Inventory Control with Lateral Transshipment Using Proximal Policy Optimization

This paper presents the proximal policy optimization algorithm to find the optimal policy for the inventory control problem with lateral transshipment. The inventory control problem with lateral transshipment in a three-echelon supply chain is formulated as a Markov decision process. The supply chain consists of one supplier, two warehouses, and two retailers, where lateral transshipment is permitted between the warehouses. The demands at the retailers are independent and stochastic. Proximal policy optimization is used to find the optimal ordering policies and the optimal lateral transshipment policies for two warehouses to maximize the profit of the inventory system. To evaluate the performance of the proximal policy optimization algorithm, value iteration is also implemented to solve the inventory control problem. The simulation method is then used to evaluate the performance of both algorithms. The experimental results show that the policy obtained by the proximal policy optimization algorithm performs close to the optimal policy obtained by value iteration without knowing the transition probability of the system.