Safety Simplex Architecture: Bridging AI Autonomy with DO-178C DAL A

Using formal verification monitors to safely bound experimental machine learning flight algorithms.

The Complex Controller vs Safety Supervisor Dualism

A high-performance deep neural network (the Advanced Controller) pilots the aircraft with superhuman agility and obstacle avoidance. Concurrently, a lightweight, mathematically verified program (the Baseline Safety Controller) runs in lockstep, observing aircraft state variables (pitch, bank, airspeed, descent rate).

Bumpless Reversionary Switching Logic

If forward reachability sets project that the AI agent's command would cause a stall within 1.5 seconds, the hardware Simplex switch overrides the neural network instantly and seamlessly, executing an aerodynamic recovery climb.

Enables FAA type certification for neural networks by proving envelope safety invariants
Zero runtime performance penalty (< 500 nanosecond switching latency)
Fully decoupled codebases certified to differing Development Assurance Levels

Browse aerospace domains for sale