Robust Switching Time Optimization for Networked Switched Systems via Model Predictive Control.

Area of Science:

• Control Systems Engineering
• Networked Systems
• Optimization Theory

Background:

• Networked switched systems are prevalent in modern control applications.
• Uncertainties in these systems pose significant challenges for optimal control.
• Existing methods often struggle with real-time optimization of switching sequences.

Purpose of the Study:

• To develop a model predictive control (MPC) strategy for optimizing switching time sequences in uncertain networked switched systems.
• To design an efficient computational framework for solving the large-scale MPC problem.
• To enable real-time optimal control of system switching behaviors.

Main Methods:

• Formulation of a large-scale MPC problem based on predicted trajectories and exact discretization.
• Development of a two-level hierarchical optimization structure using a recurrent neural network (RNN).
• Implementation of a local compensation mechanism and a real-time switching time optimization algorithm.

Main Results:

• The proposed hierarchical optimization structure effectively solves the formulated MPC problem.
• The recurrent neural network, comprising a coordination unit and local optimization units, provides a robust framework.
• The real-time algorithm successfully calculates optimal switching time sequences for uncertain systems.

Conclusions:

• The presented MPC strategy offers an effective approach for optimal control of networked switched systems with uncertainties.
• The novel hierarchical optimization using RNNs demonstrates superior performance in real-time applications.
• This work contributes to advancing the field of robust and efficient control system design.