Abstract
Archimedes drive is a composite friction planetary traction drive invented by IMSystems. Its friction
traction gifts higher efficiency to Archimedes Drive; while, it also brings higher nonlinearity (such as
creep) to the system. The creep is a microslip phenomenon triggering velocity loss. In collaboration
with IMSystems, this EngD design project is called for designing an actuator control strategy without
any torque sensor to implement on the Archimedes-based actuator for IMSystems. Our control goal
is to achieve accurate low frequency (0Hz-20Hz) torque control between [0, 15]Nm for the passive arm
orthoses and low-back exoskeletons (wearable robotics projects 7-8 https://www.wearablerobotics.nl/).
Note that the motor is equipped with a current controller. Therefore, this project only considers the
relationship between motor input and drive output torque.
To comprehend this nonlinear system better, a series of experiments were conducted to identify the
dominated dynamic behaviors. Simulation results show that the plant model, based on system identification,
can reflect real-world scenarios, even if some intricate nonlinear dynamics are not captured by our
model. Leveraging this dynamics model, two controller strategies have been developed for tracking performance
and disturbance rejection. The tracking controller emphasizes reference tracking with minimal
disturbances, while the robust controller is designed for constant reference tracking under time-varying
disturbances.
Despite the theoretical design, the simulations and experimental results revealed that the creep torque
estimator is unable to function as expected due to its excessive sensitivity. Therefore, a low-gain feedback
controller is necessary to limit the potential error during the dynamic reference tracking simulation, which
constrains the system’s performance bandwidth. On the other hand, thanks to the reliable estimation
of constant torque, the slide-mode control simulation exhibited sufficient robustness against time-varying
disturbances. However, during experiments, the creep torque estimator failed to predict even a constant
torque, thereby rejecting the possibility of testing the feedback controller. Consequently, the final implementation
of the tracking and robust controller in the setup integrates solely of the feedforward control
and compensator, achieving a bandwidth of 3 Hz. Meanwhile, the robustness experiments maintain only
marginal stability. For future work, to ensure the implementation of a closed-loop feedback controller,
equipping a high-resolution analog encoder and a single-axis gyro is recommended.
traction gifts higher efficiency to Archimedes Drive; while, it also brings higher nonlinearity (such as
creep) to the system. The creep is a microslip phenomenon triggering velocity loss. In collaboration
with IMSystems, this EngD design project is called for designing an actuator control strategy without
any torque sensor to implement on the Archimedes-based actuator for IMSystems. Our control goal
is to achieve accurate low frequency (0Hz-20Hz) torque control between [0, 15]Nm for the passive arm
orthoses and low-back exoskeletons (wearable robotics projects 7-8 https://www.wearablerobotics.nl/).
Note that the motor is equipped with a current controller. Therefore, this project only considers the
relationship between motor input and drive output torque.
To comprehend this nonlinear system better, a series of experiments were conducted to identify the
dominated dynamic behaviors. Simulation results show that the plant model, based on system identification,
can reflect real-world scenarios, even if some intricate nonlinear dynamics are not captured by our
model. Leveraging this dynamics model, two controller strategies have been developed for tracking performance
and disturbance rejection. The tracking controller emphasizes reference tracking with minimal
disturbances, while the robust controller is designed for constant reference tracking under time-varying
disturbances.
Despite the theoretical design, the simulations and experimental results revealed that the creep torque
estimator is unable to function as expected due to its excessive sensitivity. Therefore, a low-gain feedback
controller is necessary to limit the potential error during the dynamic reference tracking simulation, which
constrains the system’s performance bandwidth. On the other hand, thanks to the reliable estimation
of constant torque, the slide-mode control simulation exhibited sufficient robustness against time-varying
disturbances. However, during experiments, the creep torque estimator failed to predict even a constant
torque, thereby rejecting the possibility of testing the feedback controller. Consequently, the final implementation
of the tracking and robust controller in the setup integrates solely of the feedforward control
and compensator, achieving a bandwidth of 3 Hz. Meanwhile, the robustness experiments maintain only
marginal stability. For future work, to ensure the implementation of a closed-loop feedback controller,
equipping a high-resolution analog encoder and a single-axis gyro is recommended.
| Original language | English |
|---|---|
| Awarding Institution |
|
| Supervisors/Advisors |
|
| Award date | 22 Nov 2023 |
| Place of Publication | Enschede |
| Publisher | |
| Publication status | Published - 22 Nov 2023 |