Midterm colloquium Sam Gao

An Outlier Robust path control pipeline for Autonomous Driving

Supervisor: dr. Riccardo Ferrari

Autonomous driving systems have the potential to enhance road safety, efficiency, and sustainability. However, due to the costs and regulatory constraints associated with full-scale vehicles, model-scale autonomous ground vehicle platforms provide a more accessible alternative for research purposes. In real-world systems, non-Gaussian observational noise is prevalent and can significantly affect control performance. This work focuses on two objectives: First, developing a software framework for real-time localization data streaming and RC rover control based on the student Autonomous Vehicle platform at DCSC; Second, reviewing vehicle modeling, path-tracking, and Outlier-Robust Kalman Filters to reject outliers and integrating these components into a path-tracking control pipeline. Although the proposed framework is tailored to the platform, the methodology can be adapted to other systems experiencing noisy measurements.