
Develop reinforcement learning solutions for inventory management, production scheduling, and agv path planning to reduce holding costs, stock-outs, and idle time in dynamic environments.
Practice a project-based approach to reinforcement learning for supply chain management by building real projects, applying concepts, and not covering introductory theory.
Encourage completing at least half the course before leaving a rating. The instructor uses a clear speaking style to make reinforcement learning for supply chain management accessible to international audiences.
Master sequential decision making by examining states, actions, rewards, and policies within the Markov decision process, including value functions and the discount factor gamma.
Connect the Bellman equation to dynamic programming and reinforcement learning, showing how state value equals the immediate reward plus gamma times the expected next-state value, and explain return.
Differentiate model-based and model-free reinforcement learning, where model-based uses a transition and reward model for planning, while model-free learns from experience.
Master policy approximations for scalable, high-dimensional decision making using value-based, policy gradient, and actor-critic methods. Apply to autonomous systems, drones, finance, and healthcare with stable, data-driven learning.
Explore policy classes in reinforcement learning for supply chain management, examining how different policy approaches guide decision-making under uncertainty.
Explore reinforcement learning with Q-learning to optimize supply chain scheduling on a single production line, minimizing total weighted tardiness by prioritizing orders and comparing against EDD, SPT, and WSPT.
Solve capacitated vehicle routing problem with reinforcement learning, using an encoder-decoder and attention mechanism, guided by a critic to minimize route length and outperform greedy baselines.
Design a custom gymnasium environment for an automated guided vehicle on a grid, with four directional actions and normalized coordinates, trained with PPO and Manhattan-distance shaping.
Apply reinforcement learning to inventory management using a Q-learning agent that learns to minimize holding and shortage costs under Poisson demand, outperforming the order-up-to benchmark.
Supply chains operate in highly dynamic and uncertain environments. Traditional optimization methods often struggle when decisions must be made sequentially under uncertainty. Reinforcement Learning (RL) provides a powerful framework to address these challenges by enabling systems to learn optimal policies through interaction with the environment.
In this course, you will learn how to apply reinforcement learning techniques specifically to supply chain management problems. We begin with the foundations of RL, including Markov Decision Processes (MDPs), reward design, value functions, and policy optimization. You will then explore core algorithms such as Q-Learning, Deep Q-Networks (DQN), and Policy Gradient methods, with a strong focus on practical implementation.
The course connects theory directly to real supply chain applications. You will model and solve problems such as single- and multi-echelon inventory control, demand uncertainty management, safety stock optimization, dynamic pricing, transportation decisions, and capacity planning. Throughout the course, we emphasize how to properly define states, actions, rewards, and constraints in operational environments.
Hands-on coding exercises will guide you in building RL agents step by step, training them, and evaluating their performance against traditional policies. You will also learn how to interpret results, avoid common pitfalls, and scale solutions to more complex systems.
By the end of this course, you will be able to design and implement reinforcement learning solutions tailored to real-world supply chain challenges, bridging the gap between machine learning theory and operational decision-making.