Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Programming Manufacturing Robots with Imperfect AI: LLMs as Tuning Experts for FDM Print Configuration Selection

Ekta U. Samani, Christopher G. Atkeson

Abstract

We use fused deposition modeling (FDM) 3D printing as a case study of how manufacturing robots can use imperfect AI to acquire process expertise. In FDM, print configuration strongly affects output quality. Yet, novice users typically rely on default configurations, trial-and-error, or recommendations from generic AI models (e.g., ChatGPT). These strategies can produce complete prints, but they do not reliably meet specific objectives. Experts iteratively tune print configurations using evidence from prior prints. We present a modular closed-loop approach that treats an LLM as a source of tuning expertise. We embed this source of expertise within a Bayesian optimization loop. An approximate evaluator scores each print configuration and returns structured diagnostics, which the LLM uses to propose natural-language adjustments that are compiled into machine-actionable guidance for optimization. On 100 Thingi10k parts, our LLM-guided loop achieves the best configuration on 78% objects with 0% likely-to-fail cases, while single-shot AI model recommendations are rarely best and exhibit 15% likely-to-fail cases. These results suggest that LLMs provide more value as constrained decision modules in evidence-driven optimization loops than as end-to-end oracles for print configuration selection. We expect this result to extend to broader LLM-based robot programming.

Programming Manufacturing Robots with Imperfect AI: LLMs as Tuning Experts for FDM Print Configuration Selection

Abstract

We use fused deposition modeling (FDM) 3D printing as a case study of how manufacturing robots can use imperfect AI to acquire process expertise. In FDM, print configuration strongly affects output quality. Yet, novice users typically rely on default configurations, trial-and-error, or recommendations from generic AI models (e.g., ChatGPT). These strategies can produce complete prints, but they do not reliably meet specific objectives. Experts iteratively tune print configurations using evidence from prior prints. We present a modular closed-loop approach that treats an LLM as a source of tuning expertise. We embed this source of expertise within a Bayesian optimization loop. An approximate evaluator scores each print configuration and returns structured diagnostics, which the LLM uses to propose natural-language adjustments that are compiled into machine-actionable guidance for optimization. On 100 Thingi10k parts, our LLM-guided loop achieves the best configuration on 78% objects with 0% likely-to-fail cases, while single-shot AI model recommendations are rarely best and exhibit 15% likely-to-fail cases. These results suggest that LLMs provide more value as constrained decision modules in evidence-driven optimization loops than as end-to-end oracles for print configuration selection. We expect this result to extend to broader LLM-based robot programming.
Paper Structure (23 sections, 2 equations, 3 figures, 6 tables)

This paper contains 23 sections, 2 equations, 3 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. FDM Parameter Tuning and Orientation Selection
  4. Printability, Quality Metrics, and Failure Prediction
  5. LLMs in 3D Printing and Optimization
  6. Method
  7. Evaluator
  8. LLM Guidance Generator
  9. LLM Guidance Compiler
  10. Guided Optimization
  11. Experiments
  12. Dataset
  13. Implementation Details
  14. Baselines
  15. Optimization variants
  16. ...and 8 more sections

Figures (3)

  • Figure 1: Overview of the modular LLM-guided optimization loop for FDM 3D print configuration selection. An evaluator computes objective value $Obj(x)$ and structured diagnostics, an LLM proposes corrective changes, and a compiler maps these changes to guidance ($\mathcal{V}(x), \mathcal{I}$) used for optimization. Lighter boxes denote example intermediate outputs.
  • Figure 2: Sample efficiency of LLM-guided optimization (Ours) vs. (Unguided) optimization, showing the best-so-far objective $Obj(x)$ over iterations, averaged across 10 random initializations and objects with shaded 95% bootstrap confidence intervals over objects.
  • Figure 3: Qualitative comparison of toolpath previews and photos of 3D prints for three representative parts across methods. Our method uses GPT-5.2 (Medium Reasoning) guidance. Numbered callouts mark key print issues and configuration changes. Preview colors follow the legend.