Analysis of vehicle parking assistants

Bc. Šimon Odstrčil

Diplomová práce

Abstract:

This thesis investigates the effectiveness and reliability of modern Autonomous Parking Assistants by benchmarking their performance against average human drivers. The primary objective is to evaluate whether current automated solutions have reached a level of proficiency that meets or exceeds human intuition in terms of operational efficiency and spatial precision. The research examines two distinct technological approaches currently leading the industry: vision-only systems that rely on neural networks to interpret optical data, and multi-modal sensor fusion systems that integrate various technologies such as radar and ultrasonic sensors to achieve redundant environmental perception. The results of the comparative study reveal a distinct trade-off between mechanical consistency and contextual awareness. While the autonomous systems demonstrated a significant advantage in final orientation they were notably less efficient in terms of maneuver duration and the total number of gear shifts required. Human operators proved more effective at "tucking" the vehicle into restricted spaces and managing longitudinal depth, highlighting a persistent "logic gap" in automated safety buffers. Furthermore, the research identifies critical reliability concerns for vision-dependent systems in varied environmental conditions. Ultimately, the thesis concludes that while automation provides geometric accuracy, human drivers remain the more efficient and adaptable agents in complex urban parking scenarios.

Keywords

Camera, Lidar, Sensor, Parking Assistant

Jazyk práce: angličtina

Datum vytvoření / odevzdání či podání práce: 3. 4. 2026

Obhajoba závěrečné práce

• Obhajoba proběhla 19. 5. 2026
• Vedoucí: Ing. David Marčev, Ph.D.
• Oponent: Tomáš Kráčalík, DiS.

Česká zemědělská univerzita v Praze

Technická fakulta

Magisterský studijní program:
Silniční a městská automobilová doprava