Source author record

Umberto Ravaioli

Umberto Ravaioli appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

cond-mat.mes-hall cond-mat.mtrl-sci eess.SY Machine Learning physics.comp-ph Systems and Control

Catalog footprint

What is connected

4works

6topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2022arXiv

A Universal Framework for Generalized Run Time Assurance with JAX Automatic Differentiation

With the rise of increasingly complex autonomous systems powered by black box AI models, there is a growing need for Run Time Assurance (RTA) systems that provide online safety filtering to untrusted primary controller output. Currently, research in RTA tends to be ad hoc and inflexible, diminishing collaboration and the pace of innovation. The Safe Autonomy Run Time Assurance Framework presented in this paper provides a standardized interface for RTA modules and a set of universal implementations of constraint-based RTA capable of providing safety assurance given arbitrary dynamical systems and constraints. Built around JAX, this framework leverages automatic differentiation to populate advanced optimization based RTA methods minimizing user effort and error. To validate the feasibility of this framework, a simulation of a multi-agent spacecraft inspection problem is shown with safety constraints on position and velocity.

preprint2019arXiv

Prediction and optimization of mechanical properties of composites using convolutional neural networks

In this paper, we develop a convolutional neural network model to predict the mechanical properties of a two-dimensional checkerboard composite quantitatively. The checkerboard composite possesses two phases, one phase is soft and ductile while the other is stiff and brittle. The ground-truth data used in the training process are obtained from finite element analyses under the assumption of plane stress. Monte Carlo simulations and central limit theorem are used to find the size of the dataset needed. Once the training process is completed, the developed model is validated using data unseen during training. The developed neural network model captures the stiffness, strength, and toughness of checkerboard composites with high accuracy. Also, we integrate the developed model with a genetic algorithm (GA) optimizer to identify the optimal microstructural designs. The genetic algorithm optimizer adopted here has several operators, selection, crossover, mutation, and elitism. The optimizer converges to configurations with highly enhanced properties. For the case of the modulus and starting from randomly-initialized generation, the GA optimizer converges to the global maximum which involves no soft elements. Also, the GA optimizers, when used to maximize strength and toughness, tend towards having soft elements in the region next to the crack tip.

preprint2011arXiv

A Backscattering Model Incorporating the Effective Carrier Temperature in Nano MOSFET

In this work we propose a channel backscattering model in which increased carrier temperature at the top of the potential energy barrier in the channel is taken into account. This model represents an extension of a previous model by the same authors which highlighted the importance of considering the partially ballistic transport between the source contact and the top of the potential energy barrier in the channel. The increase of carrier temperature is precisely due to energy dissipation between the source contact and the top of the barrier caused by the high saturation current. To support our discussion, accurate 2D full band Monte Carlo device simulations with quantum correction have been performed in double gate nMOSFETs for different geometries (gate length down to 10 nm), biases and lattice temperatures. Including the effective carrier temperature is especially important to properly treat the high inversion regime, where previous backscattering models usually fail.

preprint2002arXiv

Electronic response and bandstructure modulation of carbon nanotubes in a transverse electrical field

The electronic properties of carbon nanotubes in a uniform transverse field are investigated within a single orbital tight-binding model. For doped nanotubes, the dielectric function is found to depend not only on symmetry of the tube, but also on radius and Fermi level position. Bandgap opening/closing is predicted for zigzag tubes, while it is found that armchair tubes always remain metallic, which is explained by the symmetry in their configuration. The bandstructures for both types are considerably modified when the field strength is large enough to mix neighboring subbands.