Harmonic RL: A Frequency-Domain Approach to Reinforcement Learning with Application to Active Knee Prosthesis | IEEE Conference Publication | IEEE Xplore