One way of looking at IoT deployments is this \u2013 a large array of not-particularly-sophisticated endpoints, mindlessly sending individual data points like temperature and pressure levels to either an edge device somewhere on a factory floor, or all the way out to a cloud back-end or data center.\nAnd that\u2019s largely correct, in many cases, but it\u2019s increasingly not the whole story \u2013 IoT endpoints are getting closer and closer to the ability to do their own analysis, leading to simpler architectures and more responsive systems. It\u2019s not the right fit for every use case, but there are types of IoT implementation that are already putting the responsibility for the customizing their own metrics on the devices themselves, and more that could be a fit for such an architecture.\n\nThere are three main areas where letting the endpoint do its own data analysis \u2013 in whole or in part \u2013 is becoming increasingly common \u2013 smart cities, industrial settings and transportation.\nIoT in smart cities\nIn smart cities smart cameras can do certain kinds of analysis right there on the device, helping planners understand pedestrian and motorized traffic patterns.\nThe difference between doing analytics completely on an endpoint device or partially on a device is an important one, according to Gartner research vice president Mark Hung. At the core, the analytics done by IoT implementations is about machine learning and artificial intelligence, letting systems take data provided by smart endpoints and fashion it into actionable insights about reliability, performance, and other line-of-business information automatically.\nApplying the lessons learned from sophisticated ML is easy enough, even for relatively constrained devices, but some parts of the ML process are much too computationally rigorous to happen at most endpoints. This means that the endpoints themselves don\u2019t change their instructions, but that they provide information that can be used by a more powerful back-end to customize a given IoT implementation on a per-endpoint basis.\nThe case of video analytics for smart city applications like traffic monitoring \u2013 using a system where the cameras themselves track pedestrians and motorists, then score that data against a centrally-created AI model \u2013 is an instructive one.\nEvery intersection is different, so trying to push an identical rubric for making sense of different traffic patterns and volumes to the cameras monitoring every intersection isn\u2019t going to work. Each intersection needs its own rubric. Yet the AI training that\u2019s needed to come up with them requires heavier computational lifting than the cameras alone can provide, so it has to be done somewhere in the back end. The cameras themselves can apply the lessons learned by the AI model, but they need more powerful hardware to intelligently change the instructions they\u2019re given.\n\u201cSo, to come to some preliminary analytics at the endpoint and then send that back out for further training, you have kind of a federated learning [system],\u201d Hung said.\nIndustrial IoT\nAnother key area for endpoint-based IoT analytics is the industrial and manufacturing sector. Joe Biron is the CTO of PTC, a Needham, Mass.,-based software company that makes ThingWorx, an industrial IoT software platform. Biron said that PTC\u2019s been trying to get intelligence into industrial machinery for about a decade now, with the idea being to help companies save money via predictive maintenance and other automated management and operational applications.\n\u201cTen years ago, the state of technology for doing proactive and predictive failure detection \u2026 wasn\u2019t exactly a life-changing kind of technology,\u201d Biron said. It was largely a human-intelligence-driven process that relied on a technical specialist\u2019s intimate knowledge of how the industrial components worked. Based on that knowledge, rules for detecting the parameters that predict impending failure could be hard-coded into even the \u201cdumbest\u201d of endpoints.\nThe real challenge comes when there\u2019s no one person familiar with the critical confluence of indicators that predicts a problem in the offing. For this, you need machine learning, and more specifically, a machine-learning model that can score data inputs vs. outcomes and sift out which data points are the most important to making the predictions. That\u2019s computationally expensive, according to Biron, limiting its ability to be handled on endpoints.\n\u201cOnce you\u2019ve created the model, however, now you\u2019ve got something very lightweight to score against; now this model can be fed real-time or near real-time or microbatches of recent data, and it can be used to make statistical determinations of whether the \u2026 predicted event may happen,\u201d he said. \u201cThe scoring of the model is cheap, computationally, but the training of the model is expensive.\u201d\nHumera Malik, CEO and founder of Canvass Analytics, said that the federation of these endpoints \u2013 and anything that\u2019s got a digital sensor connected to it on a factory floor is an endpoint \u2013 is critically important in the industrial sector.\n\u201cIt could be a shaft, it could be a bearing, it could be any of the assets \u2013 a turbine, a generator \u2013 all of these different assets that then, collectively, are running this process,\u201d she said.\nOn-device IoT analytics also work well in industrial settings because there, applications of IoT tech tend not to be delay-tolerant. The time it takes for data to leave a device, negotiate a complex network topology and return in the form of corrective instructions can be too lengthy for effective device management.\nIoT and smart vehicles\nThe third \u2013 and probably least well-realized \u2013 area where endpoint IoT analysis is getting popular is transportation. Hung notes that anything that requires autonomous navigation, be it a drone or a car or anything else, is a great candidate to be a relatively smart IoT endpoint.\nCars have been getting more and more heavily automated and computerized for years, and the advent of widespread IoT has only accelerated the process, as manufacturers build increasingly sophisticated smart safety features into modern vehicles and fleet management gains new tools for maintenance and tracking.\nThe increasing automation of the automobile is a great example of how this type of semi-autonomous IoT tech is supposed to work, according to Ruhollah Farchtchi, CTO of Zoomdata. \u201cThat virtuous cycle of human understanding being translated into algorithms and machine learning being deployed at the edge is a lot more of where we see the edge analytics taking shape and taking form,\u201d he said.\nIoT\u2019s future in healthcare and energy\nLooking ahead, healthcare and energy production \u2013 particularly in the oil and gas industries \u2013 are poised to become growth areas for on-device IoT analysis. Hospitals and clinics are crying out for smarter technology \u2013 witness the work being done to reduce alarm fatigue and boost interoperability in clinical environments \u2013 and having more capable computing technology built into endpoints could be a massive boon to patient care.\nThat\u2019s not to say there aren\u2019t headaches involved, particularly where the question of machine learning comes in, according to Biron. The requisite back-end for the heavier computational lifting part of the process isn\u2019t as easy to build into a medical facility\u2019s architecture.\n\u201cIt\u2019s easier to see medium-scale computation happening in an environment like [a factory floor], as opposed to let\u2019s say, a clinic, where a medical device is living \u2013 the ability to fold in high-density compute is more limited than with a manufacturing facility,\u201d he said.\nThe oil and gas industry has a particular advantages on that score, however, given the wealth of historical data about exploration and extraction available for use in training machine learning models.