DIGITAL ELECTRONICS DESIGN FEATURE
A DIGITAL SPANNER IN THE MATHEMATICAL WORKS MATLAB and Simulink suites receive an update
B
uilding on its established work with MATLAB and Simulink, MathWorks has launched its suite update, Release 2019b. For those that are less familiar with the technology, MATLAB is a programming environment for algorithm development, data analysis, visualisation and numeric computation, while Simulink instigates a block diagram environment for simulation, enabling the model-based design of multidomain and embedded engineering systems. But what is new in R2019b, what are some examples of what it enables and how does it work?
WHAT’S NEW?
In R2019b, one will find the Deep Learning Toolbox, which builds on the flexible training loops and networks introduced earlier this year. New capabilities permit users to train advanced network architectures using custom training loops, automatic differentiation, shared weights and custom loss functions. In addition, users can now build networks ranging from generative adversarial networks (GANs), Siamese networks, variational autoencoders and attention networks.
WHAT DOES IT ENABLE AND HOW? One example is the capability it offers to classify radar and communications modulation types, by using time-frequency techniques and a deep learning network. This categorisation, which has numerous applications, such as in cognitive radar and software- defined radio, is an important function for an intelligent receiver. Traditionally, to identify these waveforms and classify them, one would define meaningful features and input these into a classifier.
Additionally, R2019b introduces additional capabilities in support of the automotive industry, across multiple products, for automated driving, sensor fusion and tracking, among others. The former supports 3D simulation, augmenting Simulink’s features: the ability to develop and test driving algorithms in a 3D-generated environment, and a block that enables users to generate the velocity profile of a driving patch under kinematic constraints. The latter enables track-to-track fusion, and the architecture of decentralised tracing systems.
While effective, this procedure can require extensive effort and domain knowledge to yield an accurate classification; the Deep Learning Toolbox allows engineers to use a relevant network to extract time-frequency features from signals and automatically perform signal classification. This would be achieved firstly by computing the FFT for a few of the LFM waveforms, to show the variances in the generated set, and then inputting extracted features in place of the original signal data. Users would then store the smoothed-pseudo Wigner-Ville distribution of the signals, define network architecture, train the network with spectrogram images and use that to classify the testing data. R2019b can be also used by systems engineers to test the ability of autonomous vehicles to stay in lane, with monocular camera perception. In other words, engineers can enable system-level simulation of controls and vision-based perception programmes by integrating a control algorithm with a lane boundary and vehicle detection algorithm.
Mathworks
www.uk.mathworks.com
Hall-based E-Bike Speed Sensor
High resolution especially for hill starts Improved manipulation resistance Best suitable for ABS-Systems
/ ELECTRONICS
ELECTRONICS | NOVEMBER 2019
15
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48
