R2Deep: Recharging Recommendation System for Electric Taxis based on Deep Learning

Photo of an eTaxi.

With support from their governments, many countries, such as Unites States and China, have already partially adopted electric taxi (eTaxi) into their public transportation system. 

Electric vehicles are becoming increasingly popular due to the advantages of low-emission, zero pollution and high-efficiency. Unlike traditional Internal Combustion Engine taxis which are able to refuel in minutes, eTaxis suffer from an inherent long recharging cycle, limited number of charging stations, improper charging station deployment strategies and consequently, loss of revenue for their drivers and companies. To improve their operational efficiency as well as increase their revenue, an intelligent eTaxi recharging recommendation system is urgently needed for the widely deployment of eTaxis.

Currently, studies on recharging strategies of public transportation systems usually utilize the theory of load balance, scheduling optimization and taxi trajectory mining.  Different from the precious research work, this project intends to develop a recharging recommendation system including best recharging time and location for eTaxi drivers based on a deep learning model, called R2Deep.  

The project aims to:

  • analyze the existing eTaxi GPS trajectory data and convert them into information on the grid maps which will then be directly fed into the deep learning models;
  • utilize deep learning techniques including both Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN) to learn latent patterns behind eTaxi data sets and provide real-time suggestions on recharging time and charging stations to eTaxis drivers and
  • evaluate the R2Deep model and analyze its performance with real world data.