The electronic biped

The simple dream of a robot that vacuums your house took a long time to realise. In contrast, the reality of a multifunctional humanoid robot – capable of offering companionship – still seemed a long way off in 2005 when Bruno Maisonnier established Aldebaran Robotics, putting France on the map as a world leader in the field. “Nobody thought it was possible to manufacture a biped robot and sell it in great quantities”, recalls Rodolphe Gelin, the company’s chief scientific officer. The company reports total sales of nearly 20,000 robots since 2008.

Robots were ubiquitous at this year’s Consumer Electronics Show in Las Vegas. Aldebaran’s newest offering, Pepper, proved popular, but with tech investors backing 128 new robotics start-ups to the tune of $1.95 billion in 2016, popularity must be backed by sales.

Laying the foundations for mass market humanoid robots has not been simple for the company, despite some early wins. In 2006, Maisonnier developed NAO, his first offering. By 2008, NAO was named the standard platform for the RoboCup League, in which software developers program competing teams of footballing robots. The company seemed to be on a roll in 2010, when 20 robots were invited to dance in the France Pavilion at the Shanghai World Expo. The University of Tokyo purchased 30 NAOs for their Nakamura Lab, with hopes of developing them into active assistants. So far, so good – but the rapid momentum of those early years has slowed.

Finding its feet in education

Even though NAO sales have been disappointing, it remains a favourite among researchers and educators. By the end of 2014, more than 5,000 NAO robots were in use in 70 countries. “Though the general consumer market is not yet mature, the robot serves as a research platform”, Gelin explains. “It is a learning system for STEM and programming learning activities, support for application creation and even a tool for educating children with special needs.”

As head of the Computer-Human Interaction in Learning and Instruction (CHILI) Lab at the École Polytechnique Fédérale de Lausanne, Pierre Dillenbourg is one of the most prominent researchers exploring the potential of NAO robots for education. “Creating smooth human-like interactions is interesting, but the real challenge is to find activities in which robots can add value to create powerful learning experiences”, he explains.

The CHILI researchers have demonstrated that NAO can be used to exploit a pedagogical phenomenon known as “the protégé effect”. Through the CoWriter project, the team ran a trial with six- and seven-year-olds with literacy problems. Tasked with teaching the robot to write, the young students remained highly engaged over the four-week period, and succeeded in significantly improving the robot’s handwriting.

SoftBank’s Pepper

In December 2016, a pilot run of Pepper was conducted at B8ta, a tech retail shop in Santa Monica (USA). The store reported a 13% increase in revenue.

Shifting sands

With NAO proving popular in academic settings, but not entirely useful for the home (and still expensive at around €6,000), Aldebaran found itself at a crossroads. In 2013, the company was acquired by Japanese-owned SoftBank for $100 million and later rebranded as SoftBank Robotics. Maisonnier left and the goal, with NAO at least, pivoted from developing for a mass market to focusing specifically on academic institutions.

How close do experts believe companies like SoftBank Robotics are to developing a fully-formed humanoid? “If you’re talking about a robot that can perform the role of a real counterpart, I think it will take another 15 years before they can fully interact with us reliably”, says Appolinaire Etoundi, a senior lecturer in mechatronics at the Bristol Robotics Laboratory. He offers plenty of optimism alongside objectivity. “From shaking hands to pouring you a glass of juice or holding a basic conversation, these recent advances are huge milestones in robot-human interactions”, he says.

Is NAO the future?

SoftBank Robotics’ hopes are now firmly pinned to Pepper. NAO’s educational focus makes more sense if Pepper is stepping into the consumer and corporate space. This decision is supported by sales: 10,000 Pepper units sold worldwide since June 2015.

Pepper’s key feature is that it’s likeable, which is high praise in the world of robotics. In retail settings, Pepper is still a novelty. SoftBank conducted a pilot run of Pepper at B8ta, a tech retail shop in Santa Monica, in December and the store reported a 13% increase in revenue. Getting devices like these shipped is more than half the battle. As with Tesla’s cars, developers can apply additional intelligence at any time through remote updates; the key is to first find useful tasks the machines can comfortably handle while they’re still in a relatively “dumb” mode. On that front, SoftBank Robotics has a decade’s head start over some of its newer competitors. “Originally our company was mainly driven by the techno-push approach: we love technology and it was fun to develop a humanoid robot”, Gelin explains. “But we now have 20,000 robots ‘in the wild’, each providing an opportunity to observe people’s interactions with them.”

SoftBank won’t have to shoulder sole responsibility for innovating with Pepper. If the robot captures the imaginations of external developers worldwide, they will also help push the device’s boundaries. “The openness of the operating system is one of the most interesting strengths”, Dillenbourg says, referring back to his experiences with NAO. “We can quickly prototype new activities; if we need to track a learner’s eye movements or gestures, we can do it.”

So what is Gelin most proud of to date? “From a technical perspective, NAO’s bipedal walk and its ability to get up again if it falls are probably the features that best show off SoftBank’s humanoid robotics expertise”, he says. In an increasingly competitive market place, SoftBank Robotics will need to prove itself similarly capable of responding to knocks and setbacks to maintain its success.