Grant Powell MBCS spoke to Professor Julie McCann, Head of the Adaptive Emergent Systems Engineering (AESE) group at the Imperial College London’s Department of Computing to discuss how data-driven insights can help to develop greener, cleaner cities.
Sensor technology, artificial intelligence (AI) and machine learning (ML) are improving our understanding of how today’s cities operate, unlocking significant value. The deployment of sensors throughout a city, connected via the internet of things (IoT), can facilitate a greater understanding of a number of key elements including population movement, vehicle movement, and numerous environmental factors. When vast quantities of data are analysed using ML and AI, this produces powerful intelligence which is leading to the development of truly smart and infinitely more sustainable city environments.
How can data analysis inform more sustainable operations?
Analysis of the data that sensors collect enables us to better understand what’s happening in a city at a decent level of granularity. Armed with such data, it is then possible to build models of the city to allow a better understanding of where changes can be made that will lead to desired improvements. Using pollution levels as an example, we may want to look at how changing the shape of a road or using different traffic light configurations could improve traffic flow to reduce congestion, or even where traffic flow could be limited — perhaps through the introduction of more pollution-light zones, such as pedestrianised areas. With the knowledge that sensor data makes accessible, we are able to start to make optimisations that improve the overall sustainability of the city.
Where do machine learning and artificial intelligence come in?
When you’re modelling at a city scale, those models become very complex. This is where we can make use of ML and AI. The levels of complexity are difficult to accurately model at scale, including the many interactions between all of the sub-models. ML can be used to analyse vast amounts of data and look for relationships and patterns of causality. If any issue arises, then we can start to fix it. Beyond ML and AI, we can build mathematical representations and use fluid dynamics to determine how materials and other mediums move in space, and use this to determine how changes and improvements can be made and refined.
How can teams decide when AI or ML should be used?
It’s important to have a firm grasp on exactly what you are trying to achieve, and an awareness of the actions necessary to achieve it. This is particularly important from a sustainability perspective where you really need to be mindful about the use of energy. If you’re going to be putting a significant number of devices and hardware to use, this will of course result in the consumption of energy and production of heat. Take into account the actual cost of computing too; the machines, the sensors, and the materials going into those things.
If you don’t really need an ML system, if it’s something that could be modelled, a lot of the algorithms that use traditional or signal processing-based analytical models can help — and they’re an awful lot less heavy in terms of computing. I’d recommend being a bit conscious about what you are using your computing for and the materials that go into the devices that you deploy. Yes, AI and ML have great potential, but where we’re being cautious about our impact on the environment it makes very little sense to burn energy and resources in an attempt to become greener. It’s about carefully considering the right tool for the job.
How can sensor technology be designed to integrate with existing architecture?
In most cities, unless you’re developing from scratch — which is a real rarity across the planet — it is retrofit. There might be concrete from 50 years ago and water networks from over 150 years ago, and you have to account for all that and work around it. We talk a lot about ‘digital twins’, these big models, and here they become invaluable. For example, I have a 150-year-old pipe and want to use whatever new material to fix it or to be embedded within it. What is the impact of that over the next 100 years? What is the sustainability impact of that? You can start to ask those kinds of interesting questions and model the outcomes.
Are there issues concerning the lifecycle of IoT and sensor devices?
This is actually an ongoing issue. A lot of the technology that we use to make cities smarter is designed to last no more than, say, ten years, because technology becomes obsolete very quickly, software needs upgrading, and so on. Yet when we’re looking at city infrastructures, we’re talking a lifespan of 50, 60, 100 years and longer. There’s a mismatch here that we really need to think about. If we’re going to increasingly be designing these intelligent cities with all this smart technology, we need to look at how to ensure that the technology remains upgradeable, and therefore current.
When it comes to new structures, does the approach to tech integration change?
While it is more likely that technology will be developed to fit an existing city, it’s when we start to future gaze and look to perhaps build new buildings, which might integrate sensor technology within them as part of the build, that many new possibilities are uncovered. At Imperial College London we’ve been involved in a project with a Dutch company, MX3D, which created a 3D printed footbridge.
For you
Be part of something bigger, join BCS, The Chartered Institute for IT.
Using a network of installed sensors our researchers have been measuring and analysing the performance of the bridge as it handles pedestrian traffic. This also opens up exciting possibilities around 3D printing other structures in situ, and even progressing to the stage where the sensors and other technologies involved with data collection and analysis can be 3D printed and actually integrated into the structure as it is constructed.
How can your work with smart cities lead to better use of resources?
We’re increasingly pushing for more circularity in cities. As technology and its potential continues to grow, so we’re better able to evaluate the use of resources, make better predictions around the generation of waste, and identify opportunities for circularity. It’s then just a case of making some key decisions to make the city and its assets more efficient, more economical and more sustainable. Could, for example, a data centre be positioned close to a city, perhaps directly next to a communal heating system so that the heat from one can go into the other? We can model this and understand whether it is worth the cost to move the data centre to that position versus somewhere where it is cooled using natural elements. Those trade-offs can start to be made based on data analysis rather than assumption. Understanding circularity much better will be one of the major benefits of the evolving smart city.
Where will ongoing research and development in this area take us next?
There’s a lot of debate about urban farming at the moment, and while one could argue that it’s not particularly viable from a business perspective, it might have other viabilities that we could work out through intense modelling or use of ML. Automating the farm, for example, could ensure that precise amounts of food, water and heat create optimal conditions for plant and crop growth, and that surplus energy and waste is repurposed. For me, this is the next phase; the city that really thinks and can use data to drive intelligent automated actions. The city will not be ‘eco’ for the sake of it, but rather designed around an ability to weigh-up sustainability gains based on energy use, waste generation and a long-term outlook that can help predict which projects should be prioritised. ‘Smart by design’ is the term, and I think this is where we’re going next.
