Improving the Force Reconstruction Performance of Vision-Based Tactile Sensors by Optimizing the Elastic Body

The force feedback of vision-based tactile sensors is always expected to be accurate and robust. This work focuses on indicating and lessening the limitation of the structure of the vision-based tactile sensor's elastic body on the distributed force reconstruction performance. The conflict between the accuracy and the robustness of the reconstructed force and the limitation of the elastic body's structure on the accuracy and robustness are discussed based on a general force measurement model of vision-based tactile sensors. Aiming to improve the force reconstruction performance by optimizing the elastic body, we propose a quantitative sensitivity index to evaluate and compare the performance of the elastic body. Under the guidance of the proposed sensitivity index, the methods and principles of optimizing the structure are explored. Finally, the applications of the methods and principles in optimizing the elastic body of two representative vision-based tactile sensors, Tac3D and GelSight, are carried out and significantly improve their force reconstruction performance.