Creating a Robotic Surrogate for Co-evolving Behaviour and Design in the Real World

Basic Information

Students: Meri Mäkelä, Zhibin Yan, Elias Haarnoja, Kalle Lindgren

Project manager: Meri Mäkelä

Instructor: Kevin Luck

Starting date: 1.2.2023

Completion date: 14.6.2023

The purpose of this project was to create a robot platform for RL research by combining a RealAnt robot with the co-adaptation software. This was done by creating software, a Python Gymnasium environment with an API that could connect to the RealAnt robot. In addition, OptiTrack system was used to provide tracking data of the robot. The data from the system was gathered by using ROS.

The project also involved CAD-design and 3D printing. The goal was to create a parameterized 3D CAD-design of the robot's leg component. The parameters will be given by the RL algorithm, thus realizing the co-adaptation of behavior and morphology of the robot. 

Overall, the project was a success: by the end of the course, the platform was fully functional and ready for more testing.