Authors:

Guo Yu

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

One major challenge robots face is their lack of tactile perception. This limitation increases the risk of breaking fragile objects and makes it difficult for robots to detect slipping movements and adjust their gripping force. To alleviate this limitation, this research aims to develop a vision-based haptic sensor that can provide robots with valuable information about the objects they hold. The initial phase of research is simulating the interaction between a coherent light source and the surface deformation of a box. The shape of the deformation was parameterized using sinusoidal basis functions, and the intensity difference opposite of the light source was collected. This data was used to train a machine learning (ML) model to predict the basis parameterization given an intensity change. Simulation results indicate that it is feasible to extract the shape information from the intensity change. A physical prototype was designed to mimic the simulated system. Laser light was reflected within a sensing cavity and the intensity pattern was captured as an image using a standard CMOS camera. The sensing cavity is composed of a deformable gel (elastomer) surrounded by a reflective film. Utilizing photometric stereo, training samples for various surface deformations can be collected, and a ML algorithm can be trained to predict the surface shape based on change in intensity pattern. This research tackles the challenge of tactile sensing in robotics by designing a vision-based haptic sensor. Further improvements include making the sensor small enough to be mounted on the fingertip of a robot.

Source:

Purdue University / 2023

Topics:

No topics listed

Co-authors:

Guo Yu