Riding Pattern Identification by Machine Learning for Electric Motorcycles

Identification of riding patterns is one of the key enablers to update energy consumption strategy, optimise the energy management system and increase the range of electric motorcycles despite their weight and space limits. Considering the varying driving conditions in real applications, improving accuracy of the riding pattern recognition without significant complexity is the main challenge. In this paper a simple and efficient online classification method is introduced based on features extracted only from the motorcycle speed. The recognition mechanism is firstly developed using support vector machine technique. The effect of validation method for removing the optimism in classification and the contribution of features to the accuracy of model is then investigated. Evaluation of the method on the real riding conditions in simulation environment shows the effectiveness of the approach.