Performance implications of different $p$-norms in level-triggered sampling

David Meister, Frank Allgöwer

Abstract

This work studies the performance of an event-based control approach, namely level-triggered sampling, in a standard multidimensional single-integrator setup. We falsify a conjecture from the literature that the deployed $p$-norm in the triggering condition supposedly has no impact on the performance of the sampling scheme in that setting. In particular, we show for the considered setup that the usage of the maximum norm instead of the Euclidean norm induces a performance deterioration of level-triggered sampling for sufficiently large system dimensions, when compared to periodic control at the same sampling rate. Moreover, we investigate the performance for other $p$-norms in simulation and observe that it degrades with increasing $p$. In addition, our findings reveal the previously unknown role of the triggering rule in the cause of a recently discovered phenomenon: Previous work has shown for a single-integrator consensus setup that the commonly observed performance advantage of event-based control over periodic control can be lost in distributed settings with a cooperative control goal. In our work, we obtain similar results for a non-cooperative setting only by adjusting the norm in the level-triggered sampling scheme. We therefore demonstrate that the performance degradation found in the distributed setting originates from the triggering rule and not from the considered cooperative control goal.

Introduction

In modern networked control systems, Event-Based Control (EBC) is frequently employed to optimize resource utilization, such as energy and bandwidth. A common implementation is level-triggered sampling, where a control action is initiated only when the system state deviates beyond a predefined threshold. Historically, it was conjectured that the specific $p$-norm used to define this threshold—whether the Euclidean norm ($p=2$) or the maximum norm ($p=\infty$)—had negligible impact on performance in multidimensional single-integrator systems. This paper rigorously examines and falsifies this assumption, demonstrating that the choice of norm is a critical design parameter.

Methodology

The study focuses on an $n$-dimensional single-integrator system where the state drifts due to white noise and is reset to zero upon sampling. The performance of level-triggered sampling, defined by the condition $\|x(t)\|_p \geq \Delta_p$, is compared against standard periodic control.

Analytical derivations for the $p=2$ and $p=\infty$ cases were performed using stochastic calculus tools, including Dynkin’s formula and Bessel process properties. To explore intermediate norms ($p=3, 5, 8, 10$), the authors conducted Monte Carlo simulations with 20,000 runs per experiment, utilizing the Euler-Maruyama method with a high-precision step size of $10^{-4}$ s.

Performance Analysis and Key Results

The core finding of this research is that the Euclidean norm ($p=2$) consistently outperforms periodic sampling, while higher-order norms lead to performance degradation. For the $2$-norm, the cost ratio relative to periodic control is analytically determined to be $n/(n+2)$, which remains strictly less than 1 for all dimensions $n$.

Conversely, the maximum norm ($p=\infty$) exhibits a "curse of dimensionality." As the system dimension $n$ increases, the performance advantage of EBC diminishes. Analytical proofs and simulations show that for $n \gtrsim 40$, periodic control actually becomes more efficient than level-triggered sampling using the $\infty$-norm.

Figure 1: Cost ratio $J_p / J_{per}$ as a function of state dimension $n$. The $2$-norm (red) remains consistent, while the $\infty$-norm (black) and $8$-norm (blue) eventually exceed the periodic control threshold (ratio > 1).

Further investigation into intermediate norms reveals a monotonic degradation in performance as $p$ increases. Figure 2 illustrates the cost ratio relative to the optimal $2$-norm, highlighting that any deviation from $p=2$ towards higher $p$ values results in increased control costs.

Figure 2: Cost ratio $J_p / J_2$ versus state dimension $n$, demonstrating that performance degrades progressively as the norm index $p$ increases.

Implications for Distributed Systems

These results provide a new perspective on the performance loss often observed in distributed multi-agent systems. In such settings, decentralized triggering rules often implicitly resemble an $\infty$-norm (e.g., triggering if any agent's local error exceeds a bound). This work suggests that the performance degradation in these systems is a consequence of the triggering mathematics rather than the cooperative nature of the control task itself.

Conclusion

This study establishes that the choice of $p$-norm in level-triggered sampling is vital for maintaining the efficiency of event-based control. While the $2$-norm is optimal and consistent, higher $p$-norms can lead to performance that is worse than simple periodic sampling in high-dimensional systems. Future work aims to provide a general analytical proof for the monotonic degradation across all $p > 2$ and to extend these findings to higher-order dynamical systems.

Summary updated

Abstract

This work studies the performance of an event-based control approach, namely level-triggered sampling, in a standard multidimensional single-integrator setup. We falsify a conjecture from the literature that the deployed $p$-norm in the triggering condition supposedly has no impact on the performance of the sampling scheme in that setting. In particular, we show for the considered setup that the usage of the maximum norm instead of the Euclidean norm induces a performance deterioration of level-triggered sampling for sufficiently large system dimensions, when compared to periodic control at the same sampling rate. Moreover, we investigate the performance for other $p$-norms in simulation and observe that it degrades with increasing $p$. In addition, our findings reveal the previously unknown role of the triggering rule in the cause of a recently discovered phenomenon: Previous work has shown for a single-integrator consensus setup that the commonly observed performance advantage of event-based control over periodic control can be lost in distributed settings with a cooperative control goal. In our work, we obtain similar results for a non-cooperative setting only by adjusting the norm in the level-triggered sampling scheme. We therefore demonstrate that the performance degradation found in the distributed setting originates from the triggering rule and not from the considered cooperative control goal.
BibTeX
@misc{meister2023performance,
  author = {David Meister and Frank Allgöwer},
  title = {{Performance implications of different $p$-norms in level-triggered sampling}},
  year = {2023},
  eprint = {2309.03804},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2309.03804},
}
Figure