VUB investigated new methodologies for system identification and tuning of remote actuation in wearable robotics, focusing on flexible shafts for ankle–foot orthoses. These components can reduce distal mass and improve efficiency compared to Bowden cables or pneumatic systems, but their behaviour under dynamic use has been poorly understood. To address this, the work concentrated on system-level characterisation of flexible shafts, analysing how performance depends on mechanical constraints, bending, and routing.
Three experimental campaigns were conducted:
- Out-of-plane motion tests showed that restricting helical buckling increases stiffness, reduces hysteresis, and improves efficiency, while excessive freedom of movement introduces nonlinearities. This highlighted the importance of controlling shaft deformation when designing wearable systems.
- Fully constrained translation tests revealed that stiffness and torque transmission efficiency remain high even at large bend angles if translation is tightly constrained. This finding contradicted common assumptions in the literature that bending itself reduces stiffness, demonstrating instead that translation and buckling are the dominant factors. These insights suggest that careful mechanical routing can enable lighter, thinner shafts to achieve the required torque capacity, directly contributing to reduced device weight and improved wearability.
- Motion capture tests demonstrated that bending and endpoint translation have similar effects when measured as deviations from an ideal reference plane, reinforcing the importance of routing strategies to minimise deviations rather than focusing solely on bend radius. The experiments also revealed a consistent relationship between tensile forces and transmitted torque, which may provide a practical proxy for system monitoring in future implementations.
Together, these results establish a clear design principle: maximising translation constraint yields more predictable and efficient torque transmission. This challenges supplier data and prior models, shifting design focus from bend radius to routing and support structures.
In addition to these experimental outcomes, the project explored future methodologies for system identification and tuning. Depending on the observed degree of nonlinearity in final device prototypes, different approaches are recommended:
- For near-linear behaviour: frequency-domain methods.
- For moderate nonlinearities: polynomial state-space or nonlinear autoregressive models.
- For strongly nonlinear behaviour: hybrid strategies combining physics-based models with machine learning.
- Human-in-the-loop tuning methods are proposed to refine controller parameters based on user feedback and physiological signals, ensuring personalised and safe performance.
The main achievements of this work are:
- Demonstration that translation constraint, not bending angle, is the dominant factor affecting stiffness, hysteresis, and efficiency.
- Identification of promising pathways for system identification, modelling and tuning, adaptable to varying levels of system nonlinearity.
- Development of design guidelines for shaft routing, enabling lighter, more efficient ankle–foot orthoses.
