TL;DR
Sharp Group's recycling plant in Rainham, east London, faces staffing challenges, leading to increased reliance on robots. The facility processes 280,000 tonnes of mixed recycling annually amid a hazardous work environment.
The dust at this busy recycling plant is pervasive and the steady noise of hoppers and conveyor belts makes this a challenging environment to work in.
The facility in Rainham, east London is owned by Sharp Group, a family-run skip and waste management firm.
Along the conveyor belts runs everything you could imagine, from shoes, to old VHS cassettes and blocks of concrete.
The team here processes up to 280,000 tonnes of mixed recycling every year with 24 agency workers on its rapid conveyor belts.
This is a hazardous industry. While Sharp Group is proud of its safety record, work-related injury and ill-health in the sector is 45% higher than other industries. And the fatality rate is a sizeable multiple of the national average.
These factors, along with the unpleasant nature of the work, mean keeping workers is difficult. Annual staff turnover runs at 40%.
"The belt is moving all the time, you're constantly picking. I go through a lot of pickers because they just aren't up to the job," says line supervisor Ken Dordoy.
The firm rotates pickers through different materials every 20 minutes, and I could see the belt is stopped periodically for respite.
A potential answer to that high-staff turnover, was also on the line when I visited. A robot, known as Alpha (Automated Litter Processing Humanoid Assistant) was being trained to pick through the rubbish.
Built by RealMan Robotics in China, it's being adapted for real-world recycling operations by the British firm TeknTrash Robotics.
Automated robots are not new to the sector, but the use of a humanoid is unusual.
TeknTrash founder and CEO Al Costa argues that copying human movement allows his robot to fit into existing plants without redesigning the machinery.
Alpha is not up to speed yet, instead, it's on a training agenda and being guided through arm movements. Next to it, a plant worker wears a VR headset to record his own endeavours to demonstrate what successful picking and sorting looks like.
The learning process is two-fold. The first is identifying what's on the conveyor and the second part is actually lifting up items.
Costa says this is exactly what early-stage training looks like.
"The market thinks these robots are prêt‑à‑porter, that all you need to do is to plug them to the mains and they will work flawlessly. But they need extensive data in order to be effectively useful."
He showed me how a system called HoloLab delivers data from multiple cameras to train Alpha.
They warn it what's coming, they guide its arms, and they report failures if unpicked items stay on the belt. The passing of thousands of items delivers millions of data points every day.
The training might take time, but if it works, it could make life much easier for the firm.
"The attraction of a humanoid is that you can put it here and it stays here. It will pick all day, 24 hours a day, seven days a week. It's not going to apply for a holiday, it's not going to have a sick day," says Chelsea Sharp, plant finance director and granddaughter of company founder Tom Sharp.