Scottish scientists design robot to clean train carriages

A train-cleaning robot, designed by the National Robotarium, is set to keep seating spotless for the travelling public.
The robot has been designed by the National Robotarium and resembles K9 from Doctor WhoThe robot has been designed by the National Robotarium and resembles K9 from Doctor Who
The robot has been designed by the National Robotarium and resembles K9 from Doctor Who

The device, which looks a bit like Doctor Who’s famous K9 robotic dog, is mounted on four small wheels with flip-out brushes and will complement humans by cleaning the hard-to-reach places between and under the seats in carriages.

A National Robotarium spokesperson said: “To find waste it has a static camera in front and a stereo vision camera, mounted above, with a 180 degree field of vision.”

Hide Ad
Hide Ad

Supported by UK and Scottish Government funding of £21 million and £1.4m respectively, scientists from Heriot-Watt and Edinburgh University, at the Robotarium, have spent two years working on the robot’s design and operating system.

Research has shown that passengers want a higher standard of cleanliness on trains and a 2020 survey of 50,000 people, at the height of the pandemic, found more than 25 per cent wanted cleaner carriages.

Dr Mustafa Suphi Erden, from the National Robotarium, who is leading the project, said: “Our research uses robotics and AI to help people to solve a wide range of challenges.

“With the daily pressure on rail services, it’s essential that trains are cleaned as fast and as efficiently as possible.

Hide Ad
Hide Ad

“However, at present, this process is done entirely by hand requiring a significant amount of time for the cleaning personnel to collect each waste item one-by-one from under and in-between the seats.”

The design team has analysed more than 58,300 studio images of waste, in a variety of conditions, to help robots identify rubbish more accurately.

In addition, smaller datasets were studied, of actual waste photographed on trains, taken from the perspective of a cleaning robot.

The studio images were designed to provide as much data as possible about the features of each piece of waste, so every item was photographed at many different angles.

Hide Ad
Hide Ad

Waste items that can deform, such as newspapers or coffee cups, were represented in different states. In interviews with rail service providers, the team obtained further information regarding operating conditions to help guide the design of the robot.

The narrow under-seat area, which contains the most waste items, is extremely tight especially on older trains and this makes waste collection challenging for both humans and robots.

Some spaces were measured at just 28cm tall, with entry points as small as 31cm.

Dr Erden added: “Reaching underneath seats repeatedly over a long shift can lead to health problems.

Hide Ad
Hide Ad

“Also, cleaning staff regularly encounter hazardous and biological waste which poses a significant risk. As well as being important for health, cleaning can prevent train delays with discarded newspapers identified as a reason why train doors fail to close.”

A Robotarium spokesperson said: “The research team will now focus on the production of a flexible navigation tool to guide the robot, and a waste detection algorithm. We hope to have a working prototype by summer 2022.”

Dr Erden added: “For such a research phase system to get into the market and to be deployed in an actual business setting, one needs much more things than technical functionality, such as extensive tests and verification.

“And approvals from many different places, for a company that will pick it up for manufacturing and mass production.

“These take a lot of time, I estimate five to 10 years.

Hide Ad
Hide Ad

“It would technically be possible to physically build and deploy the robot, realistically it would be longer until we might see this little guy sweeping a train in real life due to business processes.”

Comment Guidelines

National World encourages reader discussion on our stories. User feedback, insights and back-and-forth exchanges add a rich layer of context to reporting. Please review our Community Guidelines before commenting.