Cleaning up cities with machine learning, AI

In 2018–19, Australia generated an estimated 27 million tons of waste from the construction and demolition sector — 44% of the total national waste. The sector’s contribution to waste has grown by 32% per capita over the previous 13 years.

Now, UNSW Sydney researchers are developing a new suite of design applications that aim to help architects and urban planners optimise their designs for greater sustainability.

“The construction industry produces an enormous amount of waste. 10–15% of all the materials you bring onto a construction site are going straight into the bin,” said lead researcher Associate Professor M Hank Haeusler, Director of Computational Design at UNSW’s School of Built Environment.

“It’s wasteful, it’s bad for the environment, and it doesn’t align with the United Nations’ Sustainable Development Goals [that promote inclusive, safe, resilient and sustainable cities and environmentally responsible construction],” the entrepreneur and designer said.

“We’re applying a computational eye to these global problems. As researchers we have a moral responsibility to investigate landfill, pollution, the way different materials contribute to climate change.”

The design applications use machine learning to reduce construction waste and urban heat and minimise embedded carbon footprint of buildings.

They aim to help minimise the environmental footprint of buildings by assisting built environment professionals in making more sustainable decisions around size, scale and materials, said A/Prof Haeusler, who works at the intersection of digital technologies, architecture and design. His expertise lies in computational design, including AI and machine learning, digital and robotic fabrication, virtual and augmented reality sensor technologies and smart cities.

Australian cities are experiencing unprecedented levels of overheating. Urban overheating arises from human activity such as waste heat from industry, cars and cooling, building with heat-absorbing materials and rapid urbanisation, and adversely affects health, energy and the economy.

Machine learning can interrogate vast sets of fine-grain data in real time to analyse and evaluate alternatives, A/Prof Haeusler said. In a design context, it can identify efficiencies and promote sustainable practices — in this case, reducing the heat and waste produced.

“[Within the UNSW heat reduction app,] you design your street and then a computer program does the calculation in the background [based on intelligence learned from its data sets. Then it tells you,] it looks like here, at this intersection, it will get hot because of the physics that shape urban heat islands.” 

The designer can then adjust the building height, put in green spaces and shade, change the road width and adjust other variables to improve the building’s environmental footprint.

Using machine learning, the heat reduction app helps users identify design inefficiencies that can cause urban overheating. Image: Daniel Yu.

Similarly, the UNSW waste reduction app calculates the materials required for your design and allows you to adjust its size and scale to reduce waste offcuts. Its calculations are populated with data from public hardware sites, like Bunnings.

A/Prof Haeusler is also working with architecture studio COX Architecture to develop research projects and promote educational opportunities for students. Giraffe Technology started as one such project, and is now an SME working on a digital architectural and property development application.

The one-time startup, funded by Atlassian’s Startmate accelerator program, grew out of a series of research projects aimed at making local (council) development data sets more accessible and facilitating feasibility studies for the city of Western Sydney. Giraffe Technology is like a map of the world on a browser primed for architects, he says, which means anyone with access to the internet can use it. It taps into GIS mapping to populate streets, buildings and vegetation.

Top image credit: ©stock.adobe.com/au/peshkova