Backend
Simulation Modeling and Control of an Elastic Shaping Antagonistic Variable Stiffness Actuator
arXiv cs.SY0 views
arXiv:2611.07504v1 Announce Type: new Abstract: Position--stiffness regulation of antagonistic variable stiffness actuators (VSAs) involves coupled nonlinear dynamics, parameter uncertainty, and feedback states that are not directly measured. This paper addresses this challenge through a simulation-based workflow that integrates elastic shaping, model verification, and layered nonlinear control. This study builds on a previously developed actuator comprising two crossed four-bar compliant transmission elastic units. Its optimized elastic mapping provides the structural basis, with an ADAMS torque-curve root mean square error (RMSE) of $2.57\,\mathrm{N\!\cdot\!mm}$ against the target. A control-oriented analytical model is implemented in Simulink and compared with the independently executed ADAMS model under matched inputs. The maximum joint-angle RMSE across the tested cases is $4.66\times10^{-3}\,\mathrm{rad}$. A layered controller combines extended Kalman filtering for state reconstruction, feedback linearization for nominal decoupling, and adaptive sliding-mode control for residual compensation. At the end of the nominal simulation, the joint-angle error is approximately $3.14\times10^{-5}\,\mathrm{rad}$. The stiffness error is $3.07\,\mathrm{N\!\cdot\!mm/rad}$ relative to a target of $50\,\mathrm{N\!\cdot\!mm/rad}$. Parameter-variation and ablation tests assess tracking performance and the effects of the controller modules within the evaluated conditions. The resulting workflow links structural-response modeling, state estimation, and robust control assessment, supporting repeatable comparisons beyond nominal tracking alone.
Source:https://arxiv.org/abs/2611.07504PreviousSafe Output-Regulation Flows for Constrained Optimization via Control Barrier FunctionsNextdynamicalnodes: Block Diagrams to ROS2 Graphs
Related posts
Querying Counterfactuals on Tissue Graphs with Supervised Disentanglement
0arXiv stat.ML
One Method Does Not Fit All: A Large-Scale Benchmark of Imputation Methods for Single-Cell RNA Sequencing Data
0arXiv stat.ML
The Stability of Online Algorithms in Performative Prediction
0arXiv stat.ML
FedReason: Large Language Models as Weight-Space Meta-Aggregators for Heterogeneous Federated Learning
0arXiv cs.CR
SensorGuard-RV: Sensor-to-Register Data Protection for Resource-Constrained RISC-V Devices
0arXiv cs.CR
Escalate or Close: How Fast-Tier Models Handle Held-Out Web Attack Families in Two-Tier Triage
0arXiv cs.CR