Multiple Digital Twins Competing in Manipulator Tasks

• Piatan Sfair Palar UTFPR
• Jonathan Cerbaro UTFPR
• André Schneider De Oliveira UTFPR
• João Alberto Fabro UTFPR

Resumo

This paper proposes a new system where multiple digital twin models are simulated performing a future task, while the robot is performing the current task. The simulated robots try different algorithms and compete with one another associated with indicators, where the winner is coupled with the real robot to perform the next task. When this occurs, the cycle repeats itself, with the simulated robots then trying the new future task. This is a preliminary work in a series that aims to work with different simulators, robots and time-scales since this is a scarcely explored area. In this step, the proposed system was tested with a Movemaster RV-M2 retrofitted robot with four simulated models where one is coupled with the robot and the other three are competing for the future task. The task chosen was an altered version of the Hanoi Tower puzzle, with four pegs and two towers.

Palavras-chave: Poles and towers, Education, Reinforcement learning, Manipulators, Digital twins, Planning, Task analysis, Digital Twin, Multiple Digital Twins, Hanoi Tower, Simulation