Abstract
This paper investigates the optimal local and remote linear quadratic (LQ) control problem for a networked control system (NCS) comprising multiple subsystems. Each subsystem is modeled as a general multiplicative noise stochastic system equipped with both a local and a remote controller. A key challenge is the asymmetric information structure: while downlink channels are perfect, unreliable uplink channels mean the remote controller only receives partial state information. By applying the Pontryagin maximum principle and developing a novel method to decouple the resulting group of forward and backward difference equations (G-FBSDEs), the authors derive necessary and sufficient solvability conditions. They introduce new coupled Riccati equations (CREs) and verify that the separation principle holds for these complex systems, providing a significant contribution to optimal control theory with asymmetric information.
System Architecture and Problem Statement
In modern industrial and aerospace applications—ranging from smart grids to drone swarms—control is often distributed between local units and a centralized remote hub. This paper addresses three critical hurdles in managing such Networked Control Systems (NCS):
- Asymmetric Information: Local controllers have full state visibility, but the "uplink" (local-to-remote) channel is unreliable. Consequently, the remote controller operates with a degraded, incomplete view of the system.
- Multiplicative Noise: Unlike standard additive noise, multiplicative noise scales with the system's state or control effort, a common feature in high-precision aerospace modeling that complicates standard control laws.
- Coupling of Estimation and Control: The presence of both multiplicative noise and packet loss typically causes the "separation principle" to fail, meaning estimators and controllers cannot be designed independently.
Methodology
The authors utilize a rigorous mathematical framework to solve the optimal control problem:
- Pontryagin Maximum Principle: This tool is used to define the optimality conditions.
- Decoupling G-FBSDEs: The optimization process yields a complex set of forward and backward difference equations. The authors' primary innovation is a decoupling strategy that uses specific information filtrations to make these equations solvable.
- Asymmetric Coupled Riccati Equations (CREs): To handle the information imbalance, the authors introduce asymmetric versions of the standard Riccati equations, which are fundamental to determining the optimal control gains.
Key Results and Simulation
The most significant finding is the verification that the separation principle holds for multiplicative noise NCSs with packet dropouts, provided the proposed CREs are used. This allows engineers to maintain a modular design approach even in these highly uncertain environments.
The theoretical results were validated through numerical simulations of three subsystems over 60 time steps. The simulations confirmed that the derived optimal controllers effectively drive system states to zero, ensuring stability.
As shown in the simulation for $p=0.8$ (20% dropout), the system stabilizes quickly. The authors noted that lower reliability (e.g., $p=0.3$) results in significantly more "jitter" and slower convergence, though the system remains stable under their optimal control strategy. This demonstrates the robustness of the proposed framework against varying levels of network unreliability.
Conclusion
This research provides a generalized framework for the optimal control of multiple systems with asymmetric information and multiplicative noise. While the current model assumes a perfect downlink, it sets the stage for future work involving unreliable bidirectional communication and infinite-horizon control problems. The findings demonstrate that even with significant communication gaps, optimal performance is achievable through sophisticated mathematical decoupling.
Summary updated