PhD Dissertation Proposal by Joe Waldy | RL Approach to Coordinate Real-World Multi-Agent Dynamic Routing and Scheduling.

Routing and scheduling problems are classical Combinatorial Optimization Problems (COPs) that are well-studied in the Operations Research (OR) community. New and more complex variants of these problems have increasingly been introduced and studied in the literature. These variants are usually driven and inspired by real-world applications.  Real-world routing and scheduling problems are typically characterized by the following two key features: Dynamicity and Complexity. Initially-planned routes and schedules may be disrupted by dynamically-occurring events. In addition, routing and scheduling in real-world context is made complex by the need to coordinate in multi-agent setting due to the presence of other independent entities/agents in the environment. This dissertation discusses and proposes new methodologies that incorporate relevant techniques from the field of AI (Reinforcement Learning (RL) and Multi-Agent System (MAS) more precisely) to supplement and complement classical OR techniques to solve dynamic and multi-agent variants of routing and scheduling problems. The research work in this dissertation is motivated by two real-world problems domains namely urban logistics and law enforcement.

Joe Waldy is a PhD candidate in Computer Science and is advised by Prof. Lau Hoong Chuin. His research focus is on the intersection between AI and OR specifically on how both fields can work together to solve real-world multi-agent dynamic routing and scheduling problems.