Search for: All records

Tensegrity robots are composed of rigid struts and flexible cables. They constitute an emerging class of hybrid rigid-soft robotic systems and are promising systems for a wide array of applications, ranging from locomotion to assembly. They are difficult to control and model accurately, however, due to their compliance and high number of degrees of freedom. To address this issue, prior work has introduced a differentiable physics engine designed for tensegrity robots based on first principles. In contrast, this work proposes the use of graph neural networks to model contact dynamics over a graph representation of tensegrity robots, which leverages their natural graph-like cable connectivity between end caps of rigid rods. This learned simulator can accurately model 3-bar and 6-bar tensegrity robot dynamics in simulation-to-simulation experiments where MuJoCo is used as the ground truth. It can also achieve higher accuracy than the previous differentiable engine for a real 3-bar tensegrity robot, for which the robot state is only partially observable. When compared against direct applications of recent mesh-based graph neural network simulators, the proposed approach is computationally more efficient, both for training and inference, while achieving higher accuracy. Code and data are available at https://github.com/nchen9191/tensegrity_gnn_simulator_public 

Tensegrity robots are composed of rigid struts and flexible cables. They constitute an emerging class of hybrid rigid-soft robotic systems and are promising systems for a wide array of applications, ranging from locomotion to assembly. They are difficult to control and model accurately, however, due to their compliance and high number of degrees of freedom. To address this issue, prior work has introduced a differentiable physics engine designed for tensegrity robots based on first principles. In contrast, this work proposes the use of graph neural networks to model contact dynamics over a graph representation of tensegrity robots, which leverages their natural graph-like cable connectivity between end caps of rigid rods. This learned simulator can accurately model 3-bar and 6-bar tensegrity robot dynamics in simulation-to-simulation experiments where MuJoCo is used as the ground truth. It can also achieve higher accuracy than the previous differentiable engine for a real 3-bar tensegrity robot, for which the robot state is only partially observable. When compared against direct applications of recent mesh-based graph neural network simulators, the proposed approach is computationally more efficient, both for training and inference, while achieving higher accuracy. Code and data are available at https://github.com/nchen9191/tensegrity_gnn_simulator_public 

Tensegrity robots are composed of rigid struts and flexible cables. They constitute an emerging class of hybrid rigid-soft robotic systems and are promising systems for a wide array of applications, ranging from locomotion to assembly. They are difficult to control and model accurately, however, due to their compliance and high number of degrees of freedom. To address this issue, prior work has introduced a differentiable physics engine designed for tensegrity robots based on first principles. In contrast, this work proposes the use of graph neural networks to model contact dynamics over a graph representation of tensegrity robots, which leverages their natural graph-like cable connectivity between end caps of rigid rods. This learned simulator can accurately model 3-bar and 6-bar tensegrity robot dynamics in simulation-to-simulation experiments where MuJoCo is used as the ground truth. It can also achieve higher accuracy than the previous differentiable engine for a real 3-bar tensegrity robot, for which the robot state is only partially observable. When compared against direct applications of recent mesh-based graph neural network simulators, the proposed approach is computationally more efficient, both for training and inference, while achieving higher accuracy. Code and data are available at https://github.com/nchen9191/tensegrity_gnn_simulator_public 

Abstract The recharge oscillator (RO) is a simple mathematical model of the El Niño Southern Oscillation (ENSO). In its original form, it is based on two ordinary differential equations that describe the evolution of equatorial Pacific sea surface temperature and oceanic heat content. These equations make use of physical principles that operate in nature: (a) the air‐sea interaction loop known as the Bjerknes feedback, (b) a delayed oceanic feedback arising from the slow oceanic response to winds within the equatorial band, (c) state‐dependent stochastic forcing from fast wind variations known as westerly wind bursts (WWBs), and (d) nonlinearities such as those related to deep atmospheric convection and oceanic advection. These elements can be combined at different levels of RO complexity. The RO reproduces ENSO key properties in observations and climate models: its amplitude, dominant timescale, seasonality, and warm/cold phases amplitude asymmetry. We discuss the RO in the context of timely research questions. First, the RO can be extended to account for ENSO pattern diversity (with events that either peak in the central or eastern Pacific). Second, the core RO hypothesis that ENSO is governed by tropical Pacific dynamics is discussed from the perspective of influences from other basins. Finally, we discuss the RO relevance for studying ENSO response to climate change, and underline that accounting for ENSO diversity, nonlinearities, and better links of RO parameters to the long term mean state are important research avenues. We end by proposing important RO‐based research problems. 

Developing suitable approximate models for analyzing and simulating complex nonlinear systems is practically important. This paper aims at exploring the skill of a rich class of nonlinear stochastic models, known as the conditional Gaussian nonlinear system (CGNS), as both a cheap surrogate model and a fast preconditioner for facilitating many computationally challenging tasks. The CGNS preserves the underlying physics to a large extent and can reproduce intermittency, extreme events, and other non-Gaussian features in many complex systems arising from practical applications. Three interrelated topics are studied. First, the closed analytic formulas of solving the conditional statistics provide an efficient and accurate data assimilation scheme. It is shown that the data assimilation skill of a suitable CGNS approximate forecast model outweighs that by applying an ensemble method even to the perfect model with strong nonlinearity, where the latter suffers from filter divergence. Second, the CGNS allows the development of a fast algorithm for simultaneously estimating the parameters and the unobserved variables with uncertainty quantification in the presence of only partial observations. Utilizing an appropriate CGNS as a preconditioner significantly reduces the computational cost in accurately estimating the parameters in the original complex system. Finally, the CGNS advances rapid and statistically accurate algorithms for computing the probability density function and sampling the trajectories of the unobserved state variables. These fast algorithms facilitate the development of an efficient and accurate data-driven method for predicting the linear response of the original system with respect to parameter perturbations based on a suitable CGNS preconditioner. 

Abstract We report on the preparation of and calibration measurements with a 83 mKr source for the CENNS-10 liquid argon detector. 83 mKr atoms generated in the decay of a 83 Rb source were introduced into the detector via injection into the Ar circulation loop. Scintillation light arising from the 9.4 keV and 32.1 keV conversion electrons in the decay of 83 mKr in the detector volume were then observed. This calibration source allows the characterization of the low-energy response of the CENNS-10 detector and is applicable to other low-energy-threshold detectors. The energy resolution of the detector was measured to be 9% at the total 83 mKr decay energy of 41.5 keV. We performed an analysis to separately calibrate the detector using the two conversion electrons at 9.4 keV and 32.1 keV.