Power assistant modes of e-bikes rely on manual switching, and its shock resistance is much worse than that of general vehicles and motorcycles. This work proposes a vibration reduction system for e-bikes. While cycling, the system switches the power assistant mode automatically and notifies riders according to collected sensing data for road awareness. In the system, we use the accelerometer for data collection and feature extraction on bumpy roads. Then, we combine the geographical information for classification and use the classified results for e-bikes' power assistant modes mapping. Finally, we implement a system prototype and evaluate the performance of the prototype.