Dynamic control of ticket inspectors in public transportation systems
研究通过动态控制查票员分布来增加其可见性，从而减少逃票行为，结合两阶段成本函数逼近和监督机器学习，在挪威卑尔根公交网络数据上验证了效果。
We aim to deter fare evasion on urban public transportation systems by effectively increasing the visibility of ticket inspectors. To this end, we propose a dynamic control approach that combines a two-stage cost function approximation with supervised machine learning. The idea is to balance network-wide coverage and prioritization of high-demand areas by dynamically controlling inspector distribution based on predictions of the spatial heterogeneity in passenger volumes. We do this by introducing a state-dependent constraint on the minimal average distance between inspectors. Using real-world data from Bergen’s bus network in Norway, we conduct computational experiments to evaluate the performance of our approach against various benchmarks. The results show that we can considerably increase the visibility of ticket inspectors by managing their distribution in the public transportation network more effectively. At the same time, the overall coverage of the network is increased as well.