Reducing CO2 emissions with radio base stations

The rising level of greenhouse gas emissions is driving climate change, and it’s estimated that transport is responsible for nearly a quarter of global energy-related CO2. Electric vehicles and self-driving cars are expected to lower these emissions, but we’re potentially decades away from a world where transport runs solely on electricity. Unfortunately, electric vehicles still also produce non-exhaust emissions [1]. Armin Catovic, a machine learning specialist at Ericsson, has found a new cost-efficient way to significantly reduce traffic congestion and pollution in urban areas today. Here’s what he had to say about his latest innovation.

In one sentence, what is this new solution?

With this solution, we can turn radio towers into traffic sensors that continuously measure congestion and air pollution while generating real-time data to inform city planning.

What challenges are you trying to solve with this platform?

This project solves two main challenges. First, from a societal point of view, we need to reduce the amount of pollution generated by road traffic. Secondly, from a city planning perspective, the process of measuring traffic congestion and collecting accurate data is usually time consuming, expensive and ineffective. For example, installing new on-road traffic sensors such as induction loops and ANPR cameras is a slow and costly process.  Additionally, many existing traffic sensors are faulty or erroneous, or require data to be physically extracted onsite. As a result, cities normally run one-off projects to try and measure congestion levels, which provides a limited set of data for them to work with.

How does this solution solve this challenge?

I knew we needed a solution to these challenges, and I also knew that radio signals transmitted and received by radio towers provide a lot of information about what’s happening in the local area. With this solution, we use machine learning to turn radio towers into sensors that measure the amount of traffic in their vicinity, as well as air pollution levels in that area.

Radio towers, such as those used in 4G mobile networks already exist everywhere, and with this solution, cities and local authorities no longer need to purchase and install new traffic sensors. We can also provide access to an ongoing stream of real-time data, which cities can use to inform capacity planning and reduce congestion, for example, by adding speed bumps in certain places or widening roads. They can also use this data to implement dynamic measures based on real-time traffic flow, such as variable traffic light timing and variable speed limit, which has shown to reduce traffic congestion by 50 percent [2], reduce travel time by 5 to 10 percent, and reduce accidents by 25 to 50 percent [3].

GUI screenshot

Screenshot from the MVP showing traffic flow on different roads in Stockholm; traffic flow is estimated by machine learning models using only the radio base station counters.

What’s the next step?

We are currently trialing this solution with customers in the Together with West Midlands 5G Limited and Vodafone UK, we’ve proven that the solution works and that it provides highly accurate estimates of how many cars and buses are on the road at any given time – so it has been validated from a technical perspective. The next step is to package the solution up into a product or system with a nice user interface and present it, along with the results from our initial trials, to some of our other customers.

What are the main benefits of developing the project within Ericsson ONE?

"I'm receiving a lot of support from a business and commercial standpoint, and the team is giving me lot of insights into how to develop marketing strategies and business models..."

- Armin Catovic, Machine Learning Specialist at Ericsson

I’m receiving a lot of support from a business and commercial standpoint, and the team is giving me lot of insights into how to develop marketing strategies and business models, and how to promote the product.

Ericsson has existing customer relationships in many different industries which can be leveraged to generate interest in the product. From a long-term product development perspective, this also makes it easier to explore new use cases in other areas.

It has been a real pleasure working with the team at Ericsson ONE, they aren’t just good at what they do, but are genuinely lovely people who make my day-to-day life at work more enjoyable.

What drives you to be an intrapreneur?

Intrapreneurship means you must do a lot of the same things you do as an entrepreneur, such as coming up with an idea, proving it’s feasible and building a team, but the difference is that if it fails, there’s no personal risk involved whatsoever. The tradeoff is that you work in a more constrained environment, for example, you can’t just hire whoever you want to work on your project, but sometimes this tradeoff is worth it.

I’ve always been curious and enjoyed creating new things, but at heart I’m a real skeptic. I’ve seen a lot of ideas that don’t actually work, and the desire to create something actually useful that has a clear positive impact is what drives me.

About Armin

Originally from Melbourne, Armin worked for a handful of tech startups before joining Ericsson in 2008. During his time at Ericsson, Armin spent a lot of time traveling, working on projects in Indonesia, Bangladesh, Singapore and the US. The highlight of these journeys for Armin was meeting the local people.

When he’s not trying to solve the world’s problems with machine learning, Armin spends time with his wife and two young children in Stockholm.