Jump Start Product Development with NXP Application Software Packs

Tackling the complexities of edge-connected, machine learning (ML) applications can be a daunting and long process. Combining the functions of an involved application with the intricacies of deploying that ML model on a cost- and power-effective platform requires significant effort and time. NXP’s ML-based system state monitor application software pack (App SW Pack) provides product-ready source code for rapid development of such a complex application.

Building edge-ready solutions is not an easy process, and today almost all developers avoid trying to build applications or products from scratch. Ever present time-to-market pressure means end-product manufacturers and application engineers increasingly rely on existing examples and abstraction layers wherever available to save time. This allows more focus on user experience and coding at the higher application levels for integration into final products, rather than spending time and effort on re-developing core code such as low-level drivers, middleware and communication stacks. Focus on end-product user experience and differentiation are key drivers for today’s product development efforts, so reference applications are a key enabler for fast and robust development on MCU and applications processor-based products.

NXP’s application enablement spans a wide range, from application notes addressing a specific device feature to production-ready turnkey solutions with custom-built reference hardware for specific use case that can be quickly turned into end-products. Between these approaches to enablement lies a middle ground, where a full application solution is needed as a reference, but with full flexibility to adapt that reference to a customer’s own application needs.

To address this need and complement the wide range of MCU and applications processor products, NXP has added App SW Packs, providing real-world, complete application software packages that are flexible enough to be leveraged into customer products across various technology spaces by a broad audience.

App SW Packs are a hybrid software bundle approach, offering a middle ground between turnkey solutions and example application code. They can be leveraged for productization (similar to turnkey solutions) but are also meant to be accessible by a larger development audience with an open-source policy for customized productization. Provided in a complete and easy to use package, App SW Packs deliver a combination of drivers, middleware, communication stacks and sample applications, along with comprehensive guides and training materials (application notes, lab guides, training videos, etc.) to enable fast developer ramp-up and smooth integration within products.

NXP is developing a range of App SW Packs, targeting machine learning, connectivity, remote device management, voice and vision and other application spaces.

Machine Learning-Based System State Monitor – App SW Pack

The first in the series of App SW Packs is a machine learning-based state monitoring system for time-series data.

Given the complexities of machine learning, neural networks and deep learning, coupled with the challenges to scale down to resource-constrained MCUs, ML application development can be daunting. The ML-based System State Monitor App SW Pack was created to simplify development of smart sensing applications that rely on ML processing of time-series data to generate decisions without human intervention.

This ML-based System State Monitor App SW Pack enables developers to develop and deploy neural networks on MCU-based systems. This App SW Pack shows how to create a fan vibration state monitoring and failure identification solution, and also details how to validate and evaluate the performance of a model on an MCU-based embedded target. It goes through a real use case (starting from t0) and results in an ML model being deployed and evaluated on the embedded board, demonstrating each step of the ML development:

1. How to collect data and to build a custom comprehensive dataset
2. How to define the architecture of a neural network and train a model
3. How to validate and deploy the model on the embedded device

The ML-based System State Monitor App SW Pack was developed using the eIQ^® ML development environment and comes with software for MCUs, various ML projects and helpful collateral (application note, lab guide and video, training and validation dataset). The alignment with eIQ allows the ML-based System State Monitor App SW Pack to take advantage of the essential baseline enablement within the eIQ framework including the eIQ Toolkit, eIQ Portal and MCU-based eIQ inference engines like DeepViewRT^™, Glow and TensorFlow^™ Lite Micro.

Start exploring the Application Software Pack: ML-based System State Monitor.