State Estimation of a Soft Robotic Finger with Dynamic Effect of Parameter Uncertainty

Pyshical flexibility is one of the good aspects of soft robotic hands when grasping an unknown-shaped object stably or interacting with around environment safely. On the other hand, considering controlling them dexterously, their flexibility can cause nonlinear and uncertain behaviors and this will be the barrier to accurate control. Furthermore, they deform continuously and entirely, which makes it difficult to use some sensors and to control by direct sensor feedback. For such a soft robot system, state estimation is the key to realizing accurate control. It is necessary to improve the accuracy of a model for good state estimation. Recently, probabilistic models have been proposed to represent uncertainties in soft robots, and this method can overcome traditional deterministic models that sometimes exhibit good but sometimes undesirable behaviors in terms of soft robots. We also have proposed the dynamic model of a soft finger with stochastic parameters. Because this model is the extension of the traditional dynamics, it is easy to apply the traditional state estimation method. In this paper, the state estimation method that uses the stochastic characteristics of the model is proposed. Through experiments, the efficiency of the proposed estimation is investigated.