This dissertation represents the development of a novel exoskeleton (Exo-Muscle) and corresponding evaluation and controller strategies to achieve more ergonomics and intuitive operation. At the first stage of the study, we introduce the principle, design, and mechatronics of Exo-Muscle, a novel assistive device for the knee joint. Different from the existing systems based on rigid exoskeleton structures or soft-tendon-driven approaches, the proposed device leverages a new semi-rigid principle that explores the benefits of both rigid and soft systems. Using a novel semi-rigid chain mechanism around the knee joint eliminates the misalignment between the device and the knee joint center of rotation. At the same time, it forms a well-defined route for the tendon. This results in more deterministic load compensation functionality compared to fully soft systems. The proposed device can provide up to $38Nm$ assistive torque to the knee joint. The control of the Exo-Muscle relies on the knee effort computation, which is based on the forward kinematic computation with ground reaction force measurement. To achieve the multi-task operation objective, a lightweight force solely targeting long-term operation in the scenarios of the production lines is developed. The proposed design includes six sensing elements that provide a stable support polygon between the foot and the ground. To override the latent nonlinear properties of the device under high load. The Gaussian Process Regression (GPR) based non-model calibration method is proposed. In addition, a novel fault detection method is proposed based on the Gaussian distribution from a GPR model trained for fault detection to improve reliability. To bring back the device after the partial fault of the device. A recovery approach which is also based on the GPR, is proposed. The proposed recovery method can significantly improve the accuracy of extracting the center of pressure and the nominal ground reaction force from the raw sensing reading under the fault conditions. From the control perspective, we proposed a control strategy targeting improving the ergonomics and efficiency of long-term operations. To enable the long-term realistic human state evaluation, a novel EMG-free muscle fatigue evaluation method for human-assistive robot interaction scenarios. Unlike conventional methods, which rely on complex modeling of human biological joints or noisy EMG measurements, we use the GPR to reveal the latent relation between human muscle activation (via EMG measurement) and the corresponding joint effort and state during a simple calibration stage. The trained model then online approximates the muscle activation using the human joint state and effort derived from the wearable device kinematic and sensor measurements. The approximated muscle activation is used for assessing the muscle state of the subject with a model-based fatigue model. Then the evaluation model is further implemented in a model predictive-based control schematic, which targets to provide a continuous adaptive control to the user for regulating the muscle fatigue state. In experiments following characteristics of the developed device, evaluation approaches and control are demonstrated: --The developed device was successfully validated through a series of experiments demonstrating the capacity of the device to provide the target assistive functionality to the knee joint; --The GPR-based calibration method for the developed force sole shows a comparative reliable accuracy when compared with the model-based method; --The fault detection for the force sole can effectively detect the fault of the sensing elements, while the corresponding recovery method shows a significant improvement in the accuracy for extracting the COP and NGRF under fault conditions; --The proposed EMG-free muscle fatigue evaluation approach can provide continuous and reliable evaluation of the muscle fatigue state during the operation of the exoskeleton. --The proposed muscle fatigue-based model predictive control can provide adaptive control depending on the user's muscle fatigue state.

The Development and Control of a Novel Bio-Inspired Semi-Rigid Device (Exo-Muscle) for the Knee

ZHANG, YIFANG
2023-02-27

Abstract

This dissertation represents the development of a novel exoskeleton (Exo-Muscle) and corresponding evaluation and controller strategies to achieve more ergonomics and intuitive operation. At the first stage of the study, we introduce the principle, design, and mechatronics of Exo-Muscle, a novel assistive device for the knee joint. Different from the existing systems based on rigid exoskeleton structures or soft-tendon-driven approaches, the proposed device leverages a new semi-rigid principle that explores the benefits of both rigid and soft systems. Using a novel semi-rigid chain mechanism around the knee joint eliminates the misalignment between the device and the knee joint center of rotation. At the same time, it forms a well-defined route for the tendon. This results in more deterministic load compensation functionality compared to fully soft systems. The proposed device can provide up to $38Nm$ assistive torque to the knee joint. The control of the Exo-Muscle relies on the knee effort computation, which is based on the forward kinematic computation with ground reaction force measurement. To achieve the multi-task operation objective, a lightweight force solely targeting long-term operation in the scenarios of the production lines is developed. The proposed design includes six sensing elements that provide a stable support polygon between the foot and the ground. To override the latent nonlinear properties of the device under high load. The Gaussian Process Regression (GPR) based non-model calibration method is proposed. In addition, a novel fault detection method is proposed based on the Gaussian distribution from a GPR model trained for fault detection to improve reliability. To bring back the device after the partial fault of the device. A recovery approach which is also based on the GPR, is proposed. The proposed recovery method can significantly improve the accuracy of extracting the center of pressure and the nominal ground reaction force from the raw sensing reading under the fault conditions. From the control perspective, we proposed a control strategy targeting improving the ergonomics and efficiency of long-term operations. To enable the long-term realistic human state evaluation, a novel EMG-free muscle fatigue evaluation method for human-assistive robot interaction scenarios. Unlike conventional methods, which rely on complex modeling of human biological joints or noisy EMG measurements, we use the GPR to reveal the latent relation between human muscle activation (via EMG measurement) and the corresponding joint effort and state during a simple calibration stage. The trained model then online approximates the muscle activation using the human joint state and effort derived from the wearable device kinematic and sensor measurements. The approximated muscle activation is used for assessing the muscle state of the subject with a model-based fatigue model. Then the evaluation model is further implemented in a model predictive-based control schematic, which targets to provide a continuous adaptive control to the user for regulating the muscle fatigue state. In experiments following characteristics of the developed device, evaluation approaches and control are demonstrated: --The developed device was successfully validated through a series of experiments demonstrating the capacity of the device to provide the target assistive functionality to the knee joint; --The GPR-based calibration method for the developed force sole shows a comparative reliable accuracy when compared with the model-based method; --The fault detection for the force sole can effectively detect the fault of the sensing elements, while the corresponding recovery method shows a significant improvement in the accuracy for extracting the COP and NGRF under fault conditions; --The proposed EMG-free muscle fatigue evaluation approach can provide continuous and reliable evaluation of the muscle fatigue state during the operation of the exoskeleton. --The proposed muscle fatigue-based model predictive control can provide adaptive control depending on the user's muscle fatigue state.
27-feb-2023
File in questo prodotto:
File Dimensione Formato  
phdunige_4620002.pdf

Open Access dal 28/02/2024

Tipologia: Tesi di dottorato
Dimensione 32.08 MB
Formato Adobe PDF
32.08 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11567/1108033
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact