Autonomous actions on the Intelligent Edge

Powerful, low-cost edge controllers using open source software are changing the industry landscape. Intelligent Edge Computing is bringing more power to machine learning and providing the foundation for a distributed architecture that will make machine collaboration more prevalent than ever before.

The joke in our industry is that we are either moving towards or away from centralization. The latest trend sees us moving from our present cloud power to the edge driven by the lower cost of extremely powerful controllers capable of a decentralized collaboration. Although this embodies a lot of new concepts it is not that different than the microprocessor revolution of the early 1980s when the power of mainframe and mini was pushed to stand alone controllers.

From AutomatedBuildings.com’s lead August article, Your Buildings’ 21st Century Edge:

Looking at what is happening in edge control and analytics in cars is a good way to understand how comparable technology will bring buildings into line with 21st Century service demands.
We see, for example, that new edge controllers – those that support the full software stack needed to collect, store and run analytics on time-series data from a building’s many digitized sources – are going to be a powerful force in building automation. 
  


We will follow the autonomous car's development of extremely powerful edge implementations to achieve autonomous interaction for our industry as well. Simple pattern recognition and response (“Look, a dog!”) are handled locally by the autonomous car. In our industry learned actions such as "there is no one in the room," "client arrived at 8:30 am yesterday," ''client has just returned from jog" etc. and other learned actions would likewise be turned into autonomous actions.


From last month’s well-read article, From the Clouds to the Fog comes this quote:


Back then, I advocated for enterprise energy monitoring and building control in the low-lying cumulus clouds. Today, I would look to managing energy and service delivery in local autonomous systems. These autonomous systems are the fog. Fog offers thinly disbursed nodes of intelligent actors.

From this article on evolving open source, The Edge of VOLTTRON’s Sword comes this quote:

The software stack is available free on GitHub to put onto a small form factor computer and create your very own “edge device”. For the building and HVAC space, the platform speaks BACnet and MODBUS. Out of the box and on a mini pc it can be used as a data collection device.

The article closes with:

True open source communities have been relatively rare in the building automation industry; Haystack and Sedona are the two others that immediately come to mind. It is our hope that VOLTTRON can become a robust and well-supported platform with an active community of individuals and organizations dedicated to further development. Secure and open access to data is fundamental in realizing the future potential of the building automation industry.

Microsoft CEO Satya Nadella recently outlined the state of things during his keynote address at the company’s Build 2017 conference in Seattle:

We’re moving from what is today’s mobile-first, cloud-first world to a new world that is going to be made up of an intelligent cloud and intelligent edge. 



With IoT Edge, Microsoft now makes it easier for developers to move some of their computing needs to these devices. IoT Edge can run on Windows and Linux and on devices as small as a Raspberry Pi with only 128MB of memory. The Microsoft services that can run on these devices include Azure Machine Learning, Stream Analytics (which came to Edge devices earlier this year), Azure Functions, Microsoft’s AI services and the Azure IoT Hub.

The end of last week Intel launched Deep Learning on a USB stick. This ultra-low power VPU enables users to add visual intelligence and machine learning capabilities in battery-powered products such as autonomous drones, or intelligent security cameras—at the edge, without a connection to the network, or the need for a cloud backend for your machine learning application.

This leap from cloud to tiny edge devices literally opens our world -- but how to manage machine distributed collaboration?

A possible solution from Harbour Research is waiting in the wings. Blockchain has the potential to provide the foundation for a decentralized architecture that would allow permissioning of data to proper stakeholders in the ecosystem they’re associated with, enabling devices to work together autonomously, without humans.

In Toby Considine’s August column Cryptocurrency is more than just Blockchain and Bitcoin  comes these thoughts:

I am watching closely a new CC project, IOTA, which is based on Tangle. Tangle uses directed asymmetric graphs instead of Blockchain. Tangle supports forking a database, to rejoin later, which might be critical in IoT applications that may lose connections to the cloud. IOTA can operate without the cloud if an isolated market is desired. IOTA has been demonstrated running on devices as small as a Raspberry PI.

IOTA features an element called The Tangle Ledger that is able to settle transactions with zero fees so devices can trade exact amounts of resources on-demand, as well as store data from sensors and dataloggers securely, with everything verified on the ledger.

In case your mind (like mine) is now tangled, read more at https://iota.org/

The swarmed data collected along with learned actions will be the feedstock for training machine-learning. It will create a new world where powerful edge devices will learn what they need to know by simply being present, respond with learned actions, and then share their knowledge through distributed collaboration. It will start with small victories but you will be amazed how fast this will happen. It will be like the switch from horses to cars, but the change to autonomous things will occur even faster.

Autonomous things will provide seamless interaction learning our wishes. In the beginning, we will express ourselves with data input, but quickly more autonomous devices pushed to the new ever evolving edge will turn our feelings into data for machine distributed collaboration.

Machine learning and artificial intelligence are not common in today’s building automation systems. Some systems have features built into their software that represent the beginnings of this technology, but none have been able to capture the true benefits that Autonomous Actions on the Edge will. In the very near future, increased visibility and context will turn what was once darkness and noise into easily digestible and manageable chunks of actionable information, and it will all happen autonomously on the Edge.

We are presently working on cool thoughts for our education sessions for AHRExpo January 2018 Chicago, The Future of Building Automation - Growing Autonomous from the Intelligent Edge

Be sure to join us and share your thoughts with us on this exciting topic!