CONIX Publication

Safe navigation with controlled invariant sets

Authors: Tzanis Anevlavis, Luigi Pannocchi, Marcus Lucas, Paulo Tabuada


We present a safety-critical supervisory control approach that is modular and can be combined with any arbitrary nominal controller. In supervisory control, given a system Σ, a set of safe states S, and a nominal controller, the goal is to evaluate whether a given nominal input results in the state of Σ violating the safety constraints modeled by S. If so, modify the input appropriately so that the safety constraints are always satisfied. To accomplish this goal, we leverage the theory of Robust Controlled Invariant Sets (RCIS), which provide formal guarantees that an appropriate modification always exists. We developed novel algorithms on the computation of explicit and implicit representations of RCISs and present their applicability in safe navigation and obstacle avoidance using a Crazyflie 2.0 drone.

Release Date: 10/13/2021
Uploaded File: View