Achieving transparency in series elastic actuator of sharif lower limb exoskeleton using LLNF-NARX model

Nowadays, exoskeletons have been gaining popularity due to their potential use in rehabilitation and augmentation. These robots often utilize series elastic actuators to facilitate compliant interaction with the human. Numerous studies have been carried out with the purpose of identification and control of these type of actuators. The goal of this paper is to provide a method for dynamic modeling and identification of series elastic actuators. This model is then used in the control loop as a feed-forward term to eliminate the actuator's dynamics. Each series elastic actuator used in the Sharif wearable robot, uses a brushless DC motor, a torsional spring, a harmonic drive, a timing belt, a bevel gearbox, absolute and relative encoders, etc. This actuator dissipates a major part of the torque as friction. To identify this system, a nonlinear dynamic identification method is necessary. Here, the LLNF-NARX (Local Linear Neuro-Fuzzy Nonlinear Auto Regressive with Exogenous input model) data-based method is used in this study for actuator identification and control. In this approach, the LLNF identification method is used as the main core of a NARX structure. Obtained experimental results show that the major part of the torque produced in robot is dissipated in the series elastic actuators. On the other hand, by assuming a relatively high gearbox ratio, inertial effects of the robot's links are negligible compared to that of the actuator, and as a result, it accounts for an inconsiderable amount of the torque produced by the DC motor. Therefore, a significant part of the torque produced from the control rule, is used to overcome the actuator's dynamics. Hence, the torque estimation using the proposed identification method will greatly impact the control of the series elastic actuators and Exoskeleton robot as a whole. The results from this research is used as an inner control loop to cancel the actuator dynamics, making the actuators transparent in other control loops, in a sense.