User Experience Enchanced Interface ad Controller Design for Human-Robot Interaction

The robotic technologies have been well developed recently in various fields, such as medical services, industrial manufacture and aerospace. Despite their rapid development, how to deal with the uncertain envi-ronment during human-robot interactions effectively still remains un-resolved. The current artificial intelligence (AI) technology does not support robots to fulfil complex tasks without human’s guidance. Thus, teleoperation, which means remotely controlling a robot by a human op-erator, is indispensable in many scenarios. It is an important and useful tool in research fields. This thesis focuses on the study of designing a user experience (UX) enhanced robot controller, and human-robot in-teraction interfaces that try providing human operators an immersion perception of teleoperation. Several works have been done to achieve the goal.First, to control a telerobot smoothly, a customised variable gain con-trol method is proposed where the stiffness of the telerobot varies with the muscle activation level extracted from signals collected by the surface electromyograph(sEMG) devices. Second, two main works are conducted to improve the user-friendliness of the interaction interfaces. One is that force feedback is incorporated into the framework providing operators with haptic feedback to remotely manipulate target objects. Given the high cost of force sensor, in this part of work, a haptic force estimation algorithm is proposed where force sensor is no longer needed. The other main work is developing a visual servo control system, where a stereo camera is mounted on the head of a dual arm robots offering operators real-time working situations. In order to compensate the internal and ex-ternal uncertainties and accurately track the stereo camera’s view angles along planned trajectories, a deterministic learning techniques is utilised, which enables reusing the learnt knowledge before current dynamics changes and thus features increasing the learning efficiency. Third, in-stead of sending commands to the telerobts by joy-sticks, keyboards or demonstrations, the telerobts are controlled directly by the upper limb motion of the human operator in this thesis. Algorithm that utilised the motion signals from inertial measurement unit (IMU) sensor to captures humans’ upper limb motion is designed. The skeleton of the operator is detected by Kinect V2 and then transformed and mapped into the joint positions of the controlled robot arm. In this way, the upper limb mo-tion signals from the operator is able to act as reference trajectories to the telerobts. A more superior neural networks (NN) based trajectory controller is also designed to track the generated reference trajectory. Fourth, to further enhance the human immersion perception of teleop-eration, the virtual reality (VR) technique is incorporated such that the operator can make interaction and adjustment of robots easier and more accurate from a robot’s perspective.Comparative experiments have been performed to demonstrate the effectiveness of the proposed design scheme. Tests with human subjects were also carried out for evaluating the interface design