Counterfactual Reasoning for Perception Fault Diagnosis in Autonomous Systems

Overview

Perception systems are critical for autonomous spacecraft and space robots performing rendezvous, proximity operations, and planetary exploration.

Key Idea

In this work, we integrated structural causal models (SCMs) and counterfactual reasoning into closed-loop autonomy to diagnose perception faults.

Key Contributions

• Used counterfactual queries to test perceptual reliability
• Quantified Effective Information (EI) as a measure of causal informativeness
• Formulated active fault diagnosis as a causal bandit problem
• Solved the resulting problem using Monte Carlo Tree Search

Case Study

Autonomous spacecraft proximity operations for asteroid exploration using perception-based navigation.

This represents the first integration of counterfactual reasoning with perception fault isolation in an autonomous closed-loop system.