France’s humanoid robot Romeo – Big but not dangerous

Project Romeo is an ambitious undertaking by a coalition of companies and national labs in France to develop by the end of 2011 a functional prototype of a humanoid robot that can assist the elderly and visually-challenged people at home. The goal is to develop a 1.2-1.5 meter (47-59 inches) high bipedal robot that humans can communicate with by voice and gesture and which can help a person get up in case of a fall.

 

The organizing company in this project is Aldebaran Robotics, which develops and sells the smaller humanoid robot Nao. GetRobo talked about the goal and challenges of this 10 million Euro (\1.3 billion, $14 million) project with Rodolphe Gelin, Head of Cooperative Projects at Aldebaran, who is the leader of this project. The following is an edited version of the phone interview and first appeared on Robot Watch in Japanese. (Photo: Rodolphe Gelin with Nao)

 

First of all, can you give us a brief overview of Project Romeo?

 

The idea of Project Romeo was born in March 2008 and the project actually started in January 2009. The goal of the project is to develop a functional prototype of a personal assistant humanoid robot by the end of 2011. The robot is meant to be helpful for elderly people that may have cognitive problems. Also we will be working closely with the Vision Institute (Institut de la Vision) and it’s patients so that Romeo will be able to assist visually-challenged people.

Romeo will be able to assist these people when they are alone at home. It will be able to fetch objects in another room and also assist a person to walk or get up from a chair.

The project is led by the French Cap Digital, which is a coalition of companies, labs and institutions in Paris and the surrounding Ile-de-France region, set up to cooperate in leading innovation and building a competitive industry. There are 13 partners participating in Project Romeo - 5 companies, 7 national labs and the Vision Institute.

 

The Ile-de-France region, the city of Paris and the DGCIS, which is a department within the Ministry of Economy, will fund about half of the 10 million Euros. The rest will be provided by the partners in the form of investment and human capital .

We plan to come out with our first prototype by the end of 2010. Right now we are in the specification phase. Four engineers are currently working with me at Aldebaran and I would say one at each of our partners.

How did this project come about?

 

Bruno Maisonnier, founder and CEO of Aldebaran, had the idea that France can be a leading country in robotics. He wanted to build a human-sized robot as a demonstration of what France is capable to do. And to accomplish that he thought it was very important to bring together people in robotics from different organizations. So he teamed up with Cap Digital to come up with this huge project and was able to secure funding from the public sector.

 

We think that combining the experience and technology that Aldebaran has accumulated to develop Nao with the expertise of the other partners, we will be able to accomplish our goal.

 

By the way, “Romeo” does not mean anything. It sounds like a person’s name and makes the robot more lovable. Names that end in “-eo” are popular in France these days. It sounds new and high-tech.

 

How did you get involved in the project?

 

To give a little background about myself, after receiving a masters degree in artificial intelligence at University Pierre and Marie Curie, Paris, I joined the CEA, which is a big research institute in the atomic energy field in France. I worked in the robotics unit there. Originally we worked on robots for the nuclear industry, robots that do work where there is too much radiation.

 

Then I started working on robots for disabled people. This is because one of CEA’s missions is to apply the technologies that were developed for nuclear use to other applications. The problem for severely disabled people and the problem for working in a nuclear environment is similar because in both cases you are unable to directly manipulate objects. So it became an important part of our lab to apply the technology we initially developed for use in the nuclear field to rehabilitation robotics. I was involved in this project for 15 years while gradually moving into management. Before I left CEA a year ago I was in charge of the unit that worked on robotics, virtual reality and cognitive software which consisted of about 100 people.

 

Meanwhile I was getting more interested in getting actual products out into the market. It is a major challenge these days in robotics to be able to move the robot from a prototype into a product. I had the chance to meet Bruno (founder of Aldebaran) about 3-4 years ago and told him I was interested in going out into the industry. When Project Romeo started he asked me to lead this project. Thus I joined Aldebaran in Dec. 2008. Currently Aldebaran is about 67 people, mostly engineers.

 

There are 13 organizations listed on your partners list. What are the roles of each of these partners?

 

Starting with the companies, Aldebaran is leading the project. We are developing the mechanical and electrical parts with the assistance of partners as well as the middleware and localization and navigation software.

 

Voxler is an expert in sound processing. They will work on the musical interaction aspect of the robot. Their main business is in developing games for the Nintendo DS and they have a technology that enables you to sing a song to a robot and the robot can respond to you also by singing or adding instruments to your voice. We think this is a new and good way for people to interact with a robot because it is still very difficult to communicate with a robot with speech. People are always disappointed by the limited dialogue and get bored with the robot easily. Of course, we will work on the dialogue but if you rely solely on dialogue you will always be disappointed.

 

We feel that Romeo must not only be a useful physical object inside the home but also something of a companion that you can interact with and have a relationship with. And we think that musical interaction will help.

 

Besides these two companies, there is the Acapela Group (text-to-speech software and speech recognition technology), SpirOps ( AI engine for decision making) and  AsAnAngel (dialogue specialist).

