Automated Negotiation for Future Supply Chains
4.1 PATTERN 1— DELIVERY-DATE ADJUSTMENT: A PRODUCTION DEPLOYMENT IN ELECTRONICS
A primary automated negotiation use case is the renegotiation of delivery date and quantity between a manufacturer and a parts supplier when demand or capacity shifts. The architecture, described in detail in Ando et al. [ 3 ], couples three estimators— a Needs Forecaster that predicts demand from historical and exogenous data, a True-Ordered- Quantity Calculator that aggregates open purchase orders by delivery day, and an Inventory Estimator that projects future stock levels— with two action strategies: a Trading Strategy that issues purchase orders and a Negotiation Strategy that conducts the actual back-and-forth. The negotiation strategy itself is learned by reinforcement learning, with PPO trained against a Gymnasium-compatible simulator [ 3, 6 ]. Figure 5-1 shows the architecture in its full buyer-and-supplier form.
Figure Five Deployment Patterns-2: System architecture— symmetric buyer and supplier environments connected through a standardized negotiation channel. Adapted from Ando et al. [ 3 ].
EDM Association – Journal of Innovation 105