利用彩色深度相機建立移動型機械手臂於動態環境中抓取物體系統

Abstract

Nowadays, techniques in robotics are getting mature and people expect robots to handle more complex tasks in our daily life. In high-level services, the ability of the robot to grasp objects is essential. In this thesis, to increase e the robot’s ability of interaction with human and environments, we proposed a novel manipulator grasping planner which emphasize the cooperation between two RGB-D cameras and can adapt to the dynamic environments. We assume that the object to grasp and obstacles are both dynamic. Our motion planner doesn’t need whole environmental model previously, and only requires the object information that the robot is demanded to grasp. The concept of our motion planner is based on the potential field and composed of the attractive and repulsive vectors which are generated by the distances from the manipulator to the target and obstacles respectively. Then the motion planner determines the potential vector according to the attractive and repulsive vectors in a variety of situations. Furthermore, an approach to deal with local minima is contained in the algorithm. For robot control, we take multiple control points into account and apply the potential vector with joint-level control. The mobility and the kinematic constraints of the robot are evaluated in the controller to modify the joint velocities. The experiment platform is a wheeled mobile robot with a 5-DOF manipulator which possesses a close range and a normal range RGB-D cameras as our sensors. Through several experiments, the results show that our framework is valid to accommodate the environmental changes and to grasp the object under various situations.