ModellingGoogle представила Gemini 4 Neon — модель для кодинга и агентных задачХабр · 3hModellingCoupled pricing of dynamic wireless charging and grid loadarXiv — Systems and Control (eess.SY) · 23hDigital twinAn executable digital twin for UAV/UGV swarm coordinationarXiv — Multiagent Systems (cs.MA) · 23hForecastingMPC with Gaussian-process learned critic for uncertainty-aware controlarXiv — Systems and Control (eess.SY) · 23hForecastingDifferentiable-by-design nonlinear optimization for MPCarXiv — Systems and Control (eess.SY) · 23hDigital twinTeraStor Introduces Kinetra Digital Twin and Simulation Platform for AI Data CentersThe National Law Review · yesterdayDigital twinCan activation steering replace affective priming in LLM digital twinsarXiv API — полнотекстовый поиск "digital twin" · yesterdayDigital twinConversational assistant for a human-centric building digital twinarXiv API — полнотекстовый поиск "digital twin" · yesterdayDigital twinIndustrial AI and digital twins could help Vietnam escape assembly trapBáo VietNamNet · 2dDigital twinЦифровые двойники (Digital Twin) в инженерных системах — Часть 1. Управление документациейХабр · 2dDigital twinIndustrial AI and digital twins as Vietnam's path to technological self-relianceVietnam.vn · 2dDigital twinDigital twins should be built for valueAVEVA · 2d
Differentiable-by-design nonlinear optimization for MPC
US & West ·
The authors regularize the nonlinear optimization problem around the nominal solution to obtain a surrogate derivative even when the solution map is non-differentiable. This lets MPC policies be embedded in gradient-based policy optimization.
Why it matters
It simplifies training and tuning of predictive controllers for complex control systems.
The study couples traffic, EV energy, driver response and AC distribution network models in an MPC framework. It jointly optimizes dynamic wireless charging tolls and electricity prices, explicitly accounting for charging shifted to homes and network losses.
The authors present the first executable instantiation of the Guarded Swarms architecture as a digital twin built on ROS 2, PX4 software-in-the-loop and Gazebo. The scenario uses two UAVs and one UGV in a coastal counter-UAS task, pairing a Petri-net coordination layer with an independent analog safety layer that can veto unsafe commands.
The authors propose GP-RL-MPC, using a learned critic as the MPC terminal penalty to capture costs beyond the prediction horizon. Gaussian process regression enables data-efficient learning from noisy rollouts, and a tunable risk coefficient weights epistemic uncertainty.