A preprint outlines a control-theoretic definition of harmonic stability and a computational test based on linear time-periodic and harmonic state-space models. In one numerical converter case, the method supported a local stability certificate tied to a periodic operating trajectory.
The work pairs a passivity-shortage analysis with a wave-based communication law, then examines its behavior in a nonlinear two-degree-of-freedom robot model.
A model-based wireless study tests velocity as an information channel for movable antennas, finding that receiver knowledge changes the payload while the baseline mechanical requirements remain far beyond the paper’s stated feasibility.
A new arXiv preprint describes a lower-complexity way to calculate soft-decoding inputs for noisy optical links. In modeled PAM-4 and PAM-8 systems, its results matched an exact calculation in several key comparisons, but the work remains simulation-based.
A preprint describes a nonlinear controller for stochastic watershed models and reports an 83.6% lower tracking-error norm than optimized static feedback in a high-fidelity SWMM/PySWMM simulation. The guarantee applies only under the study’s stated assumptions.
A mathematical and simulation study reports bounded synchronization for a modeled two-robot teleoperation system with fixed communication delays. The result is limited to the paper’s assumptions and has not been validated on physical hardware.
A numerical arXiv preprint describes a constrained way to optimize radar waveforms for wideband beam shaping, while stressing that the result remains an illustrative model.
An arXiv preprint reports sub-sample timing error and less than 2 dB loss versus perfect channel information in simulations of a single-user OTFS system with RF impairments. Timing estimates worsened as the number of delay bins increased, while carrier-frequency-offset and effective-channel estimates became more accurate.
A preprint reports higher modeled secrecy rates when a UAV-mounted intelligent reflecting surface’s pose and element-level reflection control are optimized alongside the base station’s beamformer. The simulations favored the proposed design over orientation-only and fixed alternatives, but supplied no exact gain or uncertainty estimate.
An arXiv preprint presents a unified pilot-control concept for a transformational eVTOL. The system uses one inceptor across hover, transition and cruise, with command filtering, full-envelope control and force feedback. Simulations were stable and smooth, while model-based tests found the filtered cases slower than a closed-loop comparison. Human-factors findings remain preliminary.
An arXiv preprint describes how operators could choose reactive-power settings before active-power injections are known. Its strongest result is conditional on a linear model and a feasible nominal-orthant setting.
An arXiv preprint presents a differentiable finite-volume framework intended to size heat-pump hardware, simulate system dynamics and support predictive control. Its benchmark results are mixed, and the control evidence remains numerical rather than laboratory-based.
A benchmark of highway traffic microsimulation methods found that no single approach led on every small-network measure. Performance declined as network size, vehicle count and simulation duration increased, while parameter choices and sensor error strongly influenced reported quality.
A preprint presents EvEMTBench, an open synthetic dataset designed for machine-learning research on power-system protection. It contains high-frequency, grid-wide measurements and labels for faults and operating events across several dataset designs.
A new sonar-tracking preprint models changing multipath background directly in raw measurements. Its proposed methods more often confirmed weak targets and generally achieved better localization in simulations, while experimental validation remains future work.
A high-effort Sonnet-5 system met the study’s hard-constraint endpoint in all 39 primary benchmark episodes, matching an engineered reference controller. The test also found stronger fault identification than a scripted supervisor, while leaving real-world deployment and broader generalization unanswered.