Learning Input Constrained Control Barrier Functions for Guaranteed Safety of Car-Like Robots
Authors:
Sven Brüggemann,
Dominic Nightingale,
Jack Silberman,
Maurício de Oliveira
Abstract:
We propose a design method for a robust safety filter based on Input Constrained Control Barrier Functions (ICCBF) for car-like robots moving in complex environments. A robust ICCBF that can be efficiently implemented is obtained by learning a smooth function of the environment using Support Vector Machine regression. The method takes into account steering constraints and is validated in simulatio…
▽ More
We propose a design method for a robust safety filter based on Input Constrained Control Barrier Functions (ICCBF) for car-like robots moving in complex environments. A robust ICCBF that can be efficiently implemented is obtained by learning a smooth function of the environment using Support Vector Machine regression. The method takes into account steering constraints and is validated in simulation and a real experiment.
△ Less
Submitted 19 February, 2024;
originally announced February 2024.
Numerical generation of vector potentials from specified magnetic fields
Authors:
Zachary J. Silberman,
Thomas R. Adams,
Joshua A. Faber,
Zachariah B. Etienne,
Ian Ruchlin
Abstract:
Many codes have been developed to study highly relativistic, magnetized flows around and inside compact objects. Depending on the adopted formalism, some of these codes evolve the vector potential $\mathbf{A}$, and others evolve the magnetic field $\mathbf{B}=\nabla\times\mathbf{A}$ directly. Given that these codes possess unique strengths, sometimes it is desirable to start a simulation using a c…
▽ More
Many codes have been developed to study highly relativistic, magnetized flows around and inside compact objects. Depending on the adopted formalism, some of these codes evolve the vector potential $\mathbf{A}$, and others evolve the magnetic field $\mathbf{B}=\nabla\times\mathbf{A}$ directly. Given that these codes possess unique strengths, sometimes it is desirable to start a simulation using a code that evolves $\mathbf{B}$ and complete it using a code that evolves $\mathbf{A}$. Thus transferring the data from one code to another would require an inverse curl algorithm. This paper describes two new inverse curl techniques in the context of Cartesian numerical grids: a cell-by-cell method, which scales approximately linearly with the numerical grid, and a global linear algebra approach, which has worse scaling properties but is generally more robust (e.g., in the context of a magnetic field possessing some nonzero divergence). We demonstrate these algorithms successfully generate smooth vector potential configurations in challenging special and general relativistic contexts.
△ Less
Submitted 27 March, 2018;
originally announced March 2018.
