Fully customizable robot to develop new algorithms

The Robot Learning & Interaction research group at Idiap has recently acquired a new tool. A small quadruped robot named SOLO12, whose mission will be to allow the researchers and engineers of the institute to develop new robotic applications based on artificial intelligence.

This new robot is not limited to walking. It can also be used to displace objects. It should be known that the recent most advanced robot demonstrations are not reusable by researchers, because they use software developed by private companies. Developed in the framework of the European H2020 robotics project "Memory of Motion - MEMMO", the greatest strength of this open-source robot lies in the fact that it is totally customizable. Whether this is at the level of its mechanical or software components, it is totally modifiable and adaptable. For example, it’s possible to add a camera or to have access to its code and modify it to adapt the robot to the needs of the institute's researchers.

Preparing a new generation of robots

Members of the Idiap robotics group will work in two stages. First, one of the students from Idiap’s Master in Artificial Intelligence will have to implement control models for quadruped robots already available in the scientific corpus.

Once this step has been completed, the objective will be to set up what are called optimization algorithms to improve the adaptation of the robot's movements. This work is a continuation of the European project MEMMO - Memory of Motion. The goal of this project is to create robots capable of adapting to a dynamic environment thanks to a technology using a memory of movements. If the robot is pushed, it will be able to reproduce movements that can help to regain its balance. It will then be able to execute these reactive movements quickly and autonomously, without the need for a user to intervene with a joystick for example.

« Our work with this robot is clearly oriented toward research. The objective is mainly to help researchers with better robotic platforms to test new algorithms. The new generations of robots that will result can then be used by the industry, » Sylvain Calinon, head of the Robot Learning & Interaction research group, explains.

More information
- Robot Learning & Interaction Group
- European project H2020 MEMMO - Memory of Motion