Multi-Agent Deep Reinforcement Learning for Multi-Echelon Inventory Management
Bibliographic record
Abstract
We apply heterogeneous-agent proximal policy optimization (HAPPO), a multi-agent deep reinforcement learning (MADRL) algorithm, to the decentralized multi-echelon inventory management problems in both a serial supply chain and a supply chain network. We also examine whether the upfront-only information-sharing mechanism used in MADRL helps alleviate the bullwhip effect. Our results show that policies constructed by HAPPO achieve lower overall costs than policies constructed by single-agent deep reinforcement learning and other heuristic policies. Also, the application of HAPPO results in a less significant bullwhip effect than policies constructed by single-agent deep reinforcement learning where information is not shared among actors. Somewhat surprisingly, compared to using the overall costs of the system as a minimization target for each actor, HAPPO achieves lower overall costs when the minimization target for each actor is a combination of its own costs and the overall costs of the system. Our results provide a new perspective on the benefit of information sharing inside the supply chain that helps alleviate the bullwhip effect and improve the overall performance of the system. Upfront information sharing and action coordination in model training among actors is essential, with the former even more essential, for improving a supply chain's overall performance when applying MADRL. Neither actors being fully self-interested nor actors being fully system-focused leads to the best practical performance of policies learned and constructed by MADRL. Our results also verify MADRL's potential in solving various multi-echelon inventory management problems with complex supply chain structures and in non-stationary market environments.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".