PUMA
Istituto di Scienza e Tecnologie dell'Informazione     
Falcini F., Lami G., Mitidieri A. C. Deep learning in automotive software. In: IEEE Software, vol. 34 (3) pp. 56 - 63. IEEE, 2017.
 
 
Abstract
(English)
DEEP LEARNING IS a branch of machine learning based on arti - cial neural networks (ANNs) that model high-level abstractions in input data by using a graph rep- resentation comprising multiple processing layers. For example, in computer-vision-based advanced driver assistance systems (ADASs), deep-learning algorithms improve the detection and recognition of multiple objects, support object classfication, and enable recognition and prediction of actions. Many automotive companies consider deep learning to be a mature, viable technology, for not only improving ADAS performance but also making progress in functional domains such as engine management and vehicle cybersecurity. However, the introduction of deep-learning technology onboard cars is creating challenges for automotive software engineering, which must harmoniously incorporate this technology into its current state of the art. Given deep learning's maturity and viability, it's fundamental to understand whether the deep-learning state of the art is aligned with aut motive demands, at both the technical and methodological levels. The idea of solving a problem by training a neural network, instead of solving the solution using domain knowledge (feature engineering), is revolutionary for automotive software applications. Here, we examine the basics of deep learning for automotive software development, introduce a development lifecycle for this pro- cess, and show how this lifecycle meshes with current standards for automotive-software development.
URL: http://ieeexplore.ieee.org/document/7927925/
DOI: 10.1109/MS.2017.79
Subject Deep Learning
Automotive Software
Software development process management
I.2 ARTIFICIAL INTELLIGENCE
D.2 SOFTWARE ENGINEERING


Icona documento 1) Download Document PDF
Icona documento 2) Download Document PDF


Icona documento Open access Icona documento Restricted Icona documento Private

 


Per ulteriori informazioni, contattare: Librarian http://puma.isti.cnr.it

Valid HTML 4.0 Transitional