Phase-Aware Guidance Injection for Recurrent MAPPO in Assembly-Line Disruption Recovery

2026-06-15

AI Summary

This study addresses the challenge of efficiently integrating heterogeneous recovery knowledge to reduce anomaly resolution time and ensure on-time delivery in industrial assembly lines under disturbances such as equipment failures, absenteeism, or urgent orders. The authors propose a phase-aware guidance injection framework that dynamically activates logit-level action biases during anomaly and recovery phases to enhance a pretrained recurrent MAPPO scheduling policy, without requiring architectural modifications. The framework unifies diverse guidance sources—including hard-coded rules, experience replay, and online large language models—and employs phase detection to ensure timely and context-specific interventions. Experimental results demonstrate that rule-based guidance yields the best performance, experience replay provides graceful degradation under partial information loss, and LLM-based guidance still offers meaningful improvements, collectively achieving significant reductions in recovery time and substantial gains in on-time delivery rates.

Abstract

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.

Problem

• disruption recovery
• assembly-line scheduling
• abnormal recovery time
• on-time delivery
• heterogeneous guidance

Innovation

• phase-aware guidance
• recurrent MAPPO
• decision-time injection
• heterogeneous recovery knowledge
• assembly-line disruption recovery