Estudio y simulación de un vehículo autopilotado en Unity 5 haciendo uso de algoritmos de aprendizaje automático
Loading...
Download
Official URL
Full text at PDC
Publication date
2018
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
En la actualidad, los vehículos autónomos han dejado de ser algo del futuro. Poco a poco se han logrado avances en este campo , hasta el punto de llegar a tener en funcionamiento taxis sin conductor.
En este proyecto hemos optado por implementar un sistema con el que experimentar diversas técnicas de aprendizaje automático en una simulación de un vehículo virtual autopilotado en Unity, basándonos en el modelo de NVIDIA “End to End”.
Esto nos permite aumentar y poner en práctica los conocimientos obtenidos en asignaturas como aprendizaje automático.
Para llevar a cabo esto, el proyecto se divide en varias partes diferenciadas:
• Investigación y estudio (Nvidia model)
• Implementación del simulador (Unity 3D)
• Implementación scripts (Red Neuronal)
• Pruebas y conclusiones (Training & Testing)
Currently, autonomous vehicles have ceased to be something of the future. Little by little, progress was made in this field, to the point of having taxis without driver in operation. In this project we have chosen to implement a system with which to experiment with various machine learning techniques in a simulation of a virtual vehicle autopiloted in Unity, based on the NVIDIA "End to End" model. This allows us to increase and put into practice the knowledge obtained in subjects such as machine learning. To carry out this, the project is divided into several differentiated parts: • Research and study (Nvidia model) • Simulator implementation (Unity 3D) • Scripts implementation (Neural network) • Testing and validation (Training & Testing)
Currently, autonomous vehicles have ceased to be something of the future. Little by little, progress was made in this field, to the point of having taxis without driver in operation. In this project we have chosen to implement a system with which to experiment with various machine learning techniques in a simulation of a virtual vehicle autopiloted in Unity, based on the NVIDIA "End to End" model. This allows us to increase and put into practice the knowledge obtained in subjects such as machine learning. To carry out this, the project is divided into several differentiated parts: • Research and study (Nvidia model) • Simulator implementation (Unity 3D) • Scripts implementation (Neural network) • Testing and validation (Training & Testing)
Description
Universidad Complutense, Facultad de Informática. Departamento de Arquitectura de Computadores y Automática, curso 2017/2018