Abstract
Control scheme design and robotic applications have attracted increasing attention from research communities in the past decades due to their ability to enhance human life in multiple aspects. Actuation systems and control schemes are two of the most important issues in robot development. The development and control of robots are challenging tasks that require knowledge of mechanical design, finite element analysis, computer vision, and controller design. This thesis will present three kinds of robots, aiming at providing state-of-the-art control schemes for robotic systems.The first one is the sanding robot, which is an industry robot. This research aims to develop an autonomous robot for the sanding task to replace human workers in hazardous industry environments. Even though industrial robots have been well developed, fully autonomous robots in some labor-intensive environments have not yet been developed. In this thesis, we developed an autonomous sanding robot with the following key features. The main contributions in developing this robot are summarized as follows:
• Autonomous robotic sanding of all surfaces on a wooden decoration box is studied. The sanding robot consists of two rotational joints and two translational joints to sand all 13 surfaces of the sanded object. 3D modeling is realized by capturing the point cloud data through a structured-light camera and reconstructing the sanded object. The adaptive impedance controller is proposed to drive the robot sanding the object automatically under the desired impedance model without knowing the exact parameters or structure information of the dynamics and disturbances.
• Autonomous sanding experiments are conducted on rectangular boxes and rounded corner boxes. The quality is assessed by autonomously counting the overexposure points shown on the captured point cloud image. The complete removal of all overexposure points on the surface proves the removal of the coat on the box. If not, then the robot will autonomously continue sanding the same surface; otherwise, it will sand the next surface.
The second one is the lower-limb exoskeleton robot, which is a medical robot. The lower-limb exoskeleton robot is widely used to enhance the strength of human beings or assist hemiplegic patients in rehab. The robot needs to interact with human beings physically, so the interaction between them should be soft to ensure safety. Compliant actuators are implemented for passive soft interaction. The human intention is included in the control scheme design to achieve human-in-the-loop control. The performance assessment of the exoskeleton robot is quantified on the basis of the comfort level of the human being who wears the exoskeleton. The main contributions are summarized as follows:
• A lower-limb exoskeleton robot with two hip and two knee joints is invented. Serial elastic actuators are implemented on the four joints to achieve soft interaction at the hardware level. The control system uses the master–slave structure, and a battery is installed on the robot to avoid an external power supply. The adaptive intention-driven variable impedance control scheme is designed to improve the comfort level in the walking experiment. The short-term and long-term human intent estimation is involved in the control scheme design.
• The sensing system to detect the human being’s status includes four surface electromyogram (sEMG) sensors, four inertial motion units detecting the real-time joint motion, and two force gauges detecting the interaction forces. With the information detected from the human being, the dual-channel robotic control scheme is proposed with neural networks to estimate the short-term human motion intention and the probabilistic model to predict the long-term human motion intention. The EMG signals are used to check the comfort level of the human being who wears the exoskeleton. A series of comparative studies, ablation studies, and the overall assessment were conducted to demonstrate the effectiveness of the designed control scheme with the developed lower-limb exoskeleton robot.
The last one is the magnetic actuation system (MAS) for actuating miniature robots. The control of the miniature robots driven by the MAS could achieve non/minimally invasive therapy. A gradient-enhanced clinical-application-oriented electromagnetic actuation system is developed. Its distinctive features include a high magnetic field gradient in a three-dimensional space, a large and extendable workspace, and a low self-weight. The developed system can generate multiple magnetic fields, such as uniform, rotating, and scattering magnetic fields. The potential applications are in knee cartilage regeneration and brain surgery.
Impedance control is a typical controller that solves soft interactions. This thesis designs the control scheme based on adaptive impedance control and implements it in sanding and exoskeleton robots. The design and implementation of the proposed control scheme for industrial and medical systems show the broad values of this research topic, which I hope could inspire the field of robotics.
| Date of Award | 29 Aug 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Gang Gary FENG (Supervisor) & Dong SUN (Supervisor) |