|Ph.D Student||Gamus Benny|
|Subject||Locomotion and Dynamics of Soft Robotic Actuators and|
|Department||Department of Mechanical Engineering||Supervisors||Professor Yizhar Or|
|Professor Amir Gat|
|Full Thesis text|
In nature, many soft-bodied creatures are capable of complex locomotion in challenging environments, inaccessible to skeletal animals. Their ability to squeeze through gaps smaller than their unconstrained body dimensions, negotiate unstructured environments and interact with humans and delicate objects has been among the motivations for the emerging bio-inspired field of soft robotics. This field revolutionizes the concept of robots' mechanical design, by introducing compliant structures with continuous actuation and deformations (rather than traditional chains of rigid links with local actuation at the joints).
The presented research starts with the investigation of soft robotic actuators, focusing on actuation by pressurized embedded fluidic networks (EFN). We present closed-form analytical solutions of such actuators, while accounting for the effects of solid inertia and elasticity, as well as fluid viscosity. This allows to model the system's step-response and frequency response, as confirmed by experiments. We also propose a novel method to significantly simplify the manufacturing process of EFN soft actuators by utilizing viscous-peeling. This method introduces extremely nonlinear phenomena into the fluid dynamics, which we model, numerically solve and validate in experiments.
We continue by studying the locomotion of soft robots that consist of several soft actuators. We propose modeling soft-robotic legged locomotion by approximating it with an articulated robot with elastic joints. For concreteness we study the quasistatic inchworm-like crawling of a soft robot which we have manufactured, via an articulated three-link model. The solution of statically indeterminate systems with passive stick-slip contact transitions (i.e. without friction manipulation) requires for a novel hybrid-quasistatic analysis. We utilize our analysis to investigate the influence of inputs' parameters on the performance of crawling gaits, with good agreement to experiments.
Finally, we extend the model of the three-link robot to a full hybrid-dynamical system, and investigate the influence of inertia and other inputs' parameters. We propose input-shaping technique and apply machine-learning based optimization, which further improves the robots performance. Both the quasistatic and dynamic locomotion were also studied for sensitivity to uncertainties in friction, which was shown to have major influence. The model and results of this part are confirmed by manufacturing a three-link robot and measuring its dynamics by image processing.
We hope that this study sheds some light on modeling methods as well as on the locomotion and dynamics of soft actuators and crawling robots, for future design and control studies.