Interval STL Runtime Assurance
Links:
This paper addresses safe control from signal temporal logic safety specifications for linear systems with bounded uncertainty in a static environment. The core move is to use interval signal temporal logic (iSTL) so the runtime-assurance problem can explicitly accommodate uncertainty instead of pretending it is negligible or hiding it inside blunt worst-case margins.
At each controller update step, the runtime-assurance algorithm takes the input proposed by a nominal controller and minimally adjusts it when necessary so that the safety specification remains satisfied at all times under all realizations of the bounded disturbance. Like related temporal-logic synthesis approaches, the method solves a mixed-integer linear program online, but iSTL keeps the uncertainty overhead comparatively modest.
The paper, with Leslie Baird and Samuel Coogan, is currently under revision for IFAC Nonlinear Analysis: Hybrid Systems (NAHS). It sits on the more formal-methods side of the same safe-autonomy thread that runs through the rest of the work on this site, but it still stays grounded in a real experimental platform: a miniature autonomous blimp.
Left: a nominal controller generates an unsafe blimp trajectory, while the runtime-assurance layer synthesizes a safe backup trajectory online under bounded disturbance. Right: the full control architecture, where a nominal policy feeds the runtime-assurance block, which synthesizes a safe reference at 2 Hz, up-samples it with polynomial splines to 50 Hz, and passes it to a Newton-Raphson Flow controller on the blimp.
Key ideas:
- Use interval signal temporal logic to encode safety requirements under bounded uncertainty
- Minimally modify the nominal controller’s proposed input instead of replacing the controller wholesale
- Solve an online MILP while retaining conditions that guarantee feasibility for all time
- Guarantee that a safe backup input is available even if a computation deadline is missed
- Demonstrate real-time computational tractability on a miniature autonomous blimp
Academic foundation:
| Venue | Status |
|---|---|
| IFAC Nonlinear Analysis: Hybrid Systems (NAHS) | Submitted, under revision |
Built around: Runtime assurance · Signal temporal logic · Interval analysis · Mixed-integer optimization · Hybrid systems