Soft Robotic Fingers and Vision Guidance Based on Deep Learning

This PhD thesis mainly includes two parts. In Part 1, we firstly propose a Venus Flytrap inspired bistable soft finger to enhance the output force for grasping. Secondly, we present a dual-actuation mechanism further to improve the performance of the bistable soft finger. Thirdly, we create an antagonistic constraint mechanism to change the effective bending length and then the bending profiles of the soft robotic finger. In Part 2, we propose vision-based sensing via deep learning to guide the robotic grasping in the complex agricultural environment.