Prescribed-time control barrier functions guarantee a system reaches a safe set by a deadline. The cost is that control effort diverges as the deadline approaches — the controller works exponentially harder to meet the constraint, demanding infinite force at the boundary. In practice, actuators saturate, and the guarantee breaks exactly when it matters most.
Gadginmath et al. solve this by constricting the tube. They construct a time-varying safe region that starts large (containing the initial condition even if it's far from the target) and smoothly contracts to match the target safe set at the deadline. The system is always required to stay inside the tube, but the tube does the geometric work of shepherding the state toward safety. Because the tube shrinks gradually, the control effort needed to stay inside it remains bounded throughout.
The key insight is that feasibility reduces to a single verifiable condition: checking whether the tube's contraction rate is compatible with the system's dynamics and input constraints. If it is, the entire trajectory is feasible with bounded control. The approach scales to 16-dimensional multi-agent systems — not a theoretical abstraction.
The structural point: the divergent control effort in prior methods wasn't a fundamental cost of guaranteed safety. It was an artifact of demanding the system reach a fixed target while allowing arbitrary initial conditions. By letting the target come to the system — the tube contracts onto the safe set rather than the system being driven into it — the geometric burden shifts from the controller to the constraint, and the infinity disappears.