Pre-Conference Talk by HAN Chung Kyun | Smart Bundling for Crowdsourced Package Deliveries

In this research, we investigate how to effectively construct task bundles for the urban crowdsourced package delivery service. Bundling is a common technique utilized by many crowd- sourced delivery service platforms. The major motivation in task bundling is to reduce worker’s planning burden and to attract workers by combining multiple low-paying tasks into a bundle that pays more. However, it is not immediately clear how to generate good bundles when the platform owner knows nothing about who would eventually participate in the delivery. One such use case is when commuters of public transport are being targeted as crowdsourced workers. In such cases, only aggregate-level flows between stations are known, while workers will reveal their intentions to work only after the bundles are created. We formulate this bundling problem as a mixed integer linear program, and propose a branch-and-price approach in solving it. Initial results suggest that our branch-and-price planning approach greatly outperforms greedy approach which mimics human planner’s heuristics. When compared to the exact approach, we see that the execution is greatly reduced, while still maintaining reasonable solution quality.

This is a pre-conference talk for the 6th INFORMS Transportation Science and Logistics Society Workshop.