In traditional digital systems, we typically sample data like clockwork—every millisecond or every second. This "Riemann sampling" is predictable, but it is often incredibly wasteful. If a sensor is monitoring a room where nothing is moving, sending a constant stream of "zero" values consumes battery and bandwidth for no gain.
The alternative is event-triggered communication, where a sensor only transmits data when something "meaningful" happens. While this saves energy, it introduces a profound mathematical shift: the absence of a message is no longer just missing data. Instead, silence becomes a structured signal.
The Signaling Effect of Silence
In an event-triggered system, the sensor (the encoder) and the remote monitor (the decoder) share a piece of logic: the triggering rule. A common rule is the threshold policy: "Only send an update if the difference between the current state and the last reported value exceeds $\delta$."
As Ashutosh Nayyar, Demosthenis Teneketzis, and Venugopal V. Veeravalli demonstrate in Optimal Strategies for Communication and Remote Estimation with an Energy Harvesting Sensor, this creates a "signaling effect." When the remote estimator receives nothing, it doesn't just hold the old value; it infers that the state must be within the range $[-\delta, \delta]$.
This "negative information" allows the estimator to refine its belief without a single bit being transmitted. In fact, Xiaobin Gao and Emrah Akyol show in Optimal Communication Scheduling and Remote Estimation over an Additive Noise Channel that in high-noise environments, this "information of silence" can actually be more accurate than a received data packet that has been corrupted by channel interference.
Joint Design of Trigger and Estimator
To fully exploit silence, the trigger and the estimator cannot be designed in isolation. If the estimator is "oblivious"—treating silence as a total lack of information—it misses the chance to reduce uncertainty.
Adam Molin argues in Event-Triggered Estimation of Linear Systems: An Iterative Algorithm and Optimality Properties that joint optimization is essential. When the trigger and estimator are coordinated, the system can handle non-standard noise (like bimodal distributions) by using silence to "trap" the state estimate within a specific region.
This coordination is the core of "semantic" or "pragmatic" communication. As explored by Touraj Soleymani and John S. Baras in Foundations of Value of Information, we move away from Shannon’s goal of reconstructing every bit. Instead, we focus on the Value of Information (VoI): we only transmit if the "innovation" (the new info) is valuable enough to justify the cost of the battery or the risk of network congestion.
Applications Across Domains
The power of implicit information spans from low-level sensor hardware to high-level AI coordination:
- Edge Analytics: Christos Anagnostopoulos and Dimitrios Pezaros describe "predictive intelligence" at the edge, where sensors only report deviations from a local model. The edge node uses the absence of data to confirm that its local forecast is still accurate, slashing traffic by up to 80% (Predictive intelligence to the edge: impact on edge analytics).
- Privacy-Preserving Traffic Data: John Krumm and Eric Horvitz show that by only transmitting traffic updates when they deviate from historical norms, a system can maintain high-fidelity maps while revealing far less about individual drivers' movements (Traffic Updates: Saying a Lot While Revealing a Little).
- Multi-Agent Teams: In the OVERSEER system, Milind Tambe and Gal A. Kaminka use "overhearing" to monitor teams. If an agent doesn't broadcast a "task complete" message, the monitor uses that silence to infer that a specific plan transition has not yet occurred (Monitoring Teams by Overhearing: A Multi-Agent Plan-Recognition Approach).
The Challenge of Ambiguity
The primary hurdle in these systems is distinguishing between intentional silence and unintentional failure. If a sensor is silent, is it because the state is stable, or because the packet was lost in the network?
Sebastian Trimpe highlights this in Event‐based state estimation: an emulation‐based approach, noting that estimators must be robust enough to handle this ambiguity. Recent work like the CALM framework by Shubham Aggarwal and Dipankar Maity uses deep reinforcement learning to help schedulers and estimators "learn" how to coordinate through these gaps, turning silence into a reliable, zero-cost communication channel (The Silence that Speaks: Neural Estimation via Communication Gaps).