Research On Intelligent Robotic Grasping System Based On Visual Perception

Since the 21st century,with the development of artificial intelligence technology,robots have become more and more intelligent.In particular,the development of com-puter vision has made it possible for robots to complete increasingly diverse industrial tasks.Traditional industrial robots usually work in a structured environment.The robots perform repetitive,singe,and regular tasks by pre-programming strategy.Once there are some changes in the environment or the production task,the robots need to be pre-programmed again.Therefore,it is of great significance to endow the robot with the visual perception ability.The robot can dynamically adjust its motion behavior in real time according to the visual perception.At the same time,the robot should be able to adjust its motion plan in real time to avoid obstacles.Finally,the robot should be able to adjust the pose of its end-effector based on the visual feedback to ensure the successful grasp of the target object.Based on the above background,this article mainly discusses the vision-based robotic grasping control,including the object recognition and localization,motion planning for obstacle avoidance and the image-based visual servoing grasping and etc.The Kinova Jaco2 j2n6s300 lightweight collaborative robotic arm is used to be the research object.And the main research contents are as follows:1.Build a vision-based robotic grasping platform,and analyze the forward and inverse kinematics model of the robotic arm,as well as the camera imaging model Also conduct the hand-eye calibration for each of the three scenarios:eye-in-hand,eye-to-hand,and planar calibration.These calibration ensures that the position information of the object recognized in the camera coordinate system can be accurately converted to the robotic arm coordinate system.2.Propose and implement a point cloud-based object positioning algorithm and a support vector machine-based object recognition algorithm to ensure accurate recogni-tion and localization of the target object.3.Analyze and implement a sampling-based rapidly-exploring random trees mo-tion planning algorithm to ensure the obstacles avoidance for the robotic arm when it is moving and grasping objects4.Propose a visual servoing grasping control method based on learning from demonstrations.The robotic arm can learn the desired grasping position autonomously,avoiding artificial settings and the fuzzy logic is introduced to dynamically adjust the controller gain,which can accelerates the convergence of the entire system.