# Neural Simplex Architecture

## TL;DR

The Neural Simplex Architecture (NSA) offers a novel runtime assurance framework for neural controllers, enabling safety guarantees and online safety improvements through retraining, demonstrated on autonomous navigation and medical control systems.

## Contribution

NSA extends the Simplex control architecture by allowing control switching back to the neural controller and enabling online safety correction and retraining.

## Key findings

- NSA provides safety guarantees for neural controllers in complex systems.
- NSA improves neural controller safety through online retraining.
- Case studies demonstrate NSA's effectiveness in autonomous navigation and medical systems.

## Abstract

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.