Phase-Aware Guidance Injection for Recurrent MAPPO in Assembly-Line Disruption Recovery
For industrial assembly line scheduling, this work provides a flexible method to inject heterogeneous recovery knowledge at decision time without retraining, though gains are incremental over existing adaptive policies.
The paper tackles disruption recovery in industrial assembly lines, proposing a phase-aware guidance injection framework that augments a trained recurrent MAPPO policy with logit-level action bias during evaluation. Results show that rule-based guidance yields strongest gains, replay-based guidance degrades smoothly, and online LLM guidance provides intermediate improvements.
Disruption recovery in industrial assembly lines requires timely decisions under machine faults, worker absence, and emergency orders. Existing methods either rely on rigid handcrafted recovery logic or learn adaptive policies that do not readily exploit heterogeneous external recovery knowledge at decision time to reduce abnormal recovery time (ART) and preserve on-time delivery (OTD). To address this gap, we propose a phase-aware guidance injection framework that augments a trained recurrent MAPPO (RMAPPO) scheduling policy through logit-level action bias during evaluation. The framework provides a unified decision-time interface for rule-based, replay-based, and online LLM-based guidance, while activating intervention only during abnormal and recovery phases. Experiments on a custom AssemblyLineEnv show that high-quality rule guidance yields the strongest gains, replay-based guidance degrades smoothly under imperfect availability, and online LLM guidance still provides useful intermediate improvements. These results show that decision-time guidance injection can exploit heterogeneous recovery hints without redesigning the actor.