A Recurrent Differentiable Engine for Modeling Tensegrity Robots Trainable with Low-Frequency Data | IEEE Conference Publication | IEEE Xplore