Abstract:

The road traffic flow data obtained from the Baidu Map Application Programming Interface (API) was used to predict the future driving energy-consumption of pure electric vehicles, and a vehicle on-road experimental verification was conducted on the in-loop platform of the cloud computing system. Used the road traffic data obtained by Baidu to calculate the remaining mileage, the path planning, the energy-management strategies, and the charging pile layout, etc. These data were combined with vehicle-driving data and used as a training data set. The future energy-consumption was predicted with the k-means cluster analysis algorithm and the support vector machine (SVM) classifcation algorithm. The predicted value of the remaining battery state of charge (SOC) was compared with the actual value obtained from the vehicle on-road experiments. The results show that the error of future driving energy-consumption prediction are limited to inset of one-standard-deviation σ for a 40 min driving condition (about 20 km), based on Baidu Map API traffic flow data. Therefore, the accuracy of the proposed prediction algorithm in this paper is verified.

Key words: autonomous vehicles, vehicle networking, driving conditions, Baidu maps API (application program interface), cluster analysis, support vector machine, cloud computing