Gesture-based robot's long-range navigation

摘要

In this paper, we deal with the problem of longrange navigating the robot in unknown environment. The robot captures the environment video in remote field, and the video is transmitted over the Internet in real-time and displayed on the control terminal. On the control side, according to the received video, the operator sends control instructions to guide the robot by using predetermined gestures. The gestures are recognized by the sensor, and convert to instructions such as forward, backward, turning left, turning right and so on, and then the instructions are transmitted to the robot over the internet. Therefore, we complete a close-loop feedback control of robot's long-range navigation. In gesture recognition, we propose a novel way to capture the tracks of users' arms by utilizing the Kinect - a product of Microsoft. With the skeleton tracing ability of Kinect, we can obtain the coordinates sequence of arms. Then we establish the library of gestures and recognize some given gestures by using DTW (Dynamic Time Warping) algorithm. In communications, we choose H.264 video codec standard for video encoding and decoding, and combine with the network communication technology, to ensure the real-time requirement of robot teleoperation.