this paper builds a mobile robotic system using industrial robots,mobile robots,and RGB-D vision sensors,and installs a visual sensor(Xtion)on the back of the entire mobile platform as an environment-aware sensor to the entire mobile machine.The arm performs kinematics modeling and analysis and derives the corresponding position relationship between the robot arm and the mobile platform in the robot coordinate system.2.Using visual sensors to build the model of obstacle avoidance scene,first use the robot hand-eye calibration algorithm to calibrate the vision sensor,obtain the coordinate conversion relationship of the camera with respect to the robot,and then build the visual sensor behind the mobile platform to obtain the entire scene cloud data.The calibration result is that the three-dimensional conversion relationship transforms the point cloud data into a robot coordinate system,and finally rasterizes the point cloud data to obtain a discretized model of the working scene,which is an obstacle space for the robot arm motion planning.3.Based on random sampling manipulator motion planning,This paper uses a fast convergence RRT* improved algorithm based on RRT* planning algorithm to carry out planning experiments,and improves the iterative speed of RRT* algorithm in path optimization.A mobile manipulator simulation platform is built in Gazebo.Known obstacles are added to the simulation environment to verify that the RRT planning algorithm can get obstacle avoidance motion trajectories,and then add unknown obstacles to obtain visual sensors.The rasterization model verifies that the RRT planning algorithm can get obstacle avoidance motion trajectories.4.Completed the experiment in a real environment.Firstly,the whole mobile manipulator is driven to reach an unknown working scene.The image processing algorithm is used to obtain the 3D pose of the target object.Then the vision sensor is used to obtain the obstacle space model.Finally,drive the manipulator to complete the obstacle avoidance motion according to the planned trajectory by the ROS robotic motion control interface.Realize the motion planning of the mobile manipulator and complete the grab and place operation experiment in unknown environment.The experimental results verify that the system constructed in this paper is feasible.