Master's Thesis

A Monocular Perception System for Autonomous Driving Vehicles for Shell Eco-Marathon

Diogo Boleto Valdrez — 2025

Key information

Authors:

Diogo Boleto Valdrez (Diogo Boleto Valdrez)

Supervisors:

João Paulo Salgado Arriscado Costeira (João Paulo Salgado Arriscado Costeira); Rita Maria Mendes de Almeida Correia da Cunha (Rita Maria Mendes de Almeida Correia da Cunha)

Published in

November 25, 2025

Abstract

Autonomous vehicles increasingly rely on sophisticated perception systems to navigate complex environments. This thesis presents a monocular camera-based perception system for the Shell Eco-Marathon Autonomous Urban Concept Challenge, demonstrating that cost-effective, energy-efficient vision-only solutions can achieve competitive performance. The proposed system integrates YOLO11 instance segmentation for track element detection, FastSAM for complementary blue pin obstacle detection, and GraphSLAM for simultaneous localization and mapping. A custom dataset of 3000+ annotated competition images was developed, achieving 0.80 mAP50 detection accuracy across all classes. Experimental validation on a 735-meter competition course revealed that obstacle-only SLAM achieves 1.784m ATE RMSE, outperforming pure odometry by 0.7% while reducing drift rate by 17.2%. Critically, lane detection constraints degraded relative pose error by 228% despite providing 12.4% drift improvement, revealing fundamental trade-offs in vision-based localization. Experimental validation against a stereo based approach revealed that the proposed system obtains a significantly closer result to ground truth in feature lacking environments (similar to the competition course). The system successfully operated in real-time on NVIDIA Jetson AGX Orin hardware at 30 Hz, demonstrating the practical viability of monocular perception for autonomous vehicles applications, with the sensors consuming only 1.9W compared to 14-30W for LiDAR alternatives.

Publication details

Authors in the community:

Fields of Science and Technology (FOS)

electrical-engineering-electronic-engineering-information-engineering - Electrical engineering, electronic engineering, information engineering

Publication language (ISO code)

eng - English

Rights type:

Embargo lifted

Date available:

October 26, 2026

Institution name

Instituto Superior Técnico