Abstract
The Internet of Things (IoT) interconnects multiple physical devices in large-scale networks. When the 'things' coordinate decisions and act collectively on shared information, feedback is introduced between them. Multiple feedback loops are thus closed over a shared, general-purpose network. Traditional feedback control is unsuitable for design of IoT control because it relies on high-rate periodic communication and is ignorant of the shared network resource. Therefore, recent event-based estimation methods are applied herein for resource-aware IoT control allowing agents to decide online whether communication with other agents is needed, or not. While this can reduce network traffic significantly, a severe limitation of typical event-based approaches is the need for instantaneous triggering decisions that leave no time to reallocate freed resources (e.g., communication slots), which hence remain unused. To address this problem, novel predictive and self triggering protocols are proposed herein. From a unified Bayesian decision framework, two schemes are developed: self triggers that predict, at the current triggering instant, the next one; and predictive triggers that check at every time step, whether communication will be needed at a given prediction horizon. The suitability of these triggers for feedback control is demonstrated in hardware experiments on a cart-pole, and scalability is discussed with a multi-vehicle simulation.
Introduction
In the Internet of Things (IoT), autonomous agents—from factory robots to vehicle platoons—must share a single wireless network to exchange state information for coordinated control. Traditional control paradigms rely on high-rate periodic communication, which quickly leads to network congestion as the number of devices scales.
While Event-Triggered Control (ETC) offers a solution by transmitting data only when a "significant event" occurs (e.g., an estimation error exceeds a threshold), it suffers from a critical reactive flaw: the decision to communicate is made instantaneously. By the time an agent decides not to use its assigned time slot, it is often too late for the network manager to reallocate that bandwidth to another device. To truly optimize shared resources, agents require a "heads-up"—a way to predict their communication needs several steps into the future.
Predictive Triggering Framework
This paper proposes a unified Bayesian decision framework that treats triggering as an optimization problem, balancing the cost of communication against the expected cost of estimation error. This framework yields two primary strategies:
- Self-Triggering (ST): At the moment of transmission, the agent calculates the exact future instant when the next update will be required based on its internal model. This allows sensors to remain in a low-power "sleep" state between triggers.
- Predictive Triggering (PT): At every time step $k$, the agent looks $M$ steps into the future. It uses current real-time data and model-based uncertainty to decide if communication will be necessary at time $k+M$. This provides the network manager with a lead time of $M$ steps to reallocate resources.
As shown in the architecture above, the system utilizes Distributed Event-Based State Estimation (DEBSE). Every agent runs a local predictor of every other agent's state, only requesting actual data updates when the predicted error is expected to become intolerable.
Experimental Evaluation and Results
The triggers were evaluated using both a high-stakes hardware experiment (an inverted pendulum cart-pole) and a 10-vehicle platoon simulation to test scalability and robustness.
Communication Efficiency
In hardware experiments, both PT and ST successfully stabilized the unstable cart-pole system while utilizing only ~20% of the communication required by traditional periodic systems. The sparse nature of these communication events is visualized through the triggering signal and the resulting decision variable $\gamma_k$.
Robustness to Disturbances
A key insight emerged from the vehicle platoon simulation during sudden braking maneuvers. While Self-Triggering (ST) is "blind" between scheduled updates and can fail to react to sudden disturbances, Predictive Triggering (PT) continuously monitors the $M$-step horizon. This allows PT to detect unexpected changes (like a lead car braking) and preemptively request bandwidth, maintaining safety where ST-based systems might collide.
Discussion and Future Work
Predictive triggering represents a vital middle ground in IoT control, combining the real-time responsiveness of event-triggering with the proactive lead time of self-triggering. The authors observed a "phase transition" in PT behavior: when communication is cheap, it behaves like a variance-driven self-trigger; when expensive, it acts as an error-driven event-trigger.
Future research will focus on extending these linear Gaussian derivations to non-linear dynamics and integrating the framework with specific low-power wireless protocols to evaluate real-time resource reservation in complex network environments.
Summary updated