13 October 2021 in Journal of Artificial Intelligence Research
Autonomous cyber-physical systems (CPS) can improve safety and efficiency for safety-critical applications, but require rigorous testing before deployment. The complexity of these systems often precludes the use of formal verification and real-world testing can be too dangerous during development. Therefore, simulation-based techniques have been developed that treat the system under test as a black box operating in a simulated environment. Safety validation tasks include finding disturbances in the environment that cause the system to fail (falsification), finding the most-likely failure, and estimating the probability that the system fails. Motivated by the prevalence of safety-critical artificial intelligence, this work provides a survey of state-of-the-art safety validation techniques for CPS with a focus on applied algorithms and their modifications for the safety validation problem. We present and discuss algorithms in the domains of optimization, path planning, reinforcement learning, and importance sampling. Problem decomposition techniques are presented to help scale algorithms to large state spaces, which are common for CPS. A brief overview of safety-critical applications is given, including autonomous vehicles and aircraft collision avoidance systems. Finally, we present a survey of existing academic and commercially available safety validation tools.
14 May 2015 in Journal of Physics: Conference Series
27 December 2017 in The Astrophysical Journal
Galactic rotation curves have proven to be the testing ground for dark matter bounds in spiral galaxies of all morphologies. Dwarf galaxies serve as an increasingly interesting case of rotation curve dynamics due to their typically rising rotation curve as opposed to the flattening curve of large spirals. Dwarf galaxies usually vary in galactic structure and mostly terminate at small radial distances. This, coupled with the fact that Cold Dark Matter theories struggle with the universality of galactic rotation curves, allow for exclusive features of alternative gravitational models to be analyzed. Recently, The H i Nearby Galactic Survey (THINGS) has been extended to include a sample of 25 dwarf galaxies now known as the LITTLE THINGS Survey. Here, we show an application of alternative gravitational models to the LITTLE THINGS survey, specifically focusing on conformal gravity (CG) and Modified Newtonian Dynamics (MOND). In this work, we provide an analysis and discussion of the rotation curve predictions of each theory to the sample. Furthermore, we show how these two alternative gravitational models account for the recently observed universal trends in centripetal accelerations in spiral galaxies. This work highlights the similarities and differences of the predictions of the two theories in dwarf galaxies. The sample is not large or diverse enough to strongly favor a single theory, but we posit that both CG and MOND can provide an accurate description of the galactic dynamics in the LITTLE THINGS sample without the need for dark matter.
1 July 2024 in Journal of Aerospace Information Systems
Estimating the probability of failure is an important step in the certification of safety-critical systems. Efficient estimation methods are often needed due to the challenges posed by high-dimensional input spaces, risky test scenarios, and computationally expensive simulators. This work frames the problem of black-box safety validation as a Bayesian optimization problem and introduces a method that iteratively fits a probabilistic surrogate model to efficiently predict failures. The algorithm is designed to search for failures, compute the most-likely failure, and estimate the failure probability over an operating domain using importance sampling. We introduce three acquisition functions that aim to reduce uncertainty by covering the design space, optimize the analytically derived failure boundaries, and sample the predicted failure regions. Results show this Bayesian safety validation approach provides a more accurate estimate of failure probability with orders of magnitude fewer samples and performs well across various safety validation metrics. We demonstrate this approach on three test problems, a stochastic decision-making system, and a neural-network-based runway detection system. This work is open-sourced ( https://github.com/sisl/BayesianSafetyValidation.jl ) and is currently being used to supplement the FAA certification process of the machine learning components for an autonomous cargo aircraft.
1 May 2017 in Journal of Physics: Conference Series
In recent years, due to the lack of direct observed evidence of cold dark matter, coupled with the shrinking parameter space to search for new dark matter particles, there has been increased interest in Alternative Gravitational theories. This paper, addresses three recent advances in conformal gravity, a fourth order renormalizable metric theory of gravitation originally formulated by Weyl, and later advanced by Mannheim and Kazanas. The first section of the paper applies conformal gravity to the rotation curves of the LITTLE THINGS survey, extending the total number of rotation curves successfully fit by conformal gravity to well over 200 individual data sets without the need for additional dark matter. Further, in this rotation curve study, we show how MOND and conformal gravity compare for each galaxy in the sample. Second, we look at the original Zwicky problem of applying the virial theorem to the Coma cluster in order to get an estimate for the cluster mass. However, instead of using the standard Newtonian potential, here we use the weak field approximation of conformal gravity. We show that in the conformal case we can get a much smaller mass estimate and thus there is no apparent need to include dark matter. We then show that this calculation is in agreement with the observational data from other well studied clusters. Last, we explore the calculation of the deflection of starlight through conformal gravity, as a first step towards applying conformal gravity to gravitaitonal lensing.