Threshold Based Transmission Policies for Remote Estimation

234 concepts149 papers2 proposals

In networked control systems and the Internet of Things (IoT), physical processes are monitored by remote sensors that send status updates across communication channels. Traditional systems rely on periodic sampling, where updates are transmitted at fixed time intervals. However, when communication channels are congested, noisy, or powered by limited batteries, periodic updates waste valuable resources by transmitting redundant information when the process is stagnant, while failing to capture sudden, critical variations.

To resolve this tension, researchers have developed threshold-based transmission policies. Instead of asking when to transmit based on a clock, these policies ask whether to transmit based on the state of the system. By mapping continuous state spaces into discrete, event-triggered decisions, threshold policies optimize the fundamental trade-off between communication cost and estimation accuracy.

The Core Intuition: Error-Triggered Sampling

The fundamental reason threshold-based policies work so efficiently is that they track the innovation—the discrepancy between the actual current state of a process and the remote estimator's current guess.

Imagine a sensor tracking a continuous stochastic process. The sensor knows the exact state of the process, and it also knows exactly what the remote receiver thinks the state is (by simulating the receiver's estimation filter locally). The local estimation error grows as the physical process drifts. Under an optimal threshold policy, the transmitter remains completely silent until this local error crosses a critical boundary.

State / Error
  ^
  |      /\             / \   <-- Physical Process State
  |---- /--\-----------/---\------- Upper Threshold
  |    /    \  Silence/     \
  |   /      \       /       \
  |--/--------\-----/---------\---- Lower Threshold
  | /          \   /           \
  +----------------------------------> Time
               ^               ^
         Transmission    Transmission

As shown by Jhelum Chakravorty et al. in Fundamental limits of remote estimation of autoregressive Markov processes under communication constraints, a fixed threshold on this local estimation error is mathematically optimal for tracking autoregressive Markov processes, significantly reducing communication frequency compared to periodic sampling. This error-triggered mechanism ensures that communication resources are strictly preserved for instances where the signal has changed enough to provide significant new information.

This principle holds across diverse stochastic processes. For instance, Yin Sun et al. demonstrated in Remote Estimation of the Wiener Process over a Channel with Random Delay that when tracking a Wiener process over a channel with random queueing delays, the optimal strategy is a threshold policy. Even when the channel becomes idle, the transmitter often waits until the signal drifts past the threshold to justify the cost of a new transmission. Similarly, for mean-reverting processes, Yuchao Chen et al. framed the tracking of an Ornstein-Uhlenbeck process in Sampling for remote estimation of an Ornstein-Uhlenbeck process through channel with unknown delay statistics as an optimal stopping problem, proving that updates should only be triggered when the expected reduction in mean square error outweighs the transmission delay.

The Power of Silence: Implicit Information

The true elegance of a threshold-based policy lies in a phenomenon known as implicit information: silence itself carries meaning.

In a periodic system, if a packet is not received, the estimator learns nothing new and its uncertainty grows. In a threshold-based system, if the transmitter is silent, it communicates a powerful mathematical guarantee: the local estimation error is currently trapped within the threshold boundaries.

This structural property changes how the remote estimator operates. Instead of guessing blindly between updates, the receiver updates its internal belief state based on the fact that the transmitter chose not to send a packet.

Edoardo David Santi et al. explored this in Remote Estimation of Markov Processes over Costly Channels: On Implicit Information Benefits, showing that communication policies that actively reason about the receiver's internal belief state achieve a superior trade-off by transmitting only when the receiver's probability of error exceeds a threshold.

Even when the communication channel introduces noise, this principle persists. Xiaobin Gao et al. demonstrated in On Communication Scheduling and Remote Estimation in the Presence of an Adversary as a Nonzero-sum Game that a symmetric threshold policy is structurally optimal because the clean, bounded information implied by a sensor's silence can be more valuable to the estimator than a noise-corrupted measurement. In Optimal communication scheduling and remote estimation over an additive noise channel, the authors further proved that for a broad range of symmetric and log-concave data distributions, the optimal sensor strategy is a symmetric threshold policy that completely avoids transmissions for data close to zero, maximizing the utility of limited channel capacity.

State-Dependent and Resource-Aware Thresholds

While a fixed threshold works well in static environments, real-world sensors often operate under fluctuating resource constraints, such as time-varying channels or energy-harvesting batteries. To maintain optimality, the transmission threshold must dynamically adapt to the internal state of the device.

