VLM-DEWM: Dynamic External World Model for Verifiable and Resilient Vision-Language Planning in Manufacturing
🤖 AI Summary
This work addresses the limitations of current vision-language models in dynamic manufacturing environments, where stateless operation often leads to world state drift and opaque reasoning processes hinder fault diagnosis and recovery. To overcome these challenges, the authors propose the Dynamic External World Model (DEWM) architecture, which decouples reasoning from state management by externalizing inference trajectories into a structured, queryable format. This design enables traceable state evolution and consistency verification, while a discrepancy-driven fault recovery mechanism enhances system resilience. Experimental results demonstrate substantial improvements in verifiability and robustness: state tracking accuracy increases from 56% to 93%, fault recovery success rate rises from below 5% to 95%, and computational overhead is simultaneously reduced.
Application Category
📝 Abstract
Vision-language model (VLM) shows promise for high-level planning in smart manufacturing, yet their deployment in dynamic workcells faces two critical challenges: (1) stateless operation, they cannot persistently track out-of-view states, causing world-state drift; and (2) opaque reasoning, failures are difficult to diagnose, leading to costly blind retries. This paper presents VLM-DEWM, a cognitive architecture that decouples VLM reasoning from world-state management through a persistent, queryable Dynamic External World Model (DEWM). Each VLM decision is structured into an Externalizable Reasoning Trace (ERT), comprising action proposal, world belief, and causal assumption, which is validated against DEWM before execution. When failures occur, discrepancy analysis between predicted and observed states enables targeted recovery instead of global replanning. We evaluate VLM-DEWM on multi-station assembly, large-scale facility exploration, and real-robot recovery under induced failures. Compared to baseline memory-augmented VLM systems, VLM DEWM improves state-tracking accuracy from 56% to 93%, increases recovery success rate from below 5% to 95%, and significantly reduces computational overhead through structured memory. These results establish VLM-DEWM as a verifiable and resilient solution for long-horizon robotic operations in dynamic manufacturing environments.
Problem
Research questions and friction points this paper is trying to address.
Vision-Language Model
Dynamic Manufacturing
State Tracking
Opaque Reasoning
World-State Drift
Innovation
Methods, ideas, or system contributions that make the work stand out.
Dynamic External World Model
Vision-Language Model
Externalizable Reasoning Trace
State Tracking
Failure Recovery
🔎 Similar Papers
No similar papers found.
