Pepper the robot at Groupe Créative

Date:
Changed on 25/02/2020
A digital services company founded in Rennes, Groupe Créative is currently modelling its premises in 3D, and visitors will soon be able to take a virtual tour. What makes this project different is that it’s a walking robot that’s been tasked with image acquisition. What’s more, it’s doing this autonomously. Launched in collaboration with Inria, this project is in line with the company’s push towards innovation.
A_Pepper_Inria-0279-463_1826x1027
© Inria / Photo C. Morel

Turnover: €20m. Employees: 340. Branches in France: 7, with more to come overseas. Over the past eleven years, Groupe Créative has become a major service provider, supporting its clients in the process of digital transformation in fields ranging from industrial applications to embedded systems. The company carries out work on behalf of SMEs, as well as groups such as Thales, Canon and BNP Paribas. Groupe Créative is currently collaborating with Rainbow, the research team specialising in robotic vision at the Inria Rennes - Bretagne-Atlantique Research Centre.

This joint project is called Explore. “The aim of this project is to model the insides of our offices in order to enable visitors to take a virtual tour using virtual reality goggles”, explains Antoine Lehuger, head of Creative Minds, the company’s innovation unit. “In order to do this, we use a robot called Pepper, which is equipped with a camera. Operating autonomously, it explores all of the rooms within our branches, capturing images which are then used to recreate 3D models. There are two aspects to this work: exploration and acquisition; and reconstruction.”

Located in Rennes, Rainbow is an Inria's project-team common to Irisa (UMR6074). It is affiliated with the Rennes 1 University, Insa Rennes and the CNRS.

 

Visual navigation

When it comes to exploration, the company draws on the expertise of the Rainbow team in the field of robotic navigation based exclusively on images, i.e. without any prior knowledge of the environment in question. “When you put the robot in a corner, it won’t know what size room it’s in, or where the walls or the corridors are. It has to discover everything on its own, without help from any maps. Using the intelligence we give it, it moves around based on feedback on what it has or hasn't seen, and is thus gradually able to build a map of the premises.” In order to do this, the robot also has to be given a tropism: “ideally, it should gravitate towards as-yet unknown areas, which could potentially contain things to discover.”

The field of smart, autonomous exploration has come up against a number of scientific obstacles. “The literature is not yet particularly extensive. There are no off-the-shelf tools.” In order to work on this subject with the research team, the company is funding a PhD within the context of the Cifre initiative (2). The person studying for this PhD, Benoît Antoniotti (3), is being jointly supervised by Antoine Lehuger and the researchers Eric Marchand and François Chaumette.

3D reconstruction

The second aspect involves 3D reconstruction. “A lot of research has been carried out in this field, and there are now methods available to us which we can use. However, given the practical context we’re operating in, this isn’t as straightforward as it might seem. You need to reach an acceptable level of image quality for the human eye when using virtual reality goggles. Our reconstructions work well for small objects, but aren’t so good at portraying larger spaces. Our biggest challenge involves finding a way of enabling the software to carry out fully automatic self-assessments on the quality of what it produces. It must be capable of detecting any errors in the shots, eliminating these imperfections and retaining only reliable information.”

For this aspect, the company recruited a young PhD. “Salah Eddine Kabbour recently had his viva for his PhD. We had taken him on a few months before. Another of our employees, an expert in C++, took responsibility for optimising the algorithms in Orb-Slam, which is opensource 3D reconstruction software. We reworked the code in order to be able to market it. We will soon be publishing this improved version on the opensource platform GitLab.”

R&D 2.0

Contrary to appearances, this project is not a conventional R&D project: the aim is not to get a product or a service ready for market. “Our approach is completely different. We launched an innovation unit in 2015, which I was tasked with heading. The goal of this unit is to coordinate the upskilling of both the company and its staff. What we had realised was that, between assignments, we sometimes had employees who were available for one or two months. We felt it would be a waste not to use this downtime in order to enable them to learn new technology, new software, etc. At times, we had close to twenty individuals available, and so we took the decision to coordinate this time, devising innovative projects in which these employees could participate in order to acquire new skills. I was given the task of coming up with subjects to match the wide range of skillsets within the company.”

Another interesting fact is that “this initiative is entirely self-funded.” But does that not mean the company is losing money? “No - in fact, we’ve noticed a return on our investment. We’re reducing staff turnover, while building loyalty among employees. As far as they are concerned, receiving training in cutting-edge technology is also an advantage with regard to clients. Furthermore, this unit strengthens our hand when it comes to consulting, helping us to stand out from other service providers.” “For example, say that one of our clients is looking for an automated identification solution for their herd of cattle. They will want to know what different methods are available to them, but most service providers will be unable to give them an answer. It’s too scientific.”

Moving towards collaborative projects

In addition to its work in robotics and 3D reconstruction, the Creative Minds unit is getting ready to explore other themes. Next on the horizon? “Blockchain”. To do a spot of data mining? “No, nothing like that. We want to develop our own internal cryptocurrency. Called Creative Coin, this will be an allowance given to our employees, which they will then be able to exchange for things, such as admission to trade shows, for example. Within the blockchain, there are concepts which may, eventually, be of interest to our clients. I’m thinking in particular of what are called smart contracts or proof of authority”. Here too, Antoine Lehuger is looking into the possibility of bringing a laboratory into the loop. “From experience, we’ve found combining our industrial expertise with that of the academic world to be effective in certain areas. Continuing on with this, we will now be looking to join public research laboratories on EU projects.”

 

Who is the Pepper robot ?

20 degrees

for more movement

15 languages

available to exchange with Pepper

1m20

height to perceive its environment

2D/3D cameras

infrared sensors, bumpers, inertial center, and sonars...

3000

companies use Pepper

2015

Pepper became a companion robot this year

7000

individuals bought a Pepper

2000

EUR by buying