Reinforcement Learning for Logistics and Transport Problems

Abstract

Modern logistics and transport systems require an extensive amount of optimisation for operational purposes. Reinforcement learning (RL) is a relatively novel method in the optimisation community but has good future potential due to its speediness. My research applies RL to a number of real-world problems by leveraging its advantages. I have conducted a thorough literature review to revise the current progress in this area. In this survey, some novel methods and research gaps were identified. In addition, I solved a real electric bus fleet operations problem with RL algorithms, which yield satisfactory results much faster than commercial solvers. In my next step, with data provided from the industry, I will further deepen my research to solve a more complicated urban delivery and pickup problem.