Event-Based Control for Resource-Constrained Systems

93 concepts92 papers3 proposals

In traditional digital control systems, the standard paradigm is time-triggered sampling: sensors, controllers, and actuators communicate on a rigid, periodic schedule governed by a central clock. While mathematically convenient, this periodic approach suffers from a fundamental inefficiency. It forces a system to sample and transmit data at a uniform rate optimized for worst-case disturbances, even when the system is perfectly stable or near equilibrium.

In modern applications—such as wireless sensor networks, internet-of-things (IoT) devices, autonomous drones, and large-scale multi-agent swarms—this constant, redundant data transmission wastes precious communication bandwidth, computational cycles, and battery power. To overcome this periodic sampling inefficiency, researchers have developed event-based control, a paradigm where communication and computation occur only when the state of the system actually demands it.

The Inefficiency of Time-Triggered Control

To understand why periodic control is inefficient, consider a system at steady state. As noted by Qing-Shan Jia in a guest editorial, constant data transmission consumes excessive bandwidth and energy even when no critical changes have occurred. Traditional embedded systems are over-engineered to handle the worst-case scenario, leading to significant oversampling during quiet periods.

This rigidity creates a harsh tradeoff. If the sampling interval is set too long to conserve resources, the system risks instability during sudden environmental disturbances. If it is set too short, it clogs shared communication channels and drains batteries. Feng Xia et al. demonstrate in Flexible Time-Triggered Sampling in Smart Sensor-Based Wireless Control Systems that rigid rates prevent systems from adapting to dynamic network congestion. Furthermore, in another work on embedded control systems, the authors leverage the non-linear relationship between sampling frequency and physical performance, showing that sampling intervals can be dramatically increased during stable states with negligible impact on control precision.

Event-Triggered vs. Self-Triggered Control

To decouple control execution from the absolute passage of time, the literature splits into two primary strategies: Event-Triggered Control (ETC) and Self-Triggered Control (STC).

Periodic:        |-------|-------|-------|-------|-------|-------| (Fixed intervals)
Event-Triggered: |-------|--|------------|---|-----------|-------| (Triggered by state error)
Self-Triggered:  |-------|---|----------|-----|------------|-----| (Next time computed ahead)

Event-Triggered Control (ETC)

In ETC, the system state is monitored continuously (or at a very high frequency), but a control update or network transmission is only executed when a specific condition is violated. Typically, this condition is defined by a threshold on the measurement error—the difference between the current state and the state at the last transmission instance.

As formalized by Romain Postoyan et al. in A Framework for the Event-Triggered Stabilization of Nonlinear Systems, this approach ensures that communication resources are only spent when the state deviates significantly from its expected trajectory. However, a core limitation of standard ETC is that the sensors must still continuously monitor the physical plant to evaluate the triggering condition.

Self-Triggered Control (STC)

To eliminate the need for continuous state monitoring, STC shifts the burden of timing from the sensor's hardware to the controller's processor. At each update instance, the controller uses a mathematical model of the system to predict how the state will evolve and computes the exact future time instance when the error threshold will be breached.

Manuel Mazo et al. provide an input-to-state stability (ISS) framework for linear controllers in An ISS self-triggered implementation of linear controllers, showing that calculating the "next best time" to run preserves stability while allowing sensors and communication hardware to enter low-power sleep states between updates. To reduce the conservatism often found in nonlinear systems, Adolfo Anta introduces an approach in Exploiting isochrony in self-triggered control that leverages isochronous manifolds to compute longer, less conservative update intervals.

Mathematical Foundations: Hybrid Systems and Stability

Because event-based control mixes continuous physical dynamics with discrete communication events, it cannot be analyzed using standard continuous or discrete control theory alone. Instead, it is modeled using hybrid systems theory.

The primary challenge is ensuring that the system remains stable despite long intervals without updates. In a unifying framework, Romain Postoyan et al. utilize hybrid Lyapunov functions to guarantee asymptotic stability. Rather than forcing the Lyapunov function to strictly decrease at every single instant, modern event-based conditions allow the system's energy to temporarily increase, provided it stays within a well-defined envelope. For example, Bhathiya Rathnayake et al. demonstrate in Performance-Barrier-Based Event-Triggered Boundary Control of a Class of Reaction-Diffusion PDEs that allowing the Lyapunov function to vary below an exponentially decaying performance barrier drastically increases the average time between updates, proving that strict monotonic decrease is an inefficient constraint for digital control.

