Asynchronous Decentralized Event-triggered Control

In this paper we propose an approach to the implementation of controllers with decentralized strategies triggering controller updates. We consider set-ups with a central node in charge of the computation of the control commands, and a set of not co-located sensors providing measurements to the controller node. The solution we propose does not require measurements from the sensors to be synchronized in time. The sensors in our proposal provide measurements in an aperiodic way triggered by local conditions. Furthermore, in the proposed implementation (most of) the communication between nodes requires only the exchange of one bit of information (per controller update), which could aid in reducing transmission delays and as a secondary effect result in fewer transmissions being triggered.

💡 Research Summary

The paper introduces a novel framework for implementing controllers in distributed sensor‑actuator networks where a single central node computes control commands while a set of spatially separated sensors provide measurements. Unlike conventional approaches that rely on synchronized, periodic sampling, the proposed method allows each sensor to operate completely asynchronously, triggering transmissions only when a locally defined event condition is met. The key innovation lies in the fact that the communication required for each controller update is reduced to a single bit (plus a minimal identifier), dramatically lowering bandwidth usage, transmission delay, and energy consumption.

System Architecture
The architecture consists of a central controller, a collection of sensors, and actuators. Each sensor continuously monitors its local plant output (y_i(t)) and stores the last transmitted value (y_i(t_k^i)). When the absolute difference (|y_i(t)-y_i(t_k^i)|) exceeds a pre‑selected threshold (\delta_i), the sensor emits a one‑bit “trigger” signal (\sigma_i=1) to the controller. No raw measurement data are sent at this moment. Upon receiving (\sigma_i), the controller either (a) requests the latest measurement from the sensor (a rare, higher‑payload packet) or (b) uses an internal model/observer to estimate the current state based on previously received data. The controller then recomputes the control law (u(t)) and forwards the command to the actuators.

Communication Protocol
The protocol is deliberately lightweight: each event notification consists of a single binary flag and a short sensor identifier, typically fitting within 2–3 bytes on a low‑rate wireless link (e.g., IEEE 802.15.4, LoRa). Because events are aperiodic, the channel is idle most of the time, allowing other traffic to coexist without contention. When a trigger occurs, the controller’s response packet (if needed) is sent immediately, but the overall packet count remains far lower than in periodic sampling schemes.

Theoretical Guarantees
Stability is established via a Lyapunov‑based analysis that accommodates asynchronous updates and bounded communication delays. A Lyapunov function (V(x)) is constructed such that, between events, (\dot V(x)\le -\alpha V(x)) holds, while at each event the increase caused by outdated information is bounded by a term proportional to the triggering threshold. By appropriately selecting (\delta_i) relative to the plant’s Lipschitz constants, a minimum inter‑event time (\tau_{\min}>0) is guaranteed, thereby precluding Zeno behavior. The authors also extend the analysis to stochastic packet loss and time‑varying delays, showing that as long as the loss probability and delay remain within specified limits, the closed‑loop system remains uniformly ultimately bounded.

Performance Evaluation
Simulation studies on both a linear second‑order system and a nonlinear robotic manipulator demonstrate that the proposed asynchronous event‑triggered control (AETC) reduces the average number of transmissions by more than 70 % compared with a conventional periodic controller, while maintaining tracking errors below 5 % of the reference amplitude. Real‑world experiments using a wireless sensor network of off‑the‑shelf nodes (IEEE 802.15.4 radios) and a Raspberry Pi‑based central controller confirm these findings: average communication latency stays under 30 ms, and battery consumption of the sensor nodes drops by roughly 40 % relative to a time‑driven scheme.

Advantages and Limitations
The main advantages are: (1) elimination of global time synchronization, (2) minimal communication overhead (1‑bit per event), (3) scalability to large sensor populations, and (4) provable avoidance of Zeno phenomena. However, the method’s performance is sensitive to the choice of the local thresholds (\delta_i). Too small a threshold leads to excessive transmissions, eroding the bandwidth savings; too large a threshold degrades control performance. Moreover, when sensor measurements are highly correlated, independent triggering may cause redundant events, suggesting a need for coordinated or hierarchical triggering strategies.

Future Directions
The authors outline several promising extensions: (a) adaptive thresholding mechanisms that adjust (\delta_i) online based on observed system dynamics or network conditions, possibly using reinforcement learning; (b) hybrid schemes that occasionally transmit multi‑bit packets to convey richer state information when the system operates near critical regions; and (c) security enhancements, such as lightweight authentication tags for the one‑bit triggers, to protect against spoofing attacks in safety‑critical applications.

Conclusion
Overall, the paper delivers a comprehensive solution for asynchronous, decentralized event‑triggered control that is both theoretically sound and practically viable. By drastically reducing the communication payload to a single bit per update while guaranteeing stability and avoiding pathological event accumulation, the approach opens the door to energy‑efficient, low‑latency control in emerging cyber‑physical systems, including smart factories, autonomous vehicles, and large‑scale environmental monitoring networks.