Feedback

Release 2018b Highlights

Release 2018b features two new toolboxes, along with enhancements for developing, testing, and deploying deep learning and automotive applications.

First, let’s talk about general improvements to MATLAB and Simulink. String arrays, or “String arrays” – like, the things in double quotes, were introduced in 16b. They’re memory-efficient and performant; there are built-in functions to manipulate and compare them, and as of 18b,  they are broadly supported throughout MATLAB and Simulink products.

Simulink continues to add to its smart editing capabilities, making it faster and more intuitive to build your models. Here, you can automatically create ports just by clicking and dragging the outline of a block. And the little underline indicates that you can rename that item without going through additional menus. 

Let’s move on to something completely new: 5G Toolbox, which is fully standards-based and standards-compliant. 5G manages wireless communication between people, between vehicles, and between IoT devices. 

So, what does this mean for the 5G developer? First, you can open up the functions and learn exactly how they’re implemented according to the standard. It’s a great learning and teaching tool, whether you’re looking for reference implementations, or learning how the signal processing works.  

Once you’ve written your algorithms, you can generate standards-compliant waveforms to test them. Furthermore, 5G Toolbox provides all the components you need for end-to-end simulation. This means your algorithms will work with other 5G devices, and you’ll have an accurate understanding of their real-world performance. 

MATLAB continues to make deep learning easier and more accessible. Labeling data for deep learning is a tedious yet important task. You can now automate the process of labeling video data in addition to image data. The Deep Network Designer app enables you to interactively design networks. It integrates with the new Network Analyzer, which is like “spellcheck” for your network, helping you visualize, check, and fix errors before training. 

MATLAB now facilitates the import and export of ONNX models so you can interoperate with other deep learning frameworks. And if you’re looking to deploy models to TensorRT, Intel, or ARM processors, GPU Coder and MATLAB Coder enable you to deploy them along with any pre-processing and post-processing steps. 

Switching gears now, pun fully intended, let’s highlight some automotive tools. You can start virtual engine calibration just by specifying your engine test data, and let MATLAB do the rest of the work. This removes many manual steps, while still allowing you to tweak things as needed. 

You can test automated driving algorithms on realistic, complex, and customizable driving scenarios, which provide a great baseline. And to make sure your algorithms are compliant, the Driving Scenario Designer app enables you to streamline that process with pre-built Euro NCAP tests.

Last, but not least, is the new Sensor Fusion and Tracking Toolbox, which bridges the worlds of sensing and controls. The toolbox enables you to develop sensor fusion and tracking algorithms, not just for automotive applications, but for any vehicle that utilizes multiple sensors and navigation systems for self-awareness and situational awareness. You can get started with reference application examples, or build these systems from scratch using inertial filters and tracking algorithms. Then, you can test them using real-world sensor data, or generate that data synthetically, especially if you need to recreate challenging scenarios. 

There’s plenty more in this release, so be sure to check out the rest of what’s new. Thanks for watching and don’t hesitate to leave us feedback.

 

Product Focus