Optimal control of linear systems with limited control actions: threshold-based event-triggered control

Burak Demirel, Euhanna Ghadimi, Daniel E. Quevedo, Mikael Johansson

Abstract

We consider a finite-horizon linear-quadratic optimal control problem where only a limited number of control messages are allowed for sending from the controller to the actuator. To restrict the number of control actions computed and transmitted by the controller, we employ a threshold-based event-triggering mechanism that decides whether or not a control message needs to be calculated and delivered. Due to the nature of threshold-based event-triggering algorithms, finding the optimal control sequence requires minimizing a quadratic cost function over a non-convex domain. In this paper, we firstly provide an exact solution to the non-convex problem mentioned above by solving an exponential number of quadratic programs. To reduce computational complexity, we, then, propose two efficient heuristic algorithms based on greedy search and the Alternating Direction Method of Multipliers (ADMM) method. Later, we consider a receding horizon control strategy for linear systems controlled by event-triggered controllers, and we also provide a complete stability analysis of receding horizon control that uses finite horizon optimization in the proposed class. Numerical examples testify to the viability of the presented design technique.

Introduction

In classical control theory, the controller is typically assumed to communicate with the actuator at every time step. However, in modern applications like remote drones or wireless sensor networks, communication is a scarce resource. Event-Triggered Control (ETC) addresses this by only sending updates when necessary—specifically, when the system state deviates significantly from the target. While efficient, this threshold-based triggering introduces a circular dependency: the decision to send a signal depends on the state, which itself depends on previous signals. This transforms a standard convex Linear Quadratic Regulator (LQR) problem into a non-convex, NP-complete optimization challenge.

Geometry and Exact Solution

The core insight of this work is the characterization of the non-convex state space. The $L_\infty$ threshold (the "dead-zone" around the origin) can be managed by partitioning the space into a finite union of $2n+1$ convex polyhedra.

Partitioning of the state space into convex polyhedra $\mathcal{C}_i$ to handle non-convexity.

By checking every possible sequence of these regions over a time horizon $N$, the global optimum can be found via exhaustive search. However, this requires solving an exponential number of Quadratic Programs (QPs), making it computationally prohibitive for long horizons.

Heuristic Optimization Methods

To overcome the complexity of the exact solution, two heuristics are proposed:

  1. Greedy Heuristic: This algorithm builds the control sequence step-by-step, solving a small set of QPs at each interval while keeping previous decisions fixed.
  2. ADMM-based Heuristic: The Alternating Direction Method of Multipliers (ADMM) is adapted to handle non-convex triggering constraints using an indicator function. A final "polishing" step ensures the resulting sequence is both feasible and high-quality.

These methods are further extended into a Receding Horizon Control (MPC) framework, which guarantees "practical stability"—ensuring the system state remains within a bounded region around the target.

Key Results and Performance

Numerical simulations on second and third-order plants demonstrate that the event-triggered approach reduces communication by 68% to 72% compared to traditional periodic control. The ADMM-based heuristic proved remarkably accurate, finding near-optimal solutions (less than 5% relative error) in the vast majority of test cases.

Comparison of state trajectories and control signals between exhaustive search and the ADMM heuristic.

As shown in the comparison above, the ADMM heuristic closely tracks the exhaustive search results while operating in polynomial time, making it suitable for real-time applications where solving thousands of QPs is impossible.

Conclusion

This paper provides a robust framework for optimal control under communication constraints. By leveraging the geometric structure of the triggering law, the authors provide both an exact solution and efficient heuristics. Future work will focus on extending these methods to handle significant stochastic noise and automating the tuning of ADMM penalty parameters.

Summary updated

Abstract

We consider a finite-horizon linear-quadratic optimal control problem where only a limited number of control messages are allowed for sending from the controller to the actuator. To restrict the number of control actions computed and transmitted by the controller, we employ a threshold-based event-triggering mechanism that decides whether or not a control message needs to be calculated and delivered. Due to the nature of threshold-based event-triggering algorithms, finding the optimal control sequence requires minimizing a quadratic cost function over a non-convex domain. In this paper, we firstly provide an exact solution to the non-convex problem mentioned above by solving an exponential number of quadratic programs. To reduce computational complexity, we, then, propose two efficient heuristic algorithms based on greedy search and the Alternating Direction Method of Multipliers (ADMM) method. Later, we consider a receding horizon control strategy for linear systems controlled by event-triggered controllers, and we also provide a complete stability analysis of receding horizon control that uses finite horizon optimization in the proposed class. Numerical examples testify to the viability of the presented design technique.
BibTeX
@article{demirel2018optimal,
  author = {Burak Demirel and Euhanna Ghadimi and Daniel E. Quevedo and Mikael Johansson},
  title = {{Optimal control of linear systems with limited control actions: threshold-based event-triggered control}},
  journal = {IEEE Trans. Control. Netw. Syst.},
  year = {2018},
  url = {https://openreview.net/forum?id=fCAk2Osx8G3},
}
Figure