Design of Architecture for Carpooling Based on Flexible Pickup and Delivery Locations

Abstract

The carpooling services allow drivers to share their rides with other passengers, reduce the passengers’ fares and time and traffic jam as well as increase the drivers’ income. In recent years, several carpooling recommendation systems have been proposed. However, most studies focus on constructing mathematical models with some improvements and then running the model simulation with the demo data. Currently, there are many apps-based taxi companies that provide carpooling service such as Uber, Grab, etc.; however, due to business secrets, these companies avoid disseminating widely the carpooling system architecture and technologies. To fill the research gaps between mathematical modeling structure to the carpooling system architecture, this paper proposes a carpooling system architecture to present an overall solution to apply the mathematical models of the carpooling services in practice. In the carpooling system architecture, we also apply big data to collect and process space–time big data and apply AI to solve prediction models such as service churn pre-diction, customer data clustering, route planning… Furthermore, this architecture design adds some components such as service quality assessment, improving efficiency with Key Performance Indicators (KPI), reporting, and statistics, etc. to narrow the gaps between research and business models in practice in the era of Industry 4.0.