Eliminating Persistent Boundary Residence via Matrosov-Type Auxiliary Functions
This addresses boundary sticking issues in safety-critical control systems, offering a general solution without modifying existing controllers, though it is incremental as it builds on standard barrier function methods.
The paper tackled the problem of trajectories sticking to the boundary of safe sets in control barrier functions, which can cause deadlock, by introducing an auxiliary function framework that eliminates persistent boundary residence while preserving safety. The result ensures trajectories can only remain near the boundary for finite time, as demonstrated in numerical examples.
Control barrier functions enforce safety by guaranteeing forward invariance of an admissible set. Under standard (non-strict) barrier conditions, however, forward invariance alone does not prevent trajectories from remaining on the boundary of the safe set for arbitrarily long time intervals, potentially leading to boundary sticking or deadlock phenomena. This paper studies the elimination of persistent boundary residence under forward-invariant barrier conditions. Inspired by Matrosov-type arguments, we introduce an auxiliary function framework that preserves forward invariance while excluding infinite-time residence within boundary layers. Sufficient conditions are established under which any trajectory can only remain in a prescribed neighborhood of the boundary for finite time, thereby restoring boundary-level liveness without altering forward invariance. The proposed construction does not rely on singular barrier formulations or controller-specific modifications, and can be incorporated into standard safety-critical control architectures. Numerical examples illustrate the removal of boundary sticking behaviors while maintaining safety across representative systems.