ModellingParallel branch MPC on GPUsarXiv — Systems and Control (eess.SY) · yesterdayModellingImpact of shared autonomous vehicles in microtransit: Atlanta casearXiv — Systems and Control (eess.SY) · yesterdayModellingHybrid simulation of circular supply chains in healthcarearXiv — Systems and Control (eess.SY) · yesterdayDigital twinFDF tool simplifies building AI digital twinsarXiv API — полнотекстовый поиск "digital twin" · yesterdayForecastingMPC and genetic algorithm for flying pendulum controlarXiv — Systems and Control (eess.SY) · yesterdayForecastingHybrid control architecture for safe vehicle lateral controlarXiv — Systems and Control (eess.SY) · yesterdayForecastingAdaptive MPC for ground vehicles: review and implementationarXiv — Systems and Control (eess.SY) · yesterdayDigital twinML-assisted Bayesian Calibration of Accelerator Digital TwinarXiv API — полнотекстовый поиск "digital twin" · yesterdayDigital twinDigital Twin Testbed for Cyberattack and Defense Analysis in Hospital IT/OT EnvironmentsarXiv API — полнотекстовый поиск "digital twin" · yesterdayModellingGenerative Agent-based Simulation of Mobility Policy Preferences from SurveysarXiv — Computers and Society (cs.CY) · 2dForecastingScalable Gaussian Process with Trigonometric Features for Safe MPCarXiv — Systems and Control (eess.SY) · 2dForecastingTwo-Layer MPC for Sustainable Data Centers with Workload Flexibility and Heat RecoveryarXiv — Systems and Control (eess.SY) · 2d
Trajectory Planning and Budgeted Querying for Digital Twin Calibration
US & West ·
Researchers proposed a method for calibrating digital twins that optimizes the selection of trajectories and timing of costly parameter measurements. It uses an RL controller, a recurrent parameter estimator with uncertainty, and a budgeted query policy. In a Pendulum experiment, a GRU trained on excitation-oriented trajectories reached a mean absolute error of 0.0066 with no queries.
Why it matters
Important for improving digital twin calibration processes, allowing to reduce data collection costs and increase modeling accuracy.
Function+Data Flow (FDF), a visual domain-specific language for specifying and validating AI/ML pipelines used in building real-time digital twins, is presented. Implemented in DesCartes Builder, it enables composition and reuse of models. An empirical study found FDF makes AI-based twin development more accessible and reliable.
The study parametrizes hard-to-measure effects and calibrates the Bmad accelerator digital twin using Bayesian methods with beam measurements at AGS Booster, BNL. Computations are accelerated with an ML emulator.
A testbed emulating hospital IT and OT infrastructure, including EHR and SCADA, with a digital twin for monitoring and experimentation, is presented. It supports controlled cyberattacks, patch evaluation, and training of defense agents, enabling safe testing of defensive mechanisms in realistic settings.
