Future of Smart Cities and Real-time Traffic Predictions (Part I)

by Brad Johnson, on Oct 11, 2017 11:00:00 AM

This is part 1 of a two part series. 

New innovations in Machine Learning, Edge Computing, and real-time data analytics are shaping the development of Smart Cities projects around the world. One benchmark use case for Smart Cities efforts is that of Intelligent Transportation Systems (ITS), which include the use of connected vehicles and transportation infrastructure to automate or optimize transportation systems. With 58% of in-development Smart Cities projects relating to either Smart Buildings or Smart Transportation efforts, and 250 million connected vehicles expected to be on the road by the year 2020, the adoption of Smart Cities technologies will have a transformative effect on the transportation systems of the future.

When utilized together, Machine Learning, Edge Computing, and real-time analytics technologies can make traffic systems smarter, cheaper, and more efficient for both public operators and citizens. In this post, the Swim team breaks down how Machine Learning and Edge Computing technologies will shape how cities can use data to the improve transportation systems of the future. We discuss real-time analytics in part 2 of this series.

Future Smart Cities 1_No URL_image

Machine Learning

Today, many traffic signals are still analog systems, which are only capable of sensing voltages and other raw outputs. These outputs create a massive amount of low value data, which is nearly unusable by traditional Big Data architectures. There is simply too much raw data, and too little significant information contained within, to make any sense of it all. But Machine Learning is ideal for finding patterns hidden within oceans of “noisy” data.

For example, Swim ESP™ has been deployed by the City of Palo Alto to monitor their traffic networks, where we have already made upwards of 240 Trillion inferences based on 1 Trillion data points (voltage readings and other sensor data). These 240 Trillion inferences are then aggregated to form predictions for light signal cycling and other traffic events. A single deployed instance of Swim ESP is currently generating 160K new inferences per second, predicting traffic signal changes with extremely high accuracy.

Machine Learning empowers cities to learn on all the high volume, low value sensor data created by existing traffic systems. By observing the “random” signal cycling of (unscheduled) traffic signals, Machine Learning algorithms can detect patterns within the voltage data use those patterns to predict when traffic signals will cycle. Because Machine Learning is able to form predictions from the raw analog voltage data, this means that municipalities can upgrade existing analog traffic signals for modern Smart City efforts instead of replacing them, a significant cost savings.

Edge Computing

Edge Computing is the processing of data at the “edge” of a network, which can be either embedded on or deployed near field devices. Edge computing is a distributed computing model that leverages the various edge devices on a network, in order to more efficiently compute sensor data as it is created. As Moore’s Law continues to provide cheaper (and smaller) computing power, individual devices are becoming their own datacenter. Edge computing provides the networking framework to leverage these “data centers on the edge,” allowing developers to compose applications logically, and compute time-sensitive data on the edge, while pushing refined data to the cloud for further processing.

Edge Computing technologies enable traffic operators to run Machine Learning instances and perform edge analytics locally, where analog traffic data is created. This eliminates the need to send raw traffic signal data over a network for central processing, which had been a major bottleneck for utilizing traffic signal data previously. Furthermore, Edge Computing allows municipalities to create “edge networks” between connected traffic devices. These edge networks link traffic devices, forming a distributed computing architecture across a region, and can provide a backbone for vehicle-to-infrastructure (V2I) and other connected infrastructure applications.

This is a part 1 of a two part series. Read more about the future of Smart Cities in part 2.

Learn More

Learn how SWIM brings together the benefits of Machine Learning, Edge Computing, and real-time analytics to power the next generation of intelligent transportation and Smart Cities applications.

Topics:Machine LearningSWIM SoftwareIndustrial IOTSmart CitiesEdge AnalyticsSWIM AI