Cycling induced by functional electrical stimulation (FES) of the lower limb muscles is a common rehabilitative treatment. In this paper, a repetitive learning controller (RLC) is developed for cadence tracking during stationary FES-cycling. The RLC is developed for an uncertain, nonlinear cycle-rider system with autonomous state-dependent switching. The stimulation pattern switches across different muscle groups based on the joint effectiveness to produce torque during different regions of the crank cycle. An electric motor provides assistance in the regions of the crank cycle where the activation of the muscle groups yields low torque production. The developed RLC provides asymptotic cadence tracking despite the presence of unknown, time-varying, bounded disturbances. A Lyapunov-like stability analysis is implemented to generate the learning feedforward term and exploits a recently developed LaSalle-Yoshizawa corollary for nonsmooth systems.