Low Cost Neural Network Hardware for Control

2001-01-3397

10/01/2001

Event
Automotive and Transportation Technology Congress and Exposition
Authors Abstract
Content
Feedforward artificial neural networks are universal function approximators and inherently parallel computing structures. Because of the lack of appropriate hardware realisations, applications of neural networks are predominantly implemented as sequential programs on digital processors. In this paper we describe an analogue integrated circuit realisation of a local response neural network (LCNN) that achieves a high degree of parallel computation in a small size, low cost and low power consumption. Because it can directly receive analog inputs from sensors and output analog control signals to actuators it is well suited as a building block for real-time control systems.
Meta TagsDetails
DOI
https://doi.org/10.4271/2001-01-3397
Pages
7
Citation
Sitte, J., "Low Cost Neural Network Hardware for Control," SAE Technical Paper 2001-01-3397, 2001, https://doi.org/10.4271/2001-01-3397.
Additional Details
Publisher
Published
Oct 1, 2001
Product Code
2001-01-3397
Content Type
Technical Paper
Language
English