A Counterfactual Reasoning Framework for Fault Diagnosis in Robot Perception Systems
Haeyoon Han, Mahdi Taheri, Soon-Jo Chung, Fred Hadaegh
AI summary
Problem
Perception faults in autonomous robots are difficult to detect and isolate because they propagate through complex multi-stage pipelines and depend heavily on environmental context, while existing diagnostic methods rely on costly physical sensor redundancy.
Approach
The framework uses structural causal models and counterfactual reasoning to define perception reliability tests and an information-theoretic metric called Effective Information. It then formulates active fault diagnosis as a causal bandit problem, solved via Monte Carlo Tree Search to find control inputs that maximize fault distinguishability.
Key results
- Introduced a counterfactual reasoning framework for perception fault diagnosis
- Defined Effective Information as a metric to quantify control input informativeness for fault detection
- Formulated active fault isolation as a causal bandit problem solved with MCTS-UCB
- Demonstrated accurate fault isolation for sensor damage and perceptual degradation in a space robot scenario
Why it matters
Provides a scalable, sensor-redundancy-free method for maintaining safety and mission success in autonomous robots, UAVs, and space vehicles.
Abstract
Perception systems provide a rich understanding of the environment for autonomous systems, shaping decisions in all downstream modules. Hence, accurate detection and isolation of faults in perception systems is important. Faults in perception systems pose particular challenges: faults are often tied to the perceptual context of the environment, and errors in their multi-stage pipelines can propagate across modules. To address this, we adopt a counterfactual reasoning approach to propose a framework for fault detection and isolation (FDI) in perception systems. As opposed to relying on physical redundancy (i.e., having extra sensors), our approach utilizes analytical redundancy with counterfactual reasoning to construct perception reliability tests as causal outcomes influenced by system states and fault scenarios. Counterfactual reasoning generates reliability test results under hypothesized faults to update the belief over fault hypotheses. We derive both passive and active FDI methods. While the passive FDI can be achieved by belief updates, the active FDI approach is defined as a causal bandit problem, where we utilize Monte Carlo Tree Search (MCTS) with upper confidence bound (UCB) to find control inputs that maximize a detection and isolation metric, designated as Effective Information (EI). The mentioned metric quantifies the informativeness of control inputs for FDI. We demonstrate the approach in a robot exploration scenario, where a space robot performing vision-based navigation actively adjusts its attitude to increase EI and correctly isolate faults caused by sensor damage, dynamic scenes, and perceptual degradation.