The Zeno Phenomenon and Periodic ETC

A critical mathematical hazard in event-triggered systems is Zeno behavior, where the triggering condition is satisfied infinitely many times in a finite time interval (e.g., the controller updates every microsecond, then every nanosecond). This completely defeats the purpose of resource conservation.

To prevent this, researchers often implement Periodic Event-Triggered Control (PETC). In PETC, the triggering condition is evaluated only at fixed clock cycles rather than continuously. D.P. Borgers et al. present a method in Periodic event-triggered control of nonlinear systems using overapproximation techniques to convert continuous triggering rules into discrete-time periodic checks, guaranteeing a minimum positive dwell time between transmissions while preserving stability.

Advanced Paradigms: Predictive, Data-Driven, and Learned Triggering

As the field has matured, researchers have moved beyond static triggering rules to incorporate predictive models, data-driven strategies, and machine learning.

Model-Based Prediction

Standard event-triggered mechanisms use a zero-order hold, keeping the control input completely flat between updates. This ignores the natural physics of the system. To improve on this, Michael Hertneck et al. propose a framework in Model-Based Nonlinear Periodic Event-Triggered Control for Continuous-Time Systems with Sampled-Data Prediction that runs a simple, discrete-time model at the actuator. By "guessing" how the system state evolves between updates, the actuator adjusts its control inputs dynamically, allowing for significantly longer intervals between network transmissions.

Data-Driven and Predictive Triggering

When a precise mathematical model of the system is unavailable, traditional STC cannot compute future update times. To circumvent this, Wenjie Liu et al. introduce a method in Data-driven Self-triggered Control via Trajectory Prediction that uses historical trajectory data to predict future states and schedule communication without explicit system identification.

Furthermore, standard event-triggered systems suffer from a lack of lead time; they decide to transmit data at the exact millisecond it is needed. Sebastian Trimpe addresses this in Predictive and Self Triggering for Event-based State Estimation by forecasting communication needs multiple steps into the future, allowing a shared network infrastructure to proactively reallocate unused bandwidth slots to other nodes.

Learning Optimal Triggering Policies

Greedy event-triggering conditions—which only look at the current state error—are often suboptimal over long time horizons. Proactively triggering an update before a threshold is reached can sometimes yield a much longer quiet period afterward.

To discover these non-intuitive, long-horizon policies, researchers have turned to reinforcement learning. Dominik Baumann et al. apply this in Deep Reinforcement Learning for Event-Triggered Control, penalizing communication frequency in the reward function to let an agent learn simultaneously how to control and when to communicate.

Greedy ETC:  |--|--|--|--|--|--| (Reacts immediately to threshold breaches)
Learned ETC: |----|--------|---| (Proactively updates early to secure long future rest periods)

To ensure that these learned policies do not violate physical safety boundaries, Pio Ong et al. propose a hierarchical framework in Hierarchical Event-Triggered Systems: Safe Learning of Quasi-Optimal Deadline Policies. This architecture decouples the safety guarantees from the resource optimization, allowing a machine learning agent to aggressively optimize communication deadlines while a formal safety layer ensures the system never breaches its critical operational boundaries.

Key Foundational Work

Enhanced Energy-Aware Feedback Scheduling of Embedded Control Systems (2008) — contributed 2 concept elements to this direction.
Flexible Time-Triggered Sampling in Smart Sensor-Based Wireless Control Systems (2008) — contributed 2 concept elements to this direction.
Self-Triggered Control: trading actuation for computation (2009) — contributed 1 concept elements to this direction.
History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 81 papers
  2. Version 1 · 2026-10-04 · 80 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_periodic_sampling_inefficiency_3fc28626,
  title = {periodic sampling inefficiency},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/periodic-sampling-inefficiency/rdr_3fc286261ab24b6dbca8775453c910c6},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 2, frozen text):

@misc{lacuna_periodic_sampling_inefficiency_3fc28626_v2,
  title = {periodic sampling inefficiency},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/periodic-sampling-inefficiency/rdr_3fc286261ab24b6dbca8775453c910c6/v/2},
  note = {Version 2, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake