Name: VINÍCIUS BRITO CARDOSO
Publication date: 22/10/2024
Examining board:
Name |
Role |
|---|---|
| ALBERTO FERREIRA DE SOUZA | Presidente |
| ANSELMO FRIZERA NETO | Examinador Interno |
| CLAUDINE SANTOS BADUE | Examinador Interno |
| KARIN SATIE KOMATI | Examinador Externo |
| MARIELLA BERGER ANDRADE | Examinador Externo |
Summary: Over the past two decades, autonomous vehicles have made remarkable
progress. However, achieving full autonomy in vehicles remains a significant hurdle,
as autonomous vehicles must accurately perceive their surroundings, make informed
decisions, and execute real-time safe and efficient driving maneuvers. These tasks are
compounded by the unpredictability of real-world scenarios, where factors such as
traffic, road conditions, and human behavior introduce substantial variability. This
thesis addresses these challenges by focusing on the decision-making system of self-driving vehicles, which is critical for guiding the vehicle from its initial position to a
predefined destination while ensuring safety and adherence to traffic regulations. The
decision-making system encompasses several subsystems, including route planning,
path planning, behavior selection, motion planning, obstacle avoidance, and control.
The contributions of this work include the development and evaluation of an enhanced
motion planning method, along with the presentation of the decision-making
architecture for an integrated self-driving vehicle software system. The proposed
method was tested in real-world scenarios on two different platforms: a passenger car
and a truck. The experimental results demonstrate the system’s high accuracy in
following the reference path, with an average distance of 0.09 meters and a standard
deviation () of 0.1 meters for the passenger car, while for the truck, the average
distance was 0.08 meters, with a standard deviation of 0.07 meters. The system also
demonstrated excellent accuracy in reverse maneuvers, with an average distance of
0.010 meters (=0.016) for the passenger car during parking maneuvers and 0.009
meters(=0.010) for the truck in maneuvers simulating loading/unloading. In the
experiment in an off-road mining truck operation it achieves an average of 0.14 m ( =
0.11). These results highlight the system’s ability to maintain high precision across
different platforms and driving scenarios, demonstrating its potential for industrial
applications.
