Until someone invents a robot that can moonwalk, you can model a gait that appears normal by simply making sure that the velocity of a foot is zero as long as it’s in contact with the ground. The Aerial-Biped robot learns how to do this through reinforcement learning in a physics simulator, and the policy transfers to the robot well enough that the legs can appear to walk as the quadrotor moves.
Right now, getting this to work on the real robot involves using motion capture on the drone, so it’s not yet suitable for out-of-lab wandering. The researchers are working on adding more degrees of freedom to both the body and the legs, with the goal of being able to physically imitate the gaits of animated characters.
For a bit more detail on this project, we spoke with lead author Azumi Maekawa from the University of Tokyo.
Where did you get the idea for this research?
We were inspired by bipedal robots that use invisible force to get stability, such as Magdan, created by Tomotaka Takahashi (an electromagnet on the bottom of its feet lets it walk on a metal plate), and BALLU (which uses buoyancy of a helium-filled balloon). The foot trajectory generation method is based on the assumption that one of the key features of walking (or at least the appearance of walking) is that the velocity of the foot in contact with the ground is zero.
What function do the legs serve on this robot? Or, what is the goal of adding legs to a quadrotor?
The goal is to develop a robot that has the ability to display the appearance of bipedal walking with dynamic mobility, and to provide a new visual experience. The robot enables walking motion with very slender legs like those of a flamingo without impairing dynamic mobility. This approach enables casual users to choreograph biped robot walking without expertise. In addition, it is much cheaper compared to a conventional bipedal walking robot.
Are there practical applications for a robot like this? What are you working on next?
Although it is at a prototype stage now, in the future, an entertainment application such as performance or animatronics can be considered. We aim to develop a biped robot that has the ability to display desired motions, including various dances, in addition to walking. I think this work has the potential to make virtual reality experiences possible in the physical world by enabling movements that have been impossible due to the constraints of the mechanisms and properties of real-world characters.