 

Moving on to the labs, below is a list of what each lab is working on for Project Romeo.

- LPPA : Physiology of perception and action. Fundamental research on human physiology.

- CEA LIST : Object recognition, tactile surfaces (for robot’s skin), mechanical design (spinal cord, limbs)

- LAAS : Path planning and motion control (bipedal walking)

- INRIA : Dynamic control (bipedal walking)

- LISV : Mechanical design, optimization of design

- LIMSI : Detection of emotion in voice

- Telecom Paris Tech : Sound signal processing

- Institut de la Vision : They will bring us the needs of the potential users of Romeo and the test bed for final evaluations.

 

The type of robot you are trying to develop is - from our point of view- very ambitious. There are multiple technical challenges that must be overcome and then all the technologies must be integrated. What do you think are the top technical challenges for Romeo to succeed?

 

A.   I agree. It is ambitious! And I believe there are 3 big technical challenges that we must overcome.

 

One is the mechatronics – how do we come up with a design that enables a big robot to be useful in a human environment but not dangerous. From the battery to the intensity of the currents in the wires, everything will be huge. And because of that, the robot could be dangerous. Force control of this robot will be very very important.

 

Also the robot must be light. We are aiming the robot’s weight at 30kg. If you need 30 motors, that’ll add up to 30kg with just the motors. This is a critical problem and we may have to develop new motors. Maybe for our second prototype we will have developed our own motors.

 

Second technical challenge is enabling the robot to move dynamically and smoothly in an everyday life environment. The walking part – how to control the joints and gait - is crucial. We are working on a new type of walking algorithms for the robot. The typical walking algorithm in current humanoid robots is the ZMP algorithm but we are attempting a new algorithm that does not use ZMP and that will enable the robot to walk faster in a dynamic and more natural way. The principle of ZMP is that first you choose where you will put your feet and then you try to bring the robot’s body to where the feet have to be. But this is really not the way humans walk. Your feet are supposed to follow where you want to go.

 

Aldebaran has accumulated knowhow in bipedal walking by developing Nao. Of course it will be more difficult with Romeo because it is taller and heavier but we hope that it will be easier because Romeo won’t have the price constraints that Nao has and we will be able to invest more money into better parts.

 

Last but not least, the third technical challenge is the man-machine interface. How to enable intuitive communication with the robot using speech and gestures? I have already told you about the musical interaction but we also want the robot to recognize human gestures. Right now we have a very simple gesture vocabulary in which there are 6 gestures that the robot has to recognize, which include pointing (give me that), ‘stop’, ‘don’t’, and ‘come here.’ We are developing a technology that will enable the robot to recognize these hand gestures in a very efficient and robust way. And we want to have this feature connect with the voice too.

 

According to your website, Romeo will "be able to help when the person would fall on the ground." In what way will the robot do this?

 

That is a major challenge, maybe the biggest, in terms of tasks. Our idea is to make the robot into something like a mobile and reconfigurable piece of furniture. If you look at the homes of elderly people, there is furniture everywhere. It’s not only because they’ve accumulated many things over the years but because it’s a way for them to be protected by objects. They can put their hands on a chair or table to avoid falls or to get up after a fall. That’s the kind of assistance we want to bring with Romeo. When you fall it can come close to you and it will offer its knee, then its arm, then its back as a support so that it can gradually pull you up higher.

 

We do not want Romeo to carry the person like the Japanese robot RIBA does. The RIBA is very impressive but it is not reasonable. For the robot to be able to carry 60kg it needs power and that could be dangerous. I don’t think people are ready to have that kind of robot inside their homes.

 

How will your robot compare with the existing humanoids?

 

Many of the humanoid robots in the world are technology platforms. We want to make Romeo into a service humanoid robot that is actually useful for elderly people. Project Romeo is led by Aldebaran, a company that has already sold more than 300 humanoid robots.  It means that Romeo is the young brother of a commercial product. We have feedbacks from customers in real life. And we can reasonably think that Romeo will become a product in 3 to 4 years after the end of the project.

 

Why do you think that this kind of assistive robot must be a humanoid?

 

The robot will firstly be a companion. It must be accepted by the users. The human-like shape is necessary for the robot to be easily accepted by people. You know where the head is and the arms, etc.

Furthermore our home environment is designed for humans. For the robot to be useful in our environment, it will be more efficient having a human shape. This point explains why we want Romeo to have two legs. Wheels are so much easier to control but homes are not designed for wheels. Ask the people who use wheelchairs.

 

Will companies outside of France be able to contribute to your project?

 

We will utilize existing technologies of our partners and of our products. But it could happen if we find subsystems available outside of France such as motors and sensors.

 

Once Aldebaran commercializes a product based on Romeo, how much do you think people would be willing to pay for it?

 

Difficult question. We think that people will not pay for the robot by themselves. Insurance companies could finance the robot to help people remain living at home longer.