Reem says hello to Yas Island

If there was a Yas Island competition for employee of the month, it is likely that last month's prize would have been awarded to a multilingual expat from Barcelona called Reem.

Standing 1.7 metres tall and weighing a robust 100 kilograms, Reem not only worked eight hours every day for the whole month, but also performed as a children's entertainer and a source of tourist information.

She led two promotional campaigns and an island-wide treasure hunt and even starred in her own short film, which received more than 150,000 views on Facebook.

Reem is a robot, and as she toured the food court linking Yas Mall with Ferrari World on her last day of service, it was clear that, as a marketing tool, her eight-hour batteries give her the power to turn heads.

"It is very interesting to have a machine interacting with you,” says Prof Klaus Weber, a German academic who was one of the few visitors to walk straight up to the robot and use its chest-mounted touchscreen to request a shake of the hand.

"I've seen something like this before on television. Our chancellor, Mrs Merkel, recently opened an exhibition in Japan that had robots like this, but I've never seen anything like this before in person.”

As shoppers posed for selfies with the robot and young children fidgeted nervously, they failed to notice the robot's "handler”, Jordan Palacios, who watched from a nearby table.

A software engineer with the Barcelona-based Pal Robotics, the Abu Dhabi-owned and funded company responsible for Reem's programming, design and manufacture, Mr Palacios is charged with keeping an eye on the Dh620,000 machine and using a touchscreen remote control to guide its movements.

"This is the first time I've worked in a live ‘demo' environment, which is wonderful for me because it's a validation of my work,” the Catalan says.

"I'm a programmer, so every day, I work in front of a screen writing lots of code. So it's rare for me to see the product in action but you can really see people interact with the robot here and, for me, that's great.”

Reem is a semi-autonomous robot, which means she does some things for herself but needs assistance with others – especially in an environment as busy, loud and confusing as a mall.

"It can navigate and if you have a map, you can also point it in that direction and the robot will find a location by itself, but for other things, like when the robot interacts with people, the operator can assist,” Mr Palacios says.

"In a quieter area, the robot would be able to hear very clearly and would be able to respond with pre-programmed responses.

"In this kind of environment, where faces and voices can get mixed, the robot may not understand exactly who is giving a command, but as the technology advances and when we are able to get the speech recognition right, the less help the robot will require.”

Before Reem went to Yas Island, the challenge of enhancing her face and speech recognition was presented to a team at the Khalifa University Robotics Institute (Kuri), who were given access to the robot for more than a week as part of a joint project with Pal Robotics.

"We developed a programme to allow the robot to concentrate on one person at events, such as exhibitions, when a lot of people would be talking,” says postgraduate student Randa Almadhoun, 22, one of the Kuri students who worked with Reem.

"It depends on localising a person's voice and face so if the person talking to the robot ends their conversation and leaves, the programme would allow the robot to recognise and start a conversation with somebody else.”

Like many of the projects under way at Kuri, the "audiovisual interaction technology” developed for Reem is still a work in progress.

But the idea is to produce a program that can be used by Pal Robotics and by Kuri for research to be used in its own robots.

The tangible evidence of those efforts can be seen in Kuri's lab at the university's Abu Dhabi campus. It is filled with whiteboards covered in handwritten code, industrial-looking robotic arms, wheeled vehicles that look like bomb-disposal devices and smaller white and orange figurines that appear to be little more than toys.

"From an intelligence point of view, they are like toys,” says Tarek Taha, one of Kuri's post-doctoral research fellows.

"There's nothing intelligent about them other than their sensing abilities, but the robots are very good at providing therapy to children with autism because they are less stimulating than a human might be.”

Despite their size and their use of a combination of voice commands and tablet computers, the robots display a quality that might almost be described as patience.

"The robot needs an awareness that there is a human in front of it to be able to respond in a convincing manner and to use behaviour that is actually part of the interaction scenario,” Mr Taha says.

"It also has to keep on monitoring the child to understand what step they are at during a learning session, evaluate whether they have finished the section correctly, repeat the exercise if they haven't or continue if they have.”

Although it is not Kuri's only research interest, the issues surrounding human-robot interaction are at the heart of its most innovative research projects.

One is what Mr Taha describes as an exercise in "cloud robotics”, a Kuri project that would enable robots to care for the elderly in their own homes.

"The robots would assist elders, not just by performing pre-programmed tasks, but through learning new behaviours that don't already exist inside the robot,” the academic says.

"For example, if a robot is trying to cook something or perform a cleaning task, it can simply scan the web or a database of collective cloud knowledge, then come back and say, ‘OK, I didn't know how to do that before but I do now'.”

As with the solution for Pal's Reem, the most profound obstacles facing anyone who wants to create a robot capable of acting independently, intelligently and of assessing their situation and sharing information are intangible.

They depend as much on an understanding of human behaviour as they do on the creation of machines.

"People underestimate the complexity of artificial intelligence and robotics software, but that's the core of robotics these days,” says Mr Taha.

"The foundation that you have to learn is that to actually achieve anything in robotics is far beyond the mechanical or the creation of a physical robot. Moving something around is easy.”

Hend Al Tair's "Collaboration Engine” is a case in a point. Her engine will consist of little more than a series of algorithms and protocols when it is finished, but if it is a success the results will allow humans and robots to work together.

"I describe it as an engine or a framework for collaboration itself,” says Ms Al Tair, 29, a PhD student from Ras Al Khaimah.

"How can I measure and enhance the collaboration between the team members, human and robot, and how can this work be enhanced during a mission?

"Humans can obviously talk to one another or look at something and then know what they have to do, but how will the robots do that.

"Let's say robots are required to look for a radiation source. They should be able to look around, understand when they have or haven't found the radiation, and then share that information with the rest of the team.

"In this scenario, the action the robot takes affects its behaviour at a local scale, but it also affects the behaviour of the group.”

If the implications of Ms Al Tair's research sounds like the stuff of science fiction, Mr Taha insists that it rests on techniques and lessons such as reinforcement learning.

"Reinforcement learning is similar to the way that children learn with negative and positive reinforcement. In the context of robotics, that means designing reward functions that guide behaviour and the idea is to define problems in a way that, if the robot takes multiple actions, will allow it to maximise its reward at the end.”

Despite their talk of programs and algorithms, procedures and protocols, it is the researcher's constant return to the subject of human interaction and learning, cooperation and understanding that is most striking.

Ultimately, the factors that may define a robot's success are the same as those that make us human.

But for Mr Taha, meaningful progress will require artificial intelligence that allows robots to learn without pre-programming or human intervention, and a change in human behaviour.

"If you are looking at human-robot interaction, the expectation normally outstrips the reality. People expect them to have the same intelligence as a human but that is not correct.

"Then they adjust their expectations and they can sometimes show their frustration. They end up thinking they're dealing with a bumbling toy.

"But I think roboticists also overstate the progress of their research. They make lots of claims and promises, but we are still not advancing as fast as we would like.

"I think we are still very far away from achieving AI.”