**Problem
**

Range anxiety is a significant reason why people are hesitant to buy e-bikes – they’re afraid of running out of battery in the middle of a ride, with no place to recharge. It doesn’t help that e-bikes often don’t have a gauge that shows how much distance is left in a bike battery. Instead, most e-bikes have a display that reveals battery voltage—though we can all agree that it’s difficult to determine the remaining range on an e-bike from this number.

Although e-bike manufacturers do publish the maximum range of their models, we found that this number isn’t accurate for all riders, since some people ride more aggressively than others.

**Solution
**Using data from a fleet of 31 sensor-equipped e-bikes used in the University of Waterloo WeBike project, combined with OpenStreetMap data, we evaluated two range prediction methods for e-bikes. The first model is a simple one, since it just takes into account the average battery usage from past trips. The second model is a more complex linear regression model that considers the characteristics of the anticipated route (such as off-road percentage, the number of stop signs, and the number of traffic lights), as well as battery temperature.

**Evaluation
**

We found that the more complex linear regression model didn’t perform much better than the simpler one. Using real trip data, our predictions using the simple model were usually within a 10% of the actual remaining range at the end of the trip.

These results should be of interest to e-bike manufacturers because the simple model we tested, which gave promising results, can be implemented as a simple on-board prediction technique. Since most e-bikes have an odometer built in, making this odometer data accessible and adding the ability to measure battery voltage and current are the only main additions needed to implement our technique inexpensively.

L. Gebhard, L. Golab. S. Keshav, and H. de Meer, “Range prediction for electric bicycles,” Proc. *ACM e-Energy 2016.*