We have all seen robots attempt to move over steps and separate platforms. It is easier said than done. Usually, cameras or LiDAR are used for these tasks but they can be affected by occlusions and computational overhead. Researchers have come up with a new approach to handle this. They have used proximity sensors at the bottom of a quadruped’s feet to allow autonomous locomotion on discrete terrain.
More like this ➡️ here
Learning Locomotion on Discrete Terrain via Minimal Proximity Sensing
As the researchers show, sparse, near-field sensing can be modeled in simulation and transferred to the real world with high fidelity. The above video shows how it all works.
[HT]
*Our articles may contain aff links. As an Amazon Associate we earn from qualifying purchases. Please read our disclaimer on how we fund this site.
