Performance modules on a modular computing architecture for autonomous driving

Fully autonomous vehicles are expected to be the new era in the upcoming future resulting in a huge increase of data from sensors. To manage this, a modular centralized computing architecture is proposed, consisting of a powerful intelligent central computing platform able to connect to several vehi...

ver descrição completa

Detalhes bibliográficos
Autor: Altemir Vidal, Irina
Formato: tesis de maestría
Fecha de publicación:2021
País:España
Recursos:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/358505
Acesso em linha:https://hdl.handle.net/2117/358505
Access Level:acceso abierto
Palavra-chave:Automobiles--Electronic equipment
Performance Module
Modular Computing Platform
Centralized In-Vehicle Networking Architecture
Autonomous Driving
Vision Application
Automotive
Automòbils--Equip electrònic
Àrees temàtiques de la UPC::Enginyeria electrònica
Descrição
Resumo:Fully autonomous vehicles are expected to be the new era in the upcoming future resulting in a huge increase of data from sensors. To manage this, a modular centralized computing architecture is proposed, consisting of a powerful intelligent central computing platform able to connect to several vehicle interfaces. Then, different powerful high-performance modules are designed to easily plugin to the main system body, known as performance modules. In this project a PCBA is designed as a performance module integrating a Jetson Xavier NX SoM to form a pluggable cartridge into another company project known as MCP, including a power, mechanical and thermal design. A software demo is developed using Nvidia Deepstream SDK to build a video processing pipeline as a vision application. The final design results as a favourable proof-of-concept of a performance module for autonomous driving, with fulfilment of the project objectives. At the end, future work steps are defined.