Researchers to teach robots how to differentiate between sandwich ingredients

Loughborough University computer scientists have teamed up with a food production automation company for a project that looks to teach AI robots how to differentiate between food items so they can make sandwiches in real-world factory environments.

Project lead Dr Baihua Li, of the School of Science, hopes the research with Millitec Food Systems will “empower a range of new automation functions on food production conveyors and pipelines”.

Millitec has developed a range of technology that can assemble and pack sandwiches, with the intention of reducing intensive labour work required on production conveyors.

Its current sandwich-assemblers are able to assist in the sandwich making process – for example, they can butter bread and cut sandwiches – but they are unable to undertake tasks that require them to differentiate between items, such as selecting salad toppings to put inside a sandwich.

The project with Loughborough University looks to take the robotic technology to the next level of intelligence.

Dr Li and her team of researchers are to develop machine vision and embedded artificial intelligence (AI) technology for the robot manipulators that will enable them to recognise a variety of items and carry out more advanced tasks such as picking-up the right elements in the right amount and order.

They will investigate using a low-cost reliable camera and sensing system on Millitec’s ‘Delta robots’, which, in essence, will allow the machines to ‘see’.

They will then develop algorithms that train the robots to recognise various breads and food ingredients, and accurately detect their location and orientation.

The academics will create the algorithms using deep learning – a method of teaching a machine how to carry out tasks automatically without explicit human instructions.

This involves collecting thousands of images and getting a computer system to analyse them, so it learns what different items look like.

As well as being taught how to identify food items, the Millitec robots will be trained to spot and deal with errors on production conveyors, such as missing ingredients or items in the wrong position.

The developed technology will be deployed as an embedded system on Millitec’s existing automation machines.

Richard Ledger, managing director of Millitec, said: “We’re thrilled to be developing the latest range of food production robotics in conjunction with Loughborough University. This development project will be a major leap forwards in food production automation.”