Monash University researchers are investigating how robotics can be utilised to reduce the amount of construction and demolition waste going to landfill
With technological advances and artificial intelligence opening up ways in which work processes can be made safer and more efficient, one area that could see significant gains is in the processing of construction and demolition (C&D) waste.
A research team at Monash University’s Automation and Sustainability in Construction and Intelligent Infrastructure Lab has begun work on developing robotic sorting arms that can recognise and sort individual pieces of material from C&D waste, with the aim of reducing the amount of waste dumped in landfill and creating a safer work environment for material processing staff.
The initial stage involves photographing the contents of construction site skips across Melbourne and training computer vision-based models on how to recognise and categorise a wide range of materials.
This was led by Monash PhD candidate Diani Sirimewan, who says that C&D waste was chosen as the focus for this research due to the untapped potential for achieving greater levels of recycling, through artificial intelligence including computer vision and robotics.
“Whenever a building is constructed, demolished or renovated, a vast amount of waste is generated,” she says.
“Unfortunately, most of this waste is disposed of in landfills and the amount of C&D waste is increasing day by day. This obviously creates environmental, land use and economic challenges, but the potential exists to recycle many of the materials found in this waste.
“For example, the concrete pieces that came from demolition sites can be washed and reused as road base or backfill material, while timber can be reused in landscaping. However, in skip bins at construction sites, the waste is mixed together and can contain concrete, timber, metal, glass, plasterboard, plastic, etc. So, we have to sort it and separate it for recycling.”
Risky work
It is this sorting of C&D waste to separate out recyclable materials that poses significant challenges for waste sorting facilities and their staff, with the process being time consuming and potentially hazardous.
“At the moment, sorting is a manual process where workers have to laboriously sift through these heavy pieces of waste, which are often contaminated with material such as asbestos, exposing themselves to occupational health and safety risks,” Sirimewan says.
“Therefore, we propose the solution of employing computer vision-based technology to automate this waste handling process at material recovery facilities where the system-assisted cameras identify the composition of materials in cluttered waste streams, including the contaminants, and pick out the valuable items that can be recycled using robotic arms.”
Sirimewan says that it isn’t just a safer approach to sorting waste – using robotic arms could also speed up the process and increase the amount of recyclable material separated out.
“One site we visited had more than 60 bins of construction waste to separate and the process currently is to lay out all of the waste on the floor and have eight to 10 labourers separate out valuable materials, with the rest sent to landfill,” she says.
“In that case more than half of the waste went to landfill because they had so many bins to process that day.
“Although we haven’t reached the point of comparing our robotic sorting arms with manual picking, previous studies have shown that robots are more efficient at this process and have already been employed in some countries.”
The system uses a method called segmentation to recognise waste objects, where each object in a waste stream is scanned and analysed. If a suitable match is found, the item is categorised as being a particular material.
“Let’s say there’s a piece of concrete inside the mixed construction waste,” Sirimewan says.
“The model draws a boundary around the object and it can mark it with a particular colour. In our proposed model, concrete was coloured green while timber was blue.
“I am currently working on a user-guided segmentation system, where the user can draw a box around an object and the model will automatically segment it, or you can use a text prompt (i.e. timber) and the model will automatically segment the timber particles it can see within a waste bin for example. This can be used for pre-screening of materials as well as sorting of the waste.”
Removing contamination
With concerns about recycled mulch and soil derived from C&D waste being contaminated with asbestos and other foreign objects, Sirimewan says a similar system could be employed as a quality control measure for recycling businesses, which will require further research.
“Over the next few months of my PhD, I’m focusing on anomaly detection, to detect contaminants that fall outside of the main material categories like concrete, timber, cardboard and plastic, so the system can flag these to the user.
“This is for quality assurance of a recycled product, so before sending it out to customers, you can check whether there are anomalies. In the case of asbestos being found in mulch, if you’re recycling timber for example, you can lay out the recycled material on a conveyor belt and check for contaminants.”
This research was published in the Journal of Environmental Management and can be accessed here. For more information, email Diani Sirimewan at diani.sirimewan@monash.edu
- Subscribe to our free weekly newsletter to receive the latest news in the earthmoving industry
- Never miss a great deal and subscribe to our monthly magazine
- Download a free copy of our latest digital magazine to catch up on the biggest news and developments in the earthmoving industry