One major challenge in robotic sensing is building tactile sensors that is thin and have wide tactile sensing areas like human skins. A typical approach is to use deformable materials such as elastomers and use a camera to observe the deformation of the material via photometric stereo. However, as the camera has to be placed at a distance to the elastomer, it is difficult to make the sensor as flat as possible. The aim of this project is to develop a compact, vision-based tactile sensor for robots via folded optics. The research will involve two components as follows:
- Design and build a tactile sensor prototype with appropriate light sources and flat elastomers.
- Make an experimental platform that can properly calibrate the sensor using computer vision techniques.
- Computer vision
- Machine learning
- Computational sensing
Signal processing, machine learning, computer vision.