Flexible grasping of robot arm based on improved Informed-RRT star

In order to grasp objects and avoid obstacles, a Grasping Informed-RRT* algorithm is proposed based on the traditional Informed-RRT* algorithm. Firstly, the three-finger pneumatic flexible gripper is selected, the flexible grasping module is designed, and the autonomous grasping system of the manipulator is constructed. Secondly, the generated residual convolutional neural network (GR-ConvNet) model is used to predict that the color image and depth image acquired by the depth camera are input and the appropriate mapping capturing pose of the object in the field of view is output. Finally, the maximum number of iterations and the adaptive function are set in advance, and the path is constrained by quadratic B-spline curve to generate the collished-free optimal trajectory of the manipulator. In order to verify the grasping effect of the robot arm, the simulation experiment and the grasping experiment on the cooperative robot arm FR3 were carried out respectively. The results show that compared with the traditional Informed-RRT* algorithm, the improved algorithm can shorten the trajectory length by 10.11% and the trajectory generation time by 62.68%. The robot arm can avoid obstacles and grasp the target object independently, which meets the requirement of autonomous grasp of the robot arm.