Pith. sign in

REVIEW 4 cited by

Failure Prediction with Statistical Guarantees for Vision-Based Robot Control

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2202.05894 v2 pith:IGOAACEZ submitted 2022-02-11 cs.RO

Failure Prediction with Statistical Guarantees for Vision-Based Robot Control

classification cs.RO
keywords failureboundsapproacherrorpredictionratescontrolerrors
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We are motivated by the problem of performing failure prediction for safety-critical robotic systems with high-dimensional sensor observations (e.g., vision). Given access to a black-box control policy (e.g., in the form of a neural network) and a dataset of training environments, we present an approach for synthesizing a failure predictor with guaranteed bounds on false-positive and false-negative errors. In order to achieve this, we utilize techniques from Probably Approximately Correct (PAC)-Bayes generalization theory. In addition, we present novel class-conditional bounds that allow us to trade-off the relative rates of false-positive vs. false-negative errors. We propose algorithms that train failure predictors (that take as input the history of sensor observations) by minimizing our theoretical error bounds. We demonstrate the resulting approach using extensive simulation and hardware experiments for vision-based navigation with a drone and grasping objects with a robotic manipulator equipped with a wrist-mounted RGB-D camera. These experiments illustrate the ability of our approach to (1) provide strong bounds on failure prediction error rates (that closely match empirical error rates), and (2) improve safety by predicting failures.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Fail2Progress: Learning from Real-World Robot Failures with Stein Variational Inference

    cs.RO 2025-09 conditional novelty 7.0

    Fail2Progress generates failure-targeted simulation data via Stein variational inference and fine-tunes skill effect models, improving long-horizon manipulation success rates and generalizing to unseen object counts a...

  2. FATE-VLA:Failue-aware test generation for vision-language-action models

    cs.RO 2026-06 unverdicted novelty 6.0

    FATE-VLA reframes VLA evaluation as active failure discovery and reports uncovering up to 29.7% more failures across four models while revealing diverse failure modes.

  3. VLAConf: Calibrated Task-Success Confidence for Vision-Language-Action Models

    cs.RO 2026-05 unverdicted novelty 6.0

    VLAConf is a one-class discriminative method that estimates step-wise task-success confidence for VLA models via anomaly scoring on frozen representations plus step-conditioned modeling, shown to be more efficient tha...

  4. RoboMD: Uncovering Robot Vulnerabilities through Semantic Potential Fields

    cs.RO 2024-12 unverdicted novelty 6.0

    A deep RL vulnerability-prediction policy trained in semantic embedding space finds up to 23% more unique robot manipulation failures than vision-language baselines and enables more efficient fine-tuning.