E-Bike Energy Needs Estimation based on Route Characteristics and Rider Behavior

Environmental awareness triggered the popularity of electric bicycles (E-Bikes) which has seen a significant rise in recent years due to technological advancement. Nevertheless, some limitations affect this continuous growth notably Range Anxiety i.e., the fear of running out of energy mid-ride. While numerous studies have investigated energy prediction in electric vehicles, research on electric bikes remains relatively scarce. Hence, this paper proposes an energy prediction machine learning model that considers route characteristics and rider behavior based on data collected from different sensors mounted on the E-Bike. Different machine learning models were used: Multiple Linear Regression, Random Forest, Light Gradient Boosting Machine, and Extreme Gradient Boosting Machine, and then evaluated using multiple metrics such as Mean Squared Error, Mean Absolute Error, Latency, CPU Utilization, and R2 Score. According to the evaluation metrics, the random forest model outperformed in terms of its R2 score, Mean Squared Error, and Mean Absolute Error, achieving the highest accuracy and lowest errors. However, it exhibited the poorest performance in terms of latency and CPU utilization.