In an average winter, Kentucky’s Department of Transportation spends between $45 million and $75 million on road salting and snow removal. But during the harsh winter of 2014-15, the state’s spending exceeded projections, putting it one storm away from exhausting its entire snow removal budget. One factor contributing to the excessive costs were inefficiencies in how the department used its plows and spread anti-icing agents.
To fix the problem, the department turned to big data, analytics and edge computing. Using sensors in roads and on snowplows, as well as data from the popular driving app, Waze, along with information from weather reports and Twitter, staff and road crews developed a more precise and better-managed process for snow removal. That process included algorithms that analyzed in how fast cars traveled over roads, the number of vehicles on a road at a given time and other factors, such as temperature, wind speed and cloud cover. As a result, the department was able to use anti-icing and plowing schedules more precisely, which resulted in less waste and saved the department from 5 to 10 percent on snow and ice removal. Every 60 seconds, more than 80,000 records of information are fed into the department’s computers, where engineers are able to deal with snow on roads more efficiently and in real time. These advances are part of a broad trend where analytics, machine learning and edge computing are starting to benefit government agencies in ways not seen before, according to Dave Shuman, industry leader for IoT and manufacturing at Cloudera, the analytics firm that developed Kentucky’s snow removal program. “There’s been a big focus on using algorithms for machine learning in advanced analytics,” he said. “The tipping point occurs when you have lots of data and want to apply lots of compute to the problem.” Besides snow removal, versions of these kinds of data rich, analytical IoT solutions are popping up in a number of government functions, from the use of ankle monitors to improve prisoner and parolee rehabilitation and reduce recidivism, to body and dashboard cameras that improve law enforcement forensics to a host of infrastructure applications that include parking, lighting, traffic, transit and even garbage removal. A big reason this kind of robust, data-driven solution has become more popular is the maturation of edge computing. The consulting firm McKinsey projects IoT devices to reach 30 billion units by 2020 and government adoption to outgrow that of the private sector by 2019. While many of the sensors will simply gather data, they are also enhancing the ability for computing — and decision-making — to take place in cars, trucks, buildings and other government infrastructure, rather than back at a central IT location or in the cloud. “Processing data at the edge allows for low-level latency on your decision-making processes,” said Shuman. “So, you are able to have devices that have enough intelligence, compute capacity and models to make the decisions that need to be made.” Every day, more examples of edge computing emerge, from building elevators that know to stop and lock their doors open if there’s an earthquake, to water valves that can shut down if there’s a leak somewhere. Snapchat, the popular smartphone app, is another example of edge computing that crunches petabytes of data to create funny images of a person’s face. More significantly, autonomous vehicles operate with edge computing in order to make split-second decisions on how to steer through an urban environment. In recent years, edge computing has generated significant attention. Gartner, the IT research company, believes edge computing will bring value where localized computing power is necessary. As IoT decentralizes an organization’s computing infrastructure, edge computing will become an advantage over cloud or centralized computing. But state and local government IT executives aren’t ready to embrace the trend quite yet, certainly not going as far as to say edge will replace the cloud. Former Boston CIO Jascha Franklin-Hodge thinks edge computing is a bit overrated. “What cloud infrastructure has taught us over the last 10 years is that centralized, high-efficiency computing infrastructure in most use cases is going to outperform distributed, lower-efficiency systems in price, performance, scalability, resiliency and all the other things we value,” he told Government Technology. Rather than viewing edge computing as an “either/or” choice, Shuman believes a hybrid approach will prevail, thanks partly to the vast amount of legacy technology that still operates in the public sector. “While there are certain things where it makes sense to take advantage of edge compute analytics, other things will need to apply more significant context to make the right decisions,” he said. The type of hybrid combinations Shuman sees unfolding are platforms where sensors are combined to drive CRM, ERP and supply chain systems in government agencies. Still, government CIOs and agency heads need to conduct due diligence before embarking on these kinds of data-driven, analytical processes. Shuman warns about the possibility of built-in biases in computer models meant to solve a problem. “We have to be sure there is an open and transparent method to it,” he said. Because governments rely not just on their own data, but what comes from data brokers, they have to watch out for the introduction of new biases that could affect outcomes negatively. “It’s less about getting data and more about what you are going to do with it,” he said. “To do that at scale, you have going to see more machine learning, advanced analytics and combinations of data driving these new cases forward.”