Self-Triggered Control
Self-triggered control predicts the next update time at the current update, eliminating the continuous monitoring that event-triggered control requires.
Predicting the next event
Event-triggered control saves updates but still needs a mechanism watching the state constantly to detect when the trigger fires. Self-triggered control removes that need: at each update it uses the model and the current state to compute in advance how long the current control can safely be held before a threshold would be crossed, then schedules the next update at that time. Between updates the system runs open-loop with no sensing at all.
How the interval is computed
The next sampling interval is derived by bounding how fast the error can grow under the held control, using the plant model and Lyapunov analysis. A conservative bound guarantees the triggering condition is not violated before the scheduled time. Because the prediction uses a worst-case bound, self-triggered intervals are typically shorter, more conservative, than the actual event-triggered intervals would be, trading some efficiency for freedom from monitoring.
Trade-offs
The scheme depends more heavily on model accuracy than event triggering, because it commits to an open-loop interval based on prediction rather than reacting to the true state. Disturbances during an interval are unaddressed until the next scheduled update, so robustness must be built into the interval computation. Combining prediction with a lightweight safety check yields hybrid schemes.
- Compute the next update time in advance
- No continuous state monitoring needed
- Intervals are conservative because prediction is worst-case
- More sensitive to model error and disturbances
Self-triggered control suits systems where continuous monitoring is itself costly, such as battery-powered sensors that can sleep between scheduled wake-ups, maximizing the benefit of skipped activity. It is the natural partner to event-triggered control in resource-aware design.
For a design-stage low-power distributed sensing network, self-triggering would let nodes sleep for predicted intervals, cutting energy use, evaluated in simulation. Both methods target networked control resource constraints.