Figure 17.

Measured tracking performance using the proposed learning-based repetitive controller, where the learning gain is $\phi =0.25$, and the phase-lead compensation level is $\gamma =10$. (a) Convergence history for the exoskeleton knee joint angle in time domain, with iterations #00 to #52 overlapped. (b) X–Y plot of the desired knee angle ${\theta }_{d-mea}\left(t\right)$ against the measured output knee angle ${\theta }_{O-mea}\left(t\right)$ for iterations #50 to #52. With the outer learning loop, the frictional hysteresis of the Bowden transmission cable is significantly reduced after learning, as compared with Figure 15b where only the inner PD loop is active.