Cloud Brigade has been providing trustworthy and innovative IT consulting services to customers, clients, and partners since 2005. The Cloud Brigade team works with businesses of all sizes to identify pain points and solve pressing challenges that other consultants simply can’t. How? Cloud Brigade’s team of Machine Learning and AI specialists.
Cloud Brigade is located in Santa Cruz, CA, which is home to an infamous intersection and stretch of road that is known for causing serious delays, backups, and frustration for local drivers. A single 2.5 mile stretch of local highway can take more than 30 minutes to pass, mostly as a result of inadequate and unresponsive traffic signals.
Realistically, road congestion and driver frustrations aren’t exclusive to Santa Cruz. Drivers around the country can likely attest to increasing traffic congestion, delayed travel times, and backups during seasons of heavy tourism.
The road management systems that worked decades ago no longer suffice. Traffic light systems that run on timers and sensors are no match for the demand of modern traffic patterns, and manually adjusting these systems is both tedious and expensive. As a result of this local issue, Cloud Brigade hit the pavement to develop a system that focused on aspects of Machine Learning, Computer Vision, and the Internet of Things (IoT).
Chris Miller, founder and CEO of Cloud Brigade, took a hands-on approach to this local pain point. In 2019, Chris had previously communicated with city officials in San Jose, California, to update existing traffic systems to a more modern and intuitive system. Although Cloud Brigade didn’t commit to the San Jose project, Chris was inspired by the potential of a new technology from Amazon Web Services (AWS) called DeepLens.
The DeepLens is a functioning computer attached to a video camera, and can employ real-time data about traffic patterns and vehicle use. (Picture credit: Amazon)
The AWS DeepLens camera is a functioning computer attached to a video camera, and it can employ real-time data about traffic patterns and vehicle use. To take this technology one step further, a Machine Learning model attached to the DeepLens microprocessor can read and interpret the direction of vehicles and where they intend to turn or go.
In an effort to reduce the resources needed to maintain such data processing, Cloud Brigade’s local Machine Engineer, Mark Davidson, used Raspberry Pi to run the models in the field (in this case–key intersections), and sync those data points to a cloud-based server.
Although this is a tailor-made solution for a particular city, stop lights around the United States rely on antiquated timer systems that circle through a set loop before changing traffic light patterns. In other cases, motion sensors detect traffic in a queue and only respond when there is a physical presence.
The downside of this old technology is that while it’s suitable for average traffic scenarios, it tends to fail whenever there’s a higher than average flow of traffic; in turn, this creates bottlenecks. These methods are simply not intuitive enough to learn real vehicle patterns, respond to live data, and make adjustments to improve the experiences of drivers.
In contrast, a Machine Learning or IoT approach is scalable and widely relevant. Minimizing wait times and improving traffic patterns is a top concern for many municipalities, and AI can accomplish those goals in significant ways. The power of AI can especially help smaller towns and counties that have limited resources and budget allocations for fixing traffic issues that only happen at predictable times throughout the year.
Once the initial system was developed, Cloud Brigade began to build a Reinforcement Learning (RL) Model to “reward” the system based on desired traffic results.
The last step in the process involves packaging the two Machine Learning models, the DeepLens Camera, the AWS Cloud Computing instance, and the IoT traffic signal control product as one single end-to-end AI device that uses Raspberry Pi and AWS cloud infrastructure to react to live traffic scenarios.
In the near future, Miller and his team also envision the possibility of creating a virtual world in which real traffic data supports hypothetical simulation and experimentation. All of these steps contribute to improved artificial intelligence, infrastructure, safety, and more.