A Reconfigurable Framework for Vehicle Localization in Urban Areas.

Area of Science:

• Robotics
• Autonomous Systems
• Sensor Fusion

Background:

• Accurate vehicle localization is critical for safe autonomous driving, especially in complex urban environments.
• Existing fault detection methods often halt vehicle operation upon failure detection.
• Temporary sensor failures, like GPS outages, pose a significant challenge to continuous localization.

Purpose of the Study:

• To develop an advanced localization framework for autonomous vehicles incorporating robust fault detection and fallback strategies.
• To enable continued vehicle operation in a degraded mode during temporary sensor failures.
• To implement a reconfiguration module that allows recovery to a non-fault state and mitigates accumulated errors.

Main Methods:

• Development of a novel localization framework with an integrated fault detection and reconfiguration module.
• Implementation of alternative positioning strategies for degraded operational modes.
• Design of a system for resetting alternative algorithms and recovering from temporary failures.
• Validation through extensive experiments in a simulated driving environment.

Main Results:

• The proposed framework ensures accurate localization for driving tasks despite sensor failures.
• The vehicle only stops operation when a critical, unrecoverable failure state is reached.
• Reconfiguration strategies effectively reset accumulated drift from alternative positioning algorithms.
• Significant improvement in overall performance and bounding of mean localization error was demonstrated.

Conclusions:

• The developed localization framework enhances the safety and reliability of autonomous vehicles in urban settings.
• The system's ability to handle temporary sensor failures and recover gracefully improves operational continuity.
• The proposed approach offers a more resilient and effective solution for autonomous vehicle localization compared to traditional methods.