Download this article in .PDF format
This file type includes high-resolution graphics and schematics when applicable.
Heading into 2016, we’re seeing a significant increase in customer interest and activity in the area of 5G research. Three points of focus are:
• New multicarrier modulations: Exploring these new multicarrier modulations will require algorithm development software that can begin on the desktop computing environment and ultimately connect to test equipment and SDR hardware.
• Massive MIMO: Massive MIMO involves using array analysis and beamforming capabilities pioneered by radar systems and applying them to wireless communications.
• Wideband RF at frequencies above 5 GHz: Wideband RF requires the convergence of digital algorithms with tunable RF hardware to ensure that RF systems are operating in the linear region across a very wide bandwidth.
All three areas present numerous technical challenges and require a new, more sophisticated design approach that closes the gaps existing across different engineering areas—such as digital signal processing, RF, and antenna design—a collection of skills rarely possessed by a single engineer.
Because of this complexity, we expect to find ourselves building bridges between these engineering skill sets and establishing workflows across different phases of the development process—from algorithm development, to simulation, prototyping, and production. These complementary talents require underlying modeling, multi-domain simulation, and prototyping tools that provide a commonly understood development platform. This will enable a clean hand-off at each stage of the design workflow, regardless of the engineer’s specific role.
The expectations of engineering teams to implement designs more quickly and cost-effectively place new demands on companies like MathWorks to deploy well-integrated design tools. Such tools should make possible rapid, efficient prototyping, as well as yield usable code that can then be implemented at both the conceptual and production phases of product design.
Making Smart Devices Smarter
In addition, we’re noticing a consumer-driven trend that’s requiring an adoption of machine learning for embedded devices. Consumers no longer want to just collect or see raw data; they want the insights that are hidden within the data. For example, a health monitor will no longer just collect heart-rate and activity information. It will learn your patterns, make recommendations on your fitness level, and, in the future, take action, such as contacting a health professional in case of emergencies.
While much of this analysis can occur in the cloud, we are seeing more and more some level of analytics being deployed to an embedded device, such as a microprocessor or microcontroller. The process by which these embedded algorithms are being developed is what will continue to change well into 2016. Machine learning will continue to be adopted as a standard way to create “classifiers,” which extract the insights from the sensor data using algorithmic methods such as support vector machines, decision trees, and logistic regression.
In 2016, we will see more microprocessor and microcontroller vendors provide their own classification algorithms running on their hardware, as well as a growing need to integrate algorithm development tools and embedded-device IDEs (integrated development environments).