This thesis experimentally explores soft robotics in its characteristics of dynamics and control. Several types of robots with capability in locomotion and manipulation are developed.

At first, the thesis presents the exploration of dynamic performance of soft actuator based robotics by validating jumping locomotion. This chapter starts with the dynamic characterization of a disc shape soft actuator and a conceptualized prototype of jumping robot. Another part is the compensation of manufacturing deficiency within the soft actuator by using a machine learning algorithm, namely making the jumping robot "aware" of its "own insufficiency" and adaptively improving its performance.

Later, we report a dexterous soft actuator for robotic hand or end-effector with complete design, manu...[ Read more ]

This thesis experimentally explores soft robotics in its characteristics of dynamics and control. Several types of robots with capability in locomotion and manipulation are developed.

At first, the thesis presents the exploration of dynamic performance of soft actuator based robotics by validating jumping locomotion. This chapter starts with the dynamic characterization of a disc shape soft actuator and a conceptualized prototype of jumping robot. Another part is the compensation of manufacturing deficiency within the soft actuator by using a machine learning algorithm, namely making the jumping robot "aware" of its "own insufficiency" and adaptively improving its performance.

Later, we report a dexterous soft actuator for robotic hand or end-effector with complete design, manufacturing, and tests that is able to realize rapid soft actuations. The dexterous soft actuator is modified from the multi-segment actuator developed in jumping robot. The static performance of dexterous soft actuator is characterized with a number of experiments. The effects of material selection, fiber reinforcement and design criterions on the dexterous soft actuator are carefully studied and validated.

One important objective of this thesis is to enable the control of soft robotics or soft actuator based robotics. Currently, several inherent disadvantages inhibit the soft robotics or actuators from being well controlled, out of which large oscillation and manufacturing deficiency are the most significant. In this thesis, an effective methodology is proposed to alleviate the undesired oscillation within a dexterous soft actuator. Therefore, the feedback bending degree control of this actuator is enabled.

A bending control mechanism for a dexterous soft actuator using flex bending sensors is also presented. To alleviate the oscillation of the dexterous soft actuator during a quick actuation, a soft spring damper system is designed and evaluated. Based on this soft spring damper system, a feedback control of bending is developed and tested. Experimental results have proven that the bending control can only be enabled with the developed soft spring damper system integrated.

In the end, the thesis points out the insufficiencies and problems facing the current development of soft robotics. Lack of theoretical foundation and insufficient experimental analysis hold back the further application of soft robotics. A series of prospective topics, including new materials, soft robotics oriented controller and on-board power system, are proposed for the potential advancement of the research in soft robotics, in both academia and industry.