New recursive CBF design tackles nested STL specs for uncertain systems
No prior disturbance knowledge needed — a breakthrough in robust control synthesis.
Control of uncertain systems under complex temporal constraints — like ensuring a robot reaches a target within a time window while avoiding obstacles — has long been a challenge. Traditional Control Barrier Functions (CBFs) struggle with nested temporal operators (e.g., 'eventually always'), and reachability-based methods that can handle such logic are computationally heavy and fail under model uncertainty. This paper from Peng, Yan, and Wang tackles exactly that gap.
The authors introduce a recursive CBF design built on a modified Signal Temporal Logic tree (sTLT). By introducing sliding window variables, they capture intricate temporal dependencies. Their key contribution: a quadratic programming (QP) controller that reconstructs the CBF online to guarantee constraint satisfaction even with unknown bounded disturbances. Unlike prior methods, it requires no prior disturbance bounds and relaxes initial safety assumptions. Simulations on benchmark examples confirm the approach works robustly.
- Recursive CBF design uses a modified STL tree (sTLT) to handle nested temporal operators like 'eventually always'.
- Sliding window variables encode complex time-window relationships without relying on reachability analysis.
- The QP-based controller requires no prior knowledge of disturbances and relaxes initial safety assumptions.
Why It Matters
Enables robust, safety-critical control for autonomous systems (e.g., drones, robots) under uncertain disturbances and complex mission logic.