Performance Measure of Vehicle Onboard Vision System: An Interval Observer-based Approach

Ifqir, Sara; Oufroukh, Naima Ait; Ichalal, Dalil; Mammar, Said · 2018 · Crossref

DOI: 10.1109/icves.2018.8519491

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the reliability evaluation of vehicle onboard vision systems used in Advanced Driver Assistance Systems (ADAS), specifically for lane keeping and departure warning. These systems are highly sensitive to environmental conditions and often fail when lane markings are obscured by adverse weather or lighting. To mitigate this, the authors propose a performance evaluation framework based on set-membership estimation theory, utilizing interval observers to quantify the accuracy of vision-based road curvature measurements. The methodology employs a switched uncertain vehicle lateral dynamics model (bicycle model) combined with vision system dynamics. The approach utilizes two distinct interval observers. First, a Switched Unknown Input Interval Observer (SUIIO) estimates the guaranteed bounds of the vehicle’s state vector (lateral velocity, yaw rate, and displacements), treating road curvature as an unknown input and accounting for tire cornering stiffness uncertainties. Second, a Switched Unknown Disturbance Interval Observer (SUDIO) reconstructs the upper and lower bounds of the road curvature based on the estimated state bounds. The observer design relies on Linear Matrix Inequalities (LMIs) and Multiple Quadratic Input-to-State Stable Lyapunov Functions to ensure stability and robustness against parameter variations and longitudinal velocity changes. The proposed framework was validated using field data acquired from an instrumented prototype vehicle. Sensors included an inertial measurement unit for yaw rate, an optical encoder for steering angle, an odometer for longitudinal velocity, and a vision system for curvature and displacement measurements. The evaluation logic checks for consistency between the vision system’s measured curvature and the estimated interval bounds. If the measured value falls outside the calculated bounds, the system flags the measurement as unreliable. The experimental results demonstrate that the scheme successfully estimates the upper and lower bounds of the road curvature and effectively confirms the reliability of the vision system measurements under the tested conditions. The significance of this work lies in providing a rigorous, observer-based indicator for the quality of vision sensor data. By establishing guaranteed bounds on road curvature, the method allows ADAS controllers to detect when vision-based inputs are corrupted by noise, occlusion, or poor visibility. This enables more robust vehicle guidance by identifying unreliable sensor data in real-time, thereby enhancing safety in scenarios where traditional vision systems might otherwise fail.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• gaze based attention detection