What is it about?
We provide a robust and realistic multi-echelon inventory optimization framework.
Featured Image
Why is it important?
We use System Dynamics reflecting the supply chain environment and Neuroevolution hashtag#ReinforcementLearning (NERL) reflecting an intelligent agent that integrates Neural Network and Evolutionary Algorithms, creating an effective decision-making model under dynamic complexity and hashtag#uncertainty. Compared to classical RL algorithms, NERL can help to reduce sample inefficiency, slower convergence, limited exploration, and lack of generalizability in representing state-action spaces.
Perspectives
Some practical implications: • It can cope with demand and supply uncertainties, as well as discount events given by suppliers that often arise unexpectedly. • It provides optimal Order Quantity & Transportation Mode decisions for each company in the supply chain. This is reasonable since suppliers/3PLs typically offer various delivery options. • It has been tested under serial and divergence supply networks and works effectively i.e. up to 58% total cost reduction compared to the optimized continuous review model with the capability for avoiding overfitting. • The cost reduction is also followed by the fill rate improvement. • It provides a more stable hashtag#inventory level among all supply chain partners while the hashtag#bullwhip effect can be damped.
Zakka Ugih Rizqi
National Taiwan University of Science and Technology
Read the Original
This page is a summary of: Neuroevolution reinforcement learning for multi-echelon inventory optimization with delivery options and uncertain discount, Engineering Applications of Artificial Intelligence, August 2024, Elsevier,
DOI: 10.1016/j.engappai.2024.108670.
You can read the full text:
Contributors
The following have contributed to this page
