Large-Scale Intelligent Taxicab Scheduling: A Distributed and Future-Aware Approach

Abstract

Intelligent taxicab scheduling systems on smartphones continue to gain popularity as they offer prominent conveniences for urban travelling as well as increase potential profits for taxicab drivers, hence inject prosperity and vitality into intelligent transportation and urban business. Existing scheduling approaches usually fall into biases and myopia due to their single target perspective of satisfying the immediate order acceptance, or maximizing the global business success rates.However, the highly complex spatiotemporal dependencies among multiple factors and the efficiency bottleneck of massive order flows make the taxicab scheduling issue still challenging. To this end, in our paper, we propose an integrated scheduling algorithm with both future-aware and context-aware mechanisms. In particular, we first present a generalized graph-based framework which aims to capture traffic dependencies, providing precise and quantified supply-demand prediction in taxicab scheduling. Then, we develop a measurement of regional supply-demand context to perform cooperative and distributed scheduling. To tackle the efficiency bottleneck in massive order flow scenario, we correspondingly design a model to learn business patterns with considering benefits from both drivers and passengers, and further promote service delivery rates via a novel bi-incentive strategy. Extensive numerical studies illustrate the remarkable significance of our method, in terms of both service delivery rates and global driver revenues. The promising results and brand-new perspectives enable our algorithm to be a paradigm in general spatiotemporal scheduling tasks.