World's first robotic personal trainer guides gym-goers through exercise programme

Researchers at Bristol Robotics Laboratory have created the world's first socially intelligent robot fitness coach.

The robot companion can give motivational voice prompts and make physical gestures to guide its pupils through an exercise regime on a treadmill. The support is personalised based on the runner's heartrate, speed, personality type, mood and fitness level.

Humanoid robot Pepper can tell jokes, show sympathy, lean in towards the runner and even change its eye colour to express different emotions.

Pepper gained its social intelligence by being 'trained' by a fitness instructor on how to be an encouraging coach. It learned when to offer praise during a workout and which cues can successfully motivate a runner to improve their performance.

The robot was developed as part of a study led by human-robot interaction experts Katie Winkle and Séverin Lemaignan, in collaboration with roboticists Paul Bremner and Praminda Caleb-Solly, psychologist Ute Leonards and rehabilitation expert Ailie Turton. Over the course of three months, 10 people participated in sessions with the robot three times a week in a gym at the University of the West of England (UWE Bristol). The robot guided them through the NHS Couch to 5k programme, designed to help beginners build up to completing a distance of 5km.

Katie said: "We wanted to test if we could transfer the intelligence of our fitness instructor, an expert with the know-how to get the best out of clients, into a robot so it could become an effective personal coach. We aimed to create an engaging motivational companion to get our runners through Couch to 5k, which is quite a long and drawn out exercise programme."

Participants received guidance from two different versions of the robot. The first was pre-programmed and gave standard instructions every 30 seconds based on the runner's performance. The second was a bespoke form of the robot that had benefitted from expert input from a UWE Bristol fitness instructor.

Results showed runners preferred engaging with the socially intelligent 'human-like' robot, and performed better with this type of companion.

Katie said: "The socially intelligent robot knew nothing to begin with but learned 'on-the-go' by observing how our fitness instructor was motivating our runners. It then replicated his behaviours until the point where it could operate intelligently by itself.

"We successfully created the first assistive robot able to learn complex social behaviours that take into account unique personalities. With the initial help of a human 'teacher', the robot developed its own unique way of interacting with its human partners, guiding them through tasks, but also appropriately challenging, encouraging or praising them, as humans would do."

Participants in the study ranged in age between their 20s and 60s. After initial scepticism, they all responded positively to the robot personal trainer.

Katie said: "Understandably, they weren't sure what to expect and when they started it was a bit of novelty. Some were energised by the robot and pushed themselves more than normal, and our fitness instructor was impressed with what the robot had learned to do.

"As time wore on, the participants began to treat the robot as a companion and the fitness instructor saw the robot as a colleague. This is really promising when we think about how robots might be used in the workplace in the future to work alongside humans."

The study was the first of its kind to feature significant participation from a human domain expert, with the UWE Bristol fitness instructor involved in designing the robot actions, teaching it and then assisting with evaluation. It is also rare for a research project to involve people interacting with a robot consistently over such a long time period.

Socially intelligent robots such as the one tested in the study could eventually be used in the fitness industry and other sectors including healthcare and education.

Dr Séverin Lemaignan, a Senior Research Fellow at the BRL, said: "Our work shows a robot could be really useful in the gym, especially for people who would perhaps feel embarrassed with a human personal trainer. It's great to see the robot was considered a colleague by our gym coach: the robot helps provide better support to the gym clients; he doesn't see it as a threat to his role.

"More widely, the study demonstrated the potential for robots to be useful and effective in the real world, with many potential applications. If you can create robots that are socially intelligent in this way, that know how to encourage you and when, we could be seeing them in health and education. I'm really excited to see what contribution we can make and where we can take it next."