The incorporation of touch as a sensing modality for robots may lead to more reliable manipulation skills. Currently, the tactile sensors available are far from ideal because they are flat or have an exceptionally narrow field of sensitivity. This means they transmit very low-resolution signals.
OmniTact is a multi-directional high-resolution tactile sensor that has been able to overcome these challenges, according to Cornell University. This sensor is designed to be used as a fingertip for robotic manipulation with robotic hands, and uses multiple micro-cameras to detect multi-directional deformations of a gel-based skin.
The micro-cameras transmit a strong signal from which a wide range of different contact state variables can be inferred by using computer vision methods and modern imaging processes.
OmniTact’s capabilities were examined by performing a challenging robotic control task, which involved:
1. Inserting an electrical connector into an outlet
2. A state estimation problem that’s representative of those typically encountered in dexterous robotic manipulation, where the goal is to infer the angle of contact of a curved finger pressing against an object.
Both tasks were performed using only touch sensing and deep convolutional neural networks to process images from the sensor’s cameras.
The researchers compared a state-of-the-art tactile sensor (that’s only sensitive on one side) and a state-of-the-art multi-directional tactile sensor, finding that OmniTact’s combination of high-resolution and multi-directional sensing is crucial for reliably inserting the electrical connector and allows for higher accuracy in the state estimation task.
Videos and supplementary material can be found here.