Control strategies for exoskeleton gait training after stroke: understanding the importance of parameter tuning
(English) The control strategies implemented in exoskeletons play a crucial role in determining the interaction between the device and user. However, it is unclear what is the most suitable control strategy and settings that maximizes neural recovery. The main objective of this thesis is to find and...
| Author: | |
|---|---|
| Format: | doctoral thesis |
| Publication Date: | 2023 |
| Country: | España |
| Institution: | Universitat Politècnica de Catalunya (UPC) |
| Repository: | UPCommons. Portal del coneixement obert de la UPC |
| Language: | English |
| OAI Identifier: | oai:upcommons.upc.edu:2117/404665 |
| Online Access: | https://hdl.handle.net/2117/404665 https://dx.doi.org/10.5821/dissertation-2117-404665 |
| Access Level: | Open access |
| Keyword: | Àrees temàtiques de la UPC::Enginyeria biomèdica |
| Summary: | (English) The control strategies implemented in exoskeletons play a crucial role in determining the interaction between the device and user. However, it is unclear what is the most suitable control strategy and settings that maximizes neural recovery. The main objective of this thesis is to find and contribute to solving the open challenges of the exoskeleton controllers for post-stroke gait training/rehabilitation. For this purpose, the barriers and opportunities of exoskeleton controllers are analysed first through a systematic review. We have identified that most of the exoskeleton-based training programs are limited to assistance and do not explore alternative types of training, like functional resistance training. There is a consensus on the fact that exoskeletons should promote the active participation of the patients by providing adapted assistance, and that one of the factors that hinders this requirement is parameter tuning. The tuning of the exoskeleton control parameters has shown to have a big impact on the training outcomes. However, there is a lack of systematic or automatic procedures to help clinicians in the selection of appropriate exoskeleton control parameter values. One of the reasons for this shortage might be the unknown relationship between the control parameters and the post-stroke gait biomechanics. In the other section of this review, we saw that little is known on the clinical efficacy of the exoskeleton control strategies due to the low homogeneity of the experimental protocols used. Engineers and therapists do not fully understand how to use exoskeletons effectively for post-stroke gait rehabilitation, since there are large number of unexplored combinations of control parameters, and there is still not solid evidence on which control strategies are more effective. To address these open challenges, we developed an ankle (ABLE-S) and a knee exoskeleton (ABLE-KS) that can provide time-adapted assistance through the whole gait cycle. The next Part of this thesis focuses on supplementing current evidence on the exoskeleton control strategies and validating alternative training methods. We carried out two clinical pilot studies with five (with ABLE-S) and six (with ABLE-KS) post-stroke participants. The comprehensive experimental analysis and protocols reveals that the exoskeletons were capable of correcting the main post-stroke knee and ankle impairments in comparison to walking without the exoskeleton. With the ABLE-KS, we also examined the use of a novel robotic training that combined assistance and resistance modes together with auditory feedback to improve peak knee flexion angle. Our preliminary findings suggest that the proposed training approach can produce similar or larger improvements in post-stroke individuals than other studies with knee exoskeletons that used higher training intensities and were only based on providing assistance training. The following Part examined the interaction between different levels of assistance of the ABLE-S and ABLE-KS and the gait biomechanics to guide the future design of adaptive control strategies. Results showed that variations of the plantarflexion peak torque magnitude and timing and the peak knee flexion torque affected a wider range of the analyzed gait metrics in comparison with the dorsiflexion peak torque magnitude, which did not show strong relationship with any of the outcome metrics. Moreover, we validated the feasibility of using shank-worn Inertial Measurement Units (IMUs) for clinical gait analysis after stroke, and evaluate their preliminary applicability in designing an automatic and adaptive controller for the ABLE-KS knee exoskeleton. Finally, we performed an off-line validation of two machine learning models, i.e., linear regression and neural network, that accounted for variations of the maximum vertical displacement estimated by the shank-worn IMU to adapt the peak knee flexion torque of the ABLE-KS exoskeleton. |
|---|