Abstract
This report investigates the optimal design of event-triggered estimation for first-order linear stochastic systems. The problem is posed as a two-player team problem with a partially nested information pattern. The two players are given by an estimator and an event-trigger. The event-trigger has full state information and decides whether the estimator shall obtain the current state information by transmitting it through a resource-constrained channel. The objective is to find an optimal trade-off between the mean squared estimation error and the expected transmission rate. The proposed iterative algorithm alternates between optimizing one player while fixing the other player. It is shown that the solution of the algorithm converges to a linear predictor and a symmetric threshold policy if the densities of the initial state and the noise variables are even and radially decreasing functions. The effectiveness of the approach is illustrated on a numerical example. In the case of a multimodal distribution of the noise variables, a significant performance improvement can be achieved compared to a separate design that assumes a linear prediction and a symmetric threshold policy.
Introduction
In networked control systems—such as autonomous robot swarms or distributed power grids—communication is a finite and often sparse resource. Continuous data transmission (periodic sampling) can lead to unnecessary battery drain and network congestion. Event-triggered estimation addresses this by transmitting data only when a specific condition is met, such as when the system state deviates significantly from its predicted value.
Traditionally, the event-trigger (the "when") and the estimator (the "how") are designed independently. Standard designs often assume a simple linear predictor for the estimator and a symmetric threshold for the trigger. However, this decoupling ignores the fact that the trigger's decision-making process inherently shapes the information available to the estimator. This paper explores the joint design problem, seeking to optimize both components simultaneously to achieve a superior trade-off between estimation accuracy and communication costs.
Methodology
The problem is modeled as a two-player team problem with a partially nested information pattern. The Event-Trigger ($\mathcal{E}$) monitors the full state $x_k$ and generates a binary decision $\delta_k$ to transmit or remain silent. The Estimator ($\mathcal{S}$) receives the transmitted data $z_k$ (or an erasure symbol) and produces an estimate $\hat{x}_k$.
To find the optimal configuration, the authors propose an iterative scheme (Algorithm 1) that alternates between optimizing each player while holding the other constant:
- Estimator Optimization: Given a fixed trigger policy, the optimal estimator is derived using a least-squares approach.
- Trigger Optimization: Given a fixed estimator, the optimal triggering policy is found using dynamic programming (Bellman equations).
By treating this iterative process as a discrete-time dynamical system, the authors utilize Lyapunov stability theory to prove convergence properties, particularly for symmetric and unimodal noise distributions.
Key Insights and Results
The research challenges the "independence myth"—the idea that separate design is always sufficient. For standard unimodal (bell-shaped) symmetric noise, the iterative algorithm indeed converges to the standard independent design. In these cases, the linear predictor and symmetric threshold are optimal.
However, the joint design provides a significant advantage when dealing with multimodal noise. For bimodal distributions (where noise is likely to push the state toward one of two distinct peaks), the optimal trigger becomes asymmetric, and the estimator becomes biased to account for the implicit information contained in "silence."
Numerical simulations demonstrate that as the noise becomes increasingly bimodal (represented by $\mu \to 1$), the joint iterative design can reduce the total cost (a combination of estimation error and communication penalty) by up to 45% compared to traditional symmetric designs. This performance gain stems from the estimator's ability to "infer" which peak the state is likely near even when no data is transmitted, provided the trigger and estimator are tuned to each other.
Conclusion
This work demonstrates that joint design is crucial for optimizing networked systems operating under non-Gaussian or multimodal noise conditions. While the current study is limited to first-order scalar systems, it provides a rigorous theoretical foundation for extending these methods to more complex multidimensional systems and sensor networks. Future research will focus on overcoming the computational "curse of dimensionality" in larger systems and exploring coordination strategies for multiple sensors sharing a central estimator.
Summary updated