Inventory Management + DRL

Inventory management is a challenging problem that involves making a sequence of optimal decisions under uncertainty. Traditional methods often rely on heuristics or simplifying assumptions that may not capture the complexity and dynamics of real-world scenarios. Motivated by developing an automatic decision-making policy without over-reliance on developing special heuristics tailored to specific conditions, we investigate a framework in which Deep reinforcement learning (DRL) is applied to automatically search, feedback, and update, obtain the optimal policy for inventory management problems. DRL is a powerful technique that can learn from trial and error to handle high-dimensional, nonlinear, and stochastic problems. First of all, to enhance the decision-making effectiveness to obtain long-term returns under stochastic demand environments, we explore how an agent can learn efficiently from historical information to infer the replenishment policy. Second, we will discuss how an effective deep learning model is designed to optimize the inventory policy that can discover the latent pattern in a stochastic demand environment and respond in advance. Finally, we are extending our proposed model to more complex inventory management problems of censored demand, random supply, etc.