Light Petroleum Products Distribution Based on Reinforcement Learning
Rejected manuscript
Venues: UAI; Annals of Operations Research
Reason for rejection: Rejected by both venues over concerns about insufficient novelty and research contributions.
Abstract
The importance of rapidly finding optimal routes in the field of transportation is undeniable. This study aimed to apply the reinforcement learning (RL) approach to the petroleum station reshipment problem (PSRP). To answer this question, we applied Recurrent Proximal Policy Optimization (RecurrentPPO) to the multi-period PSRP. We used an exact solver as a baseline and experimented on a small synthetic dataset. Our results showed the RL model’s ability to adapt to a list of requirements for such a task on a limited-scale problem. Our work also suggests an environment for learning the RL algorithm to solve the PSRP problem, which allows us to perform further experiments with optimization in transportation optimization problems.