Machine learning and computer vision for Smart Robots – Arès

The increasing automation of dangerous, tedious and repetitive tasks is a symbol of technological progress at the service of man and bearer of productivity and competitiveness. In France, as in Western Europe, the sustainability of an industrial site is conditioned to its high productivity and the flexibility and scalability of its production tool in a context of globalized market with strong competition and whose needs change rapidly.

The objective of this Labcom project aims to develop new machine learning and computer vision methods to create within 3 years flexible, adaptable, autonomous robotized PICKING / KITTING tools able to guarantee high productivity and manage ultra-flexibility required in many industrial applications. This is to contribute to improving the performance of the task of sorting and placement of products in context of high diversity and high flexibility, a task often very tedious and repetitive that is transverse to all industrial manufacturing processes (foundry, chassis, machining, stamping, sheet metal, assembly), and a problem that we frequently find in companies with significant internal logistic flows such as parcel packaging or any type of manufacturing. The challenge is the management of the diversity of packaging components and the ability to reduce by 100 the adjustment time of robotised Picking/Kitting systems currently parameterized by experts. This also to eliminate, by mean of robotic automation, dangerous or repetitive tasks with high cognitive and physical hardship, allowing operators to be reengaged on other tasks with high added value, ie, the assembly of parts in the automotive sector.

For all the aforementioned goals, the Labcom ARES project combines two complementary partners who will share their capabilities of research, innovation and development and formalize their already strong cooperation in R & D:
- A laboratory: LIRIS (Laboratoire d'InfoRmatique en image and information systems) to develop new methods and techniques of machine learning and computer vision in the service of adaptable and flexible robotics.
- An SME: Siléane, specialized in robotics, who will design and produce autonomous PICKING / KITTING robots.

The Lacom ARES project fits perfectly with the program of "Industry of the Future" supported by the government. After a joint Cifre thesis focused on generic and dynamic engine of vision-based reconnaissance of objects in vrac and a FUI project, PIKAFLEX, involving Renault, Siléane and Liris ECL to develop robotised devracage tools for mechanical items at the service of the automotive industry, this project is another step that strengthens the partnership between the Liris laboratory at ECL and Siléane to conduct R & D work that will endow Siléane's robot lines with learning and vision capabilities, allowing them to pick and kit products from vracs which could be homogeneous or heterogeneous, composed of items of different nature, e.g., food, cosmetic, mechanical. In extending the skills of Siléane's robots to deal with an increasing diversity of vracs further to mechanical items, the research results of the project will be a real growth driver that will enable the SME to industrialize the developed solutions and quickly address the logistics market in Europe. Many jobs will be created as consequence of the project, especially in the Loire department where PICKING / KITTING products will be designed and produced. In addition, the smart Picking/Kitting robots as targeted by this LabCom project will address a key issue of French industry, e.g., automotive, food, allowing in particular to increase their competitiveness.