A Neural-embedding-based Recommender System to Get the Most out of EV Recharge Times

Abstract

Long recharge times are one of the main challenges in providing a seamless experience for users of electric vehicles (EVs) during long trips. Recommender systems play a crucial role in optimizing the recharge process and in providing a better overall experience, as witnessed by a number of solutions presented in the literature. Most of these works, however, focused on minimizing waiting times by suggesting where and when to recharge the EV, leveraging for example historical demand patterns and/or real-time information.

In this work, we tackle the problem from the different – and complementary – perspective of suggesting relevant and tailored activities users can perform to make the most out of their EV recharge times. To this end, we present a novel recommender system that leverages open data from the OpenStreetMap project and uses a specifically-trained neural embedding model to capture the semantic relationships between user preferences and nearby available activities and Points of Interest (PoIs).

The recommender system, whose implementation we make freely available for interested researchers and practitioners, provides personalized recommendations to users based on their preferences and past behaviour. We assessed the effectiveness of the proposal by conducting a study involving real users with different backgrounds, and the results showed that the proposal is effective in providing relevant and personalized recommendations to users, during recharge times.