On the Real-time Vehicle Placement Problem
This work addresses the challenge of minimizing wait times for riders in ride-sharing platforms, which is an incremental improvement in optimizing real-time vehicle placement under dynamic conditions.
The paper tackles the problem of placing vehicles to reduce rider wait times in ride-sharing platforms by introducing a real-time vehicle placement problem that accounts for dynamic request patterns influenced by human mobility. It proposes distributed online learning algorithms and bounds their expected performance using a dataset of ten million ride requests from four major U.S. cities, showing significant self-similarity in the requests.
Motivated by ride-sharing platforms' efforts to reduce their riders' wait times for a vehicle, this paper introduces a novel problem of placing vehicles to fulfill real-time pickup requests in a spatially and temporally changing environment. The real-time nature of this problem makes it fundamentally different from other placement and scheduling problems, as it requires not only real-time placement decisions but also handling real-time request dynamics, which are influenced by human mobility patterns. We use a dataset of ten million ride requests from four major U.S. cities to show that the requests exhibit significant self-similarity. We then propose distributed online learning algorithms for the real-time vehicle placement problem and bound their expected performance under this observed self-similarity.