TacEvo: Self-Evolving Architecture Discovery for Robotic Tactile Perception via LLM-Driven Quality-Diversity Search
2026-06-29 • Robotics
Robotics
AI summaryⓘ
The authors present TacEvo, a system that uses a large language model to automatically improve neural network designs for vision-based tactile sensors, which turn touch information into images. TacEvo evolves network architectures by making code changes and selects the best ones based on performance and diversity, without much human effort. Their method works well on tasks like force prediction and texture classification, matching or beating expert-designed models. This shows that using AI to evolve sensor designs can help robots better understand touch.
Vision-based tactile sensingNeural architecture searchLarge language modelMAP-ElitesForce regressionGrating classificationArchitectural diversityEfficiency ratioRobotic sensingSelf-evolving architectures
Authors
Mohammed AbuSadeh, Lan Wei, Dandan Zhang
Abstract
Vision-based tactile sensing converts contact-induced surface deformation into images, enabling robots to infer contact forces and fine surface textures that are not accessible through conventional vision alone. However, tactile images are sensor- and physics-specific, so effective architectures often require expert intuition and extensive manual iteration. Existing neural architecture search (NAS) pipelines can reduce this burden, but they are often computationally expensive and restricted to hand-designed search spaces, which limits architectural novelty and diversity. We introduce TacEvo, a self-evolving architecture discovery framework that improves network designs from downstream feedback. TacEvo uses an LLM to generate code-level mutations and crossovers, and a MAP-Elites quality-diversity loop that preserves diverse elite architectures while preferentially reusing prompts that consistently yield improvements. Exploration is guided by two behavioural descriptors, Architectural Diversity and Efficiency Ratio, which encourage coverage across structural variations and compute-size trade-offs. On ViTacTip force regression and grating classification, TacEvo achieves high autonomous generation reliability (96.0%/94.5% trainable) and improves best validation fitness over 20 generations by 56.1%/96.1%. In a 20-seed post-search high-fidelity evaluation, TacEvo matches the expert baseline on force prediction and outperforms it on fine-grained grating classification. These results suggest that LLM-driven self-evolving search constitutes a practical paradigm for AI-assisted scientific discovery in specialised robotic sensing.