Full Article on EETimes
SAN JOSE, Calif. — It’s still early days in commercial markets for the internet of things and especially machine learning, according to a startup making Bluetooth gateways. Rigado wants to spread Bluetooth mesh networking beyond use in lighting and see chip vendors improve the way that they manage their open-source offerings.
The startup aims to deliver products for asset tracking and smart buildings, markets in which “the frameworks and solutions to make [IoT] easy aren’t there yet … you don’t have ubiquitous gateways and IT people who understand IoT,” said Rigado’s co-founder and CTO, Justin Rigling.
Much of Rigado’s work is focused on delivering software to make Bluetooth-based IoT deployments friendly for mainstream users. It implemented additions and APIs to the BlueZ Linux software for Bluetooth that enables use of Python, Node.js, and other popular programming environments.
“Languages like Python are the only way [that some users] know how to build systems,” said Rigling. “Developers using Embedded C are a tiny fraction of those who do Java … we make our system look like an app in a container.”
Rigado is also integrating into its gateways support for cloud-based IoT services such as Amazon’s Greengrass, Microsoft’s Azure Edge, and ClearBlade. “Their cloud-based consoles let users build Javascript apps and deploy agents on our gateway like an AWS Lambda service,” he said.
The software work requires low-level software that is robust and up to date. Here, silicon vendors sometimes fall short.
“Some silicon vendors fork their open-source stacks,” he said. “Rather than merge them into the general repository, they keep stack additions as upstream forks, but as Linux gets better, those forked repositories are left behind on the latest updates.”
Rigling claims that he has designed dozens of IoT modules for networks including LoRa, Thread, and Zigbee. Rigado’s gateways support Thread and Zigbee, but the company focused on Bluetooth from its start. “They all will have a place over the next decade, but we really made a bet on Bluetooth because its ubiquitous in mobile phones, and that makes it easier to deploy,” he said.
The Bluetooth Special Interest Group that Rigling attends is doing a lot of work optimizing the mesh protocol for asset tracking, one of the hottest applications in IoT today. To date, few users have deployed the mesh spec, and most of those who have are focused on lighting apps.
“We’re steering our customers toward mesh, but it’s still new,” said Rigling. “So far, just one supports it … we use a Nordic nRF52840 SiC in our gateway that supports mesh along with the BlueZ software. The Zephyr RTOS has good Bluetooth mesh support, Silicon Labs has a Bluetooth mesh stack, and NXP has implemented at least part of the mesh spec, but in end products and projects, its use is still limited.”
Separately, Rigado is looking to upgrade its host processor, a single-core 800-MHz Cortex-A7. To support as many as four apps running simultaneously, “I want two to four cores running around 1.8 GHz for flexibility, and there are a lot of vendors,” said Rigling.
Security will be a big factor in choosing a chip. “Processors need to support a secure boot and secure enclave, we need full-drive encryption, and we’ve looked at adding a secure element to store private keys — that’s a big selling point,” he said.
Meanwhile, end users are just starting to think about machine learning in the IoT. “A lot of our customers are in the phase of collecting data into the cloud, and later, they will build inference models and push them down to the edge … with support for Greengrass, we have some of that capability already.”