Search

We present the Neural Simplex Architecture (NSA), a new approach to runtime assurance that provides safety guarantees for neural controllers (obtained e.g. using reinforcement learning) of autonomous and other complex systems without unduly sacrificing performance. NSA is inspired by the Simplex control architecture of Sha et al., but with some significant differences. In the traditional approach, the advanced controller (AC) is treated as a black box; when the decision module switches control to the baseline controller (BC), the BC remains in control forever. There is relatively little work on switching control back to the AC, and there are no techniques for correcting the AC's behavior after it generates a potentially unsafe control input that causes a failover to the BC. Our NSA addresses both of these limitations. NSA not only provides safety assurances in the presence of a possibly unsafe neural controller, but can also improve the safety of such a controller in an online setting via retraining, without overly degrading its performance. To demonstrate NSA's benefits, we have conducted several significant case studies in the continuous control domain. These include a target-seeking ground rover navigating an obstacle field, and a neural controller for an artificial pancreas system., Comment: 12th NASA Formal Methods Symposium (NFM 2020)

In earlier work, we presented a process algebra, PACSR, that uses a notion of resource failure to capture probabilistic behavior in reactive systems. PACSR also supports an operator for resource hiding. In this paper, we carefully consider the interaction between these two features from an axiomatic perspective. For this purpose, we introduce a subset of PACSR, called "PACSR-lite", that allows us to isolate the semantic issues surrounding resource hiding in a probabilistic setting, and provide a sound and complete axiomatization of strong bisimulation for this fragment. © 2001 Elsevier Science B.V. All rights reserved. 80 1 3 13

PACSR, a probabilistic extension of the real-time process algebra ACSR, is presented. The extension is built upon a novel treatment of the notion of a resource. In ACSR, resources are used to model contention in accessing physical devices. Here, resources are invested with the ability to fail and are associated with a probability of failure. The resulting formalism allows one to perform probabilistic analysis of real-time system specifications in the presence of resource failures. A probabilistic variant of Hennessy-Milner logic with until is presented. The logic features an until operator which is parameterized by both a probabilistic constraint and a regular expression over observable actions. This style of parameterization allows the application of probabilistic constraints to complex execution fragments. A model-checking algorithm for the proposed logic is also given. Finally, PACSR and the logic are illustrated with a telecommunications example. © 1998, Springer Verlag. All rights reserved. 1466 389 404 Sponsors: Conference code: 143949 Cited By :10

Neural state classification (NSC) is a recently proposed method for runtime predictive monitoring of hybrid automata (HA) using deep neural networks (DNNs). NSC trains a DNN as an approximate reachability predictor that labels an HA state x as positive if an unsafe state is reachable from x within a given time bound, and labels x as negative otherwise. NSC predictors have very high accuracy, yet are prone to prediction errors that can negatively impact reliability. To overcome this limitation, we present neural predictive monitoring (NPM), a technique that complements NSC predictions with estimates of the predictive uncertainty. These measures yield principled criteria for the rejection of predictions likely to be incorrect, without knowing the true reachability values. We also present an active learning method that significantly reduces the NSC predictor’s error rate and the percentage of rejected predictions. We develop two versions of NPM based, respectively, on the use of frequentist and Bayesian techniques to learn the predictor and the rejection rule. Both versions are highly efficient, with computation times on the order of milliseconds, and effective, managing in our experimental evaluation to successfully reject almost all incorrect predictions. In our experiments on a benchmark suite of six hybrid systems, we found that the frequentist approach consistently outperforms the Bayesian one. We also observed that the Bayesian approach is less practical, requiring a careful and problem-specific choice of hyperparameters.

In this paper, we investigate the transmission-power assignment problem for κ-connected Mobile Ad hoc NET-works (MANETs), the problem of optimizing the lifetime of a MANET at a given degree κ of connectivity by minimizing power consumption. Our proposed solution is fully distributed and uses a model-based transmission power adaptation strategy based on model-predictive control. Specifically, a stochastic model of the network is used by a state estimator to predict the network's future degree of connectivity and remaining energy. The predicted states are used by an optimizer to derive an optimal transmission power assignment sequence which tracks the desired connectivity level κ. while minimizing energy consumption. Our experimental results on a simulated wireless sensor network comprising 100 mobile nodes reveals that our localized topology control algorithm provides an almost identical control policy to that of a globalized scheme which is solving a fully observable problem. The difference, of course, is in the scalability of our localized solution, which requires much less communication bandwidth and energy than the globalized approach. © 2008 IEEE.