A scheme of fuzzy training and learning applied to Elebike control system

This paper proposes an fuzzy intelligent control approach to Elebike (electrical power-aided bicycle). The so-called fuzzy intelligence is focused on two phases: the swivel handle training and phase, and the self-learning of the fuzzy logic control (FLC). For the phase 1, voltage level of swivel handle changes in response to road conditions and acceleration non-satisfaction. In phase 2, an initial configuration of FLC processor is given that includes: 1) definition and membership function of each fuzzy input/output variable, 2) 30 fuzzy rules (15 voltage rules and 15 current rules), 3) dynamic weighting of linear combination of voltage and current, 4) coupling of training switch and learning switch. The pseudo swivel handle model built in phase 1 is to be transformed into an appropriate fuzzy rule table in FLC in terms of tuning up the partitioning of membership functions and the levelling of fuzzy input/output variables. Performance index is also provided to monitor the achievement level of learning. In this way, after sufficient learning, FLC intelligence will gradually and eventually replace the swivel handle entirely.