ISE Seminar Series - Justin Goodson

TITLE: Ditch the Distribution: ETAs for Truck Scheduling

 ABSTRACT: Logistics synchronization problems often require coordinating vehicles with resources at their destinations. Uncertain travel times complicate this coordination. Existing models typically assume known travel time distributions and no new information about travel times until vehicles reach their destinations. In practice, navigation systems provide operators with frequent estimated times of arrival (ETAs), creating opportunities to improve synchronization through real-time information. Motivated by our collaboration with the logistics division of the Italian post, we study inbound truck scheduling at warehouse unloading docks with frequent ETA updates. We formulate the problem as a partially observable Markov decision process (POMDP) that represents arrival uncertainty through belief distributions updated with each ETA. We develop a lookahead policy for dynamic resource allocation under evolving ETA information. To evaluate the policy, we derive an information relaxation dual bound, which yields a challenging penalized perfect information problem. We develop a decomposition algorithm that separates truck-to-dock assignment and sequencing from service start time decisions. Experiments show that the dual bound substantially improves upon conventional perfect information bounds, the lookahead policy achieves near-optimal performance when arrival time distributions are known, and incorporating ETAs reduces expected waiting time by 29 percent relative to current industry practice.

 BIO: Justin Goodson is a professor at Saint Louis University. He researches methods to solve sequential decision problems. He teaches analytics, supply chain management, and reinforcement learning. He studied business analytics and industrial engineering at the University of Iowa and the University of Missouri. He enjoys photography, writing, and musical composition.