A Reinforcement-Learning-Based Assisted Power Management With QoR Provisioning for Human–Electric Hybrid Bicycle

In this paper, a reinforcement-learning-based assisted power management (RLAPM) with quality-of-riding (QoR) provisioning is proposed for the human-electric hybrid bicycle or the pedelec, which is a light electric vehicle (LEV) driven mainly by human´s pedal force with the assisted force from a battery-powered electric motor. By learning from the changes in the riding environment, the proposed RLAPM adaptively dispatches appropriate assisted motor power to support the rider in meeting the QoR requirement, i.e., safety and comfort, of the pedelec. With the RLAPM, not only pedelec rider´s QoR is guaranteed, but energy utilization of battery is improved as well. Simulations of the RLAPM and the RLAPM enhancement (RLAPME) for the pedelec are performed under different road types. Experimental results demonstrate that the achievability of the comfort riding of the RLAPME is improved by 24%, while the energy utilization is improved by 50% by comparing with other existing assisted power methods for the pedelec in urban riding.