Driving · 2026-10-06
Generative World Models Enable Predictive Control of Laser Melt Pool Dynamics
Yiyang Yan, Markus Bambach, Mohamadreza Afrasiabi
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
The authors introduce a generative latent world model for localized 3D laser melt-pool dynamics, predicting evolution from histories of temperature and phase morphology under candidate laser actions. The model’s differentiable generative latent dynamics are used as a predictive control plant: MPC optimizes laser schedules via gradients through imagined futures, and an amortized latent policy produces actions in a single forward pass. They re-validate control schedules in high-fidelity CFD.
Why it matters
This work targets melt-pool depth regulation in laser powder bed fusion, where high-fidelity simulation is too slow for control and depth is important for avoiding defects like lack of fusion and overmelting. By learning differentiable, generative dynamics that remain stable for recursive rollouts and support gradients for model predictive control, the authors provide a technical proof-of-concept for predictive control in a manufacturing setting with partial observations and transient, highly dynamic melt-pool behavior.
Method
- Encode temperature and phase-indicator fields (T, α) into a finite-scalar-quantized (FSQ) latent state via an autoencoder; decode physical fields for melt-pool depth/volume evaluation.
- Use a history-conditioned flow-matching generative latent transition (diffusion-transformer) that is differentiable w.r.t. laser actions to support recursive rollouts under transient inputs and gradient-based optimization.
- Implement predictive control with (i) receding-horizon MPC that backpropagates a depth-tracking objective through latent rollouts and (ii) an amortized latent policy trained with short-horizon actor-critic (SHAC), then re-evaluate chosen schedules in CFD.
Limitation
The authors note that repeated evaluation of candidate action sequences against the CFD model is infeasible within a control loop, motivating the world-model as the fast plant.
Abstract (from arXiv)
World models, which learn how environments respond to actions, are emerging as a powerful paradigm for planning through imagined futures, transforming decision-making across games, robotics and autonomous driving. Bringing this capability to manufacturing could enable process decisions on timescales inaccessible to high-fidelity simulation. Here we introduce a generative world model for localized highly dynamic laser melt pool that predicts evolution from histories of temperature and phase morphology under candidate actions. Its generative latent dynamics capture the effects of unresolved melt flow, enabling more accurate recursive rollouts than deterministic regressors under transient laser inputs. Because the learned dynamics are differentiable, the model can serve directly as a predictive control plant. Gradients through imagined futures optimize laser schedules that regulate melt-pool depth over previously unseen geometry, path, initialization. We further distil this optimization into an amortized policy that produces control actions in a single forward pass, providing a proof of concept for real deployment on machines.
Related papers
- Variational Streaming Flow: Probabilistic Forecasting in Physical Time
- WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction
- MiniWAM: Learning Compact Future Targets for Efficient World-Action Modeling
- Copper-Policy: Focus on the Representation for Robust Robot Manipulation
- {\Delta}WAM: Distilling Action Tangent Fields into World Action Models
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-05 · Indexed by Pub-AI: 2026-10-06 UTC
Generative World Models Enable Predictive Control of Laser Melt Pool Dynamics
Preprint posted: 2026-10-05 · Original record ↗
- arxiv
- 2610.06250
Explicit source versions (1)
- v1 · 2026-10-05
- Preprint available
Source date: 2026-10-05 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
World models, which learn how environments respond to actions, are emerging as a powerful paradigm for planning through imagined futures, transforming decision-making across games, robotics and autonomous driving. Bringing this capability to manufacturing could enable process decisions on timescales inaccessible to high-fidelity simulation. Here we introduce a generative world model for localized highly dynamic laser melt pool that predicts evolution from histories of temperature and phase morphology under candidate actions. Its generative latent dynamics capture the effects of unresolved melt flow, enabling more accurate recursive rollouts than deterministic regressors under transient laser inputs. Because the learned dynamics are differentiable, the model can serve directly as a predictive control plant. Gradients through imagined futures optimize laser schedules that regulate melt-pool depth over previously unseen geometry, path, initialization. We further distil this optimization into an amortized policy that produces control actions in a single forward pass, providing a proof of concept for real deployment on machines.