Files
Abstract
Soft robots are constructed of softer material than rigid robots, which have a higher potential to safely interact with humans, are more adaptive to the target objects, and is flexible to work in confined environments. A lot of invertebrates in nature move or function without born or skeleton. These features are excellent analogs for soft robotic design because it can prevent some unnecessary mistakes if we do not have enough understanding about the motion. In addition, bionic methods can also help robots better navigate real-world environments and perform interactive tasks. In the soft robotic control, the softness property gives the robot high locomotion flexibility, allowing the robot to own multiple degrees of freedom. However, the high flexibility also brings difficulties to motion detections. Since soft robots are susceptible to gravity and their bodies are easily deformed by the environment, traditional pressure and resistance sensing cannot provide accurate motion information in these situations. Here in this thesis, we draw inspiration from nature, design and developed soft robot that mimic the motion of the creatures, fiber optic sensors are embedded in the soft robot, to detect motion and event as well as the environmental perturbations. The perceptual soft robotics project presented in this thesis include human hand inspired soft robotic gripper that can detect the target object weight and actuation motions, twining plant inspired single-channel pneumatic soft spiral gripper that is sensitive to the twinning angle and target object diameter, and an inchworm-inspired soft robotic climber with 3 fiber optic sensor embedded to provide motion information, which is processed and assessed by machine learning, to achieve shape reconstruction consequently. The fiber optic sensor based soft robotic system is demonstrated to be effectively providing fast, precise, stable, and repeatable data for motion prediction, making it promising for close loop feedback controlling.