Graduation Semester and Year
Winter 2025
Language
Engilsh
Document Type
Thesis
Degree Name
Master of Science in Computer Engineering
Department
Computer Science and Engineering
First Advisor
William J. Beksi
Second Advisor
Chengkai Li
Third Advisor
Farhad A. Kamangar
Abstract
Vision-language-action (VLA) models have emerged as generalist robotic controllers capable of mapping visual observations and natural language instructions to continuous action sequences. However, VLAs provide no calibrated measure of confidence in their action predictions, which limits their reliability in real-world settings where uncertainty and failures must be anticipated. In this thesis, we present a novel VLA framework that produces uncertainty-guided and failure-aware control signals. Concretely, our approach applies conformal prediction directly to the action token outputs of pretrained VLA policies, yielding calibrated uncertainty estimates that correlate with execution quality and task success. Additionally, we extend conformal prediction to the robot state space to detect outliers or unsafe states before failures occur, providing a simple yet effective failure detection mechanism that complements action-level uncertainty. We evaluate our method in both simulation and via physical robot experiments across diverse manipulation tasks. Our results show that conformalized action predictions consistently improve failure anticipation, reduce catastrophic errors, and provide a calibrated measure of confidence without retraining or modifying the underlying VLA. In summary, our methodology offers a practical path toward reliable and uncertainty-aware control for generalist robot policies.
Keywords
AI-based methods, robot safety, failure detection and recovery, probability and statistical methods.
Disciplines
Robotics
License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
Recommended Citation
Chen, Lingling, "An Uncertainty-Guided and Failure-Aware Vision-Language-Action Model Control Framework for Robotics" (2025). Material Science and Engineering Theses. 145.
https://mavmatrix.uta.edu/materialscieng_theses/145