• DMP
• mobile robot
• topologial maps
• without coding

Today robotics is one of the so-called exponential technologies. While the first robots were only used in factories to carry out repetitive, dangerous or dirty jobs, in protective cages, today they are increasingly used in hospitals, domestic sectors etc. and interact with people. Widespread use in various fields and among non-expert users requires accessible and intuitive robotics. In this direction, Learning from Demostration is a promising method as it transfers skills from humans to robots. It is a very developed method in various applications such as those that require the repetition of paths such as robots that have to carry out inspections, or those that guide inside museums or those that collect data. It will describe an intuitive method for transferring skills from human to robot and which allows even the beginner to use robots “without writing code”, after a single demonstration. The robot, in fact, will be teleoperated along the trajectory to be learned and then will be able to perform it autonomously.
The robot will also be able to build a decision tree, not containing redundant information, and will be able to learn more trajectories in different scenarios and re-execute them adequately and autonomously when required. The trajectories learned will be generated through the DMPs so that even if we change the starting point with respect to the demonstration, convergence towards the desired goal is guaranteed.
This system was tested in simulation and on the Robotnik SUMMIT-XL STEEL at the CROSSLAB in Navacchio.