Contact-Based Gesture Recognition on an Omni-directional Mobile Robot for a Robot Companion

Human Robot Interaction (HRI) is being highlighted since the applications of robots has been extended from industry to a common human environment. The main objective of HRI is delivering an operator’s intention to a robot and the state of the robot to the operator. To deliver human intention to robot, various kinds of methods can be used such as dedicated input devices, gesture, voice, and touch. Typical industrial robots have been commanded by experienced workers using dedicated input devices such as teach pendants, however, such complicated devices are not suitable for general public to control service robots, and easier methods are needed. If a robot is used as a companion helping people achieve their daily goals rather than an object to be controlled, new HRI methods rather than complex multi-functional teach pendants are necessary because the people have to focus on their own tasks and not to be distracted by controlling robots. Therefore, the HRI should be not only intuitive and simple, but also not interfere or be interfered by operators’ task. One of the advanced intuitive methods for HRI is delivering the human intention with a body gesture. The most famous method for a gesture recognition is using a depth image made from an infrared projection such as KinectTM, and many scientists have developed precise gesture recognition techniques using it. However, the workspace is limited to an indoor environment because the infrared is vulnerable to the sunlight. The alternative to the limited infrared projection is a laser sensor. The main application of a laser sensor was measuring a distance, but it has been extended to object detection and 3D mapping. In addition, some laser sensors which are produced for 3D mapping can scan around it, so robots can recognize the gestures of all the people around it. To use a gesture recognition as a HRI for a robot companion, we need to differentiate a commanding gesture from a working motion, and it is a very difficult work in which a human behavior should be closely analyzed. On the other hand, touching became a pervasive input method for handheld devices which have touch screen components. In the case of touch screen devices, almost all the human intentions contained in touch gestures are welldefined, but touch or contact sensing on a robot body is not investigated enough compared to touch devices. One approach to use touch gesture in HRI is utilizing tactile sensors. In this case, tactile sensors are attached to the skin