Nachricht

Special date: On Tuesday, June 11, 2019, room P 1.4.17, Ke Lu will present the results of his master's thesis with the title:

Gesture-based Control System for a Robot Arm

Abstract:

Nowadays, robot arms are being used extensively to execute many precise or dangerous tasks. They are usually controlled by a joystick or some similar controllers with buttons. However, the traditional control method is complicated and inefficient that requires a lot of experience of the operators. This thesis addresses this problem by proposing a new control method for robot arms using hand gestures recognition. The aim of this thesis is to reduce the difficulty of controlling a robot arm as well as save the learning time for operators. In this thesis, a new control system for the robot arm was developed. It uses a human-machine communication interface between the Leap Motion sensor and the robot arm of a rescue robot, called GETjag. On the one hand, by using a hand tracking algorithm supplied by the Leap Motion SDK, the control system allows the robot arm’s gripper to follow the movement and the orientation of the operator’s right hand simultaneously. And, on the other hand, it can also drive the GETjag’s tracks, flippers, and RGB-D camera by moving the left hand. The new control system is also able to recognize some static gestures, which are trained and classified by a Convolutional Neural Network (CNN). After recognizing a gesture, the control system sends a signal to invoke a continuous action with a preset route, such as rotate an object or open a door. Besides, a first-order low-pass filter using dynamically adjustable parameters is used for smoothing the robot arm’s movement. Furthermore, an anti-self-collision algorithm for the robot arm and a graphical user interface(GUI) for the whole system has also been designed. For evaluation purposes, extensive tests were first carried out in the simulator. After that, the system was also tested on the real robot successfully and has shown a good performance.