Perception with Guarantees: Certified Pose Estimation via Reachability Analysis

2026-02-10

AI Summary

This work addresses the lack of formal worst-case safety guarantees in conventional pose estimation methods for safety-critical applications, which often rely on untrusted external services such as GPS. The authors propose a novel monocular 3D pose estimation approach that leverages a single image and prior knowledge of the target object’s geometry. For the first time, this method integrates reachability analysis with formal verification of neural networks to provide rigorous, mathematically certified error bounds on the estimated pose. Crucially, it operates without dependence on external positioning systems and demonstrates high efficiency and accuracy in both synthetic and real-world scenarios, thereby significantly enhancing system reliability and safety through provably bounded estimation errors.

Technology Category

• Intelligent Robots: State Estimation
• Computer Vision: Biometrics, Face, Gesture & Pose
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Security and Privacy: Large-scale security measurements
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

Agents in cyber-physical systems are increasingly entrusted with safety-critical tasks. Ensuring safety of these agents often requires localizing the pose for subsequent actions. Pose estimates can, e.g., be obtained from various combinations of lidar sensors, cameras, and external services such as GPS. Crucially, in safety-critical domains, a rough estimate is insufficient to formally determine safety, i.e., guaranteeing safety even in the worst-case scenario, and external services might additionally not be trustworthy. We address this problem by presenting a certified pose estimation in 3D solely from a camera image and a well-known target geometry. This is realized by formally bounding the pose, which is computed by leveraging recent results from reachability analysis and formal neural network verification. Our experiments demonstrate that our approach efficiently and accurately localizes agents in both synthetic and real-world experiments.

Problem

• pose estimation
• safety guarantees
• cyber-physical systems
• formal verification
• reachability analysis

Innovation

• certified pose estimation
• reachability analysis
• formal verification
• neural network verification
• safety-critical perception