Failure Prediction with Statistical Guarantees for Vision-Based Robot Control

TL;DR

This paper introduces a method for predicting failures in vision-based robotic systems with guaranteed error bounds, enhancing safety by effectively anticipating failures using high-dimensional sensor data.

Contribution

It presents a novel approach combining PAC-Bayes theory with class-conditional bounds to synthesize failure predictors with guaranteed error bounds for high-dimensional sensor data.

Findings

Provides strong bounds on failure prediction error rates

Demonstrates improved safety through failure prediction in experiments

Achieves close alignment between theoretical bounds and empirical errors

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.