Researchers from ETH Zurich in Switzerland have managed to successfully use a drone to open doors. Technically speaking, the team used a Reinforcement Learning (RL) approach to learn reactive motion behaviors for a manipulation task while producing policies that are robust to disturbances and modeling errors.
In other words, they train a policy to perform a door-opening task with an Omnidirectional Micro Aerial Vehicle. This policy is trained in a physics simulator and then showcased in the real world, where it is able to generalize door closing tasks never before seen in training. They also compared this new method against a state-of-the-art Model Predictive Path Integral (MPPI) solution in simulation, which showed a considerable increase in robustness and speed.
No products found.
