Last week’s conference gave me the exciting opportunity to discuss Artificial Intelligence (AI) in the waste sector. Having had several discussions from colleagues across the sector, we all agreed that AI has a place in improving safety, but the consensus was that it’s still a long way off being ‘Intelligent’.
Whilst risk assessments can now be conducted using a mobile phone with a camera, which can scan a site to identify hazards, this does raise concerns around competency, compliance and reliability. It also requires ‘knowledge’ to identify the hazard to include within the assessment.
It’s apparent that one of most significant pitfalls with AI relates to the requirement for machine learning. Essentially, AI systems need to be ‘trained’ for a specific purpose. Whether it is used to identify a person operating at the rear of an RCV, completing a risk assessment, or identifying hazardous materials at a HWRC, we need to remember that one of the fundamental concepts of AI is pattern recognition or accessing a digital image bank for verification.
AI can work very well alongside robotics on a MRF sorting line, which use pattern recognition and probability to identify items such as plastic pots, tubs or trays. These systems cross-reference the image of a single item against potentially tens of thousands of images, which have had to be uploaded to a digital image database.
These optical sorting systems using robotics and AI can work very well as long as they are maintained. It’s safe to say (excuse the pun) that AI and robotics can improve health and safety in the sector with automation reducing manual handling risks. However, safe systems of work are still required to separate machinery from people!
When it comes to operating at the rear of an RCV, it’s not quite so simple as identifying a yogurt pot against an image bank because operatives, PPE, bins/bags and waste material all vary in size and shape, and don’t remain static. Therefore, a variety of systems are required to identify and differentiate between a body and an overloaded bin present during the lifting cycle.
I’m aware that a German company has teamed up with a camera vision company that has experience with HGVs operating in London to alert drivers to cyclists and pedestrians in blind spots. A new AI camera system is being developed to see whether it’s possible to identify human form identification, and early indications are that this looks quite promising. It’s clear that vehicle manufacturers are providing new and innovative solutions to improve safety, but we are still a long way off from a true AI system and it’s unlikely that Skynet will take control anytime soon!