Battery-Driven Adaptation

For sensors powered by solar or ambient energy, the threshold acts as a valve that regulates energy expenditure against battery depletion. When the battery is low, the sensor must conserve power, so it raises its threshold, becoming highly selective and only transmitting massive innovations. When the battery is full, it lowers the threshold to prevent incoming harvested energy from overflowing and going to waste.

Ashutosh Nayyar et al. formalized this in Optimal Strategies for Communication and Remote Estimation With an Energy Harvesting Sensor, proving that the optimal policy uses a dynamic threshold that decreases monotonically as battery levels increase. In Data-Driven Stochastic Models and Policies for Energy Harvesting Sensor Communications, Meng-Lin Ku et al. highlighted that this state-dependent threshold structure simplifies real-time hardware implementation, replacing complex multi-dimensional look-up tables with basic conditional logic based on current energy reserves.

Channel-State Adaptation

When the communication channel has memory (e.g., fading or packet drops modeled as a Markov chain), the threshold must adapt to the likelihood of transmission success. If the channel is in a deep fade, attempting to transmit is likely to waste energy without updating the receiver.

Jhelum Chakravorty et al. showed in Remote estimation over a packet-drop channel with Markovian state that the optimal transmission policy remains threshold-based, but the exact error limit shifts dynamically based on the current channel quality. Aditya Mahajan et al. confirmed this in Structure of optimal strategies for remote estimation over Gilbert-Elliott channel with feedback, proving that conditioning the threshold on the previous state of a bursty channel allows the sensor to automatically tolerate higher local errors when the channel is poor, waiting for favorable conditions to maximize energy efficiency.

Queue-Aware and Buffer-Driven Thresholds

Threshold policies are not limited to physical tracking errors; they are equally powerful when applied to data queues and network buffers to optimize transmission latency.

In buffered communication systems, the transmission rate or power is adjusted based on the number of packets waiting in the queue. To minimize average delay under a power constraint, the system uses discrete queue-occupancy thresholds. As the queue fills up and crosses a threshold, the transmitter steps up its transmission rate to flush the buffer before latency penalties accumulate.

Xiaoyu Zhao et al. analyzed this mechanism in Delay-Optimal and Energy-Efficient Communications with Markovian Arrivals, demonstrating that the optimal transmission rate is a monotonic function of queue length. Crucially, these thresholds are ordered such that better channel conditions lower the queue-length triggers for higher transmission rates, ensuring the system prioritizes high-throughput transmission during favorable channel states. This matches the structural properties derived in Delay-Optimal Buffer-Aware Scheduling with Adaptive Transmission, where the optimal scheduling policy increases data throughput only when the buffer occupancy reaches specific discrete levels.

This queue-threshold framework generalizes well to broader scheduling problems:

Quantifying the Value of Information

To calibrate the exact numerical value of a threshold, researchers must define what makes an update "worth it." This requires moving beyond traditional metrics to formalize the Value of Information (VoI).

Touraj Soleymani et al. operationalized this in Value of Information in Minimum-Rate LQG Control, proposing an event-driven sampling policy where data is transmitted only when its calculated VoI—the immediate reduction in control performance penalty—surpasses a predefined threshold. When a packet drop occurs, the authors showed in Transmit or Retransmit: A Tradeoff in Networked Control of Dynamical Processes Over Lossy Channels With Ideal Feedback that a threshold switching policy evaluates the residual value of information to dynamically choose between transmitting a fresh sensor measurement or retransmitting old data.

Alternatively, when tracking state correctness over time, Ali Maatouk et al. introduced a metric called the Age of Incorrect Information (AoII) in The Age of Incorrect Information: A New Performance Metric for Status Updates. Under strict power constraints, the authors proved that the optimal policy keeps the transmitter completely idle until the accumulated penalty for incorrect information crosses a specific numerical threshold, at which point an update is triggered to resynchronize the receiver.

By shifting from rigid, clock-driven schedules to flexible, threshold-driven decisions, modern remote estimation systems achieve mathematical optimality. They exploit the informational value of silence, adapt dynamically to internal energy and queue states, and ensure that every single transmission directly justifies its resource cost.

History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_threshold_based_transmission_policy_d3b02936,
  title = {threshold-based transmission policy},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/threshold-based-transmission-policy/rdr_d3b029369773467cba46b0cff697cae3},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_threshold_based_transmission_policy_d3b02936_v1,
  title = {threshold-based transmission policy},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/threshold-based-transmission-policy/rdr_d3b029369773467cba46b0cff697cae3/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake