Mobile App Developers, Enterprise IT Vendors and Controls System Integrators Compete and Cooperate to Make Machine Learning Accessible to the Masses and Robust Enough for Commercial Building Operations and Facilities Management
Mobile app developers would have you believe that tiny acts of automation — like dimming a connected light, locking a door or adjusting a thermostat — are a big deal. Maybe they are. Enterprise IT software vendors posit that an automation revolution is upon us; it’s the ‘Internet of Things’ and it will sell a lot of their their cloud infrastructures, data management and security applications. Maybe it will. Meanwhile, the traditional masters of commercial building automation — the system integrators that make big equipment from disparate vendors work together — continue their steady, incremental progress towards control systems that meet ever higher energy performance and comfort design goals. Absolutely, they are. All this market activity and noise constitutes not so much a battle between the different worlds for the ultimate automation platform for building operations and facilities management, as it is a ballet. This early in the program, there is great opportunity for partnering and emulation of technologies and best practices. In just the last few weeks there were some notable announcements about automation and machine learning coming from all corners. Here’s my summary:
First, on February 19th, IFTTT introduced the ‘Do’ Button. The five-letter acronym company name stands for IF-THIS-THEN-THAT — capturing the essence of its service, which is to string together functions from mobile and web apps in conditional series to carry out some action. Users are essentially coding machine rules. One of the fastest growing IFTTT user communities is home automation enthusiasts that have already invested in connected bulbs, locks and thermostats. These early adopters are sure to love the ‘Do Button’ because whatever action they wanted done with IFTTT, they can now accomplish in one tap on their phone or tablet, ie turn on lights, open the garage, change the temperature. IFTTT has a channel list to match this target group – Hue and LIfx bulbs, Lutron controls, Nest thermostats, Scout alarms, etc.
The power of IFTTT goes well beyond these simple ‘Do’ use cases. Its growing popularity is giving more and more mobile device users hands-on training in coding machine rules.
Any dance has its forward and backward steps, though, and the mobile-app-centric smart home received some very bad publicity last month too. A contributor to the popular consumer gadget-press site gizmodo wrote a scathing review about his home hub and a Google employee published a ‘nightmarish’ youtube video documenting the buggy nature of the Nest smoke alarm. Such reviews should be expected of tech that is still in the early adopter phase. These products are highly dependent on the home owner playing the role of system integrator and investing many hours in making technology work. The bad press that comes from over promising and under delivering is going to make main-stream home users even more hesitant to invest in such devices. The chasm that must be crossed for these brands to reach the commercial building world just got bigger.
There would probably not be so many brave start-ups, like IFTTT, offering free or nearly-free mobile apps, if they could not get cloud-hosting services at surprisingly low cost. As it stands, service providers like Amazon Web Services (AWS) and Google have been at battle cutting prices on cloud storage and data managment and luring mobility start-ups into their camps. Both companies also have advertising and marketing businesses at their core. which drives a hunger for media content and contextual data of all kinds. Their public cloud services are like ‘loss leaders’ that keep them at the top of the ‘big data’ game. Microsoft’s business doesn’t have the same dynamic, but it and other traditional enterprise IT software vendors also offer cloud hosting. A distinction is that they often offer private clouds that adhere to the data ownership, security and compliance requirements of the customer. But, even for these private clouds and hybrid public/private clouds, the trend is for customers to demand heterogenous open platforms that don’t lock them into one vendor. This is an aspect of the automation dance that has enterprise IT and building management system software in lockstep. Commissioning Engineer Matt Schwartz of Altura Associates explains why ‘open’ is so important to building control automation in his article this month.
Microsoft announced another way it is distinguishing its cloud services at the Strata + Hadoop World conference in San Jose in February. This is Azure ML, a machine-learning-as-a-cloud service that is now released for general availability. Like IFTTT, Azure ML is aimed at making it easy for the non-data-scientist to jump in and code machine rules. Carnegie Mellon University used Azure ML to design the predictive analytics it uses across campus to save energy in building operations and facilities management. With the help of OSIsoft, best known for its Operational Intelligence software, the PI System, researchers at CMU’s Center for Building Performance and Diagnostics built the plaftform. Data streams from multiple building automation systems are integrated for control through this common platform, and the Azure ML interface is used to write analytics rules to run against this data. This BuildingContext.me blog post explains more about the deployment and links to a detailed presentation by CMU researcher, Bertrand Lasternas.
A recent general availability release native to the control automation world also challenged conventional notions of what comprises a building automation platform. This is J2 Innovations FIN 3.0, on display at January’s AHR Expo in Chicago. The announcement and demonstration had talking points common to the mobility and enterprise IT themes of 2015: making it easier to define machine rules and combine these into predictive analytics; mobile-app UI; open architecture. One stand-out differentiator is that FIN 3.0 embraces tagging and data modeling conventions as defined by the Haystack Community. It will be a highlight of the upcoming Haystack Connect 2015 Conference to learn more about the rollout of FIN 3.0.
The conference will be another great place to observe how the cross-pollination of ideas among mobile computing start-ups, enterprise IT and commercial and industrial controls integrators are advancing the state of automation.