Information-Theoretic Dual Adaptive Control Revisited: Multivariable Extension with Applications to Fault-Tolerant Control

TL;DR

This paper extends an information-based control framework to handle complex systems with multiple inputs and outputs, improving fault tolerance in real-time.

Contribution

The paper introduces a multivariable extension of dual adaptive control that preserves parameter learning and performance balance in fault-tolerant systems.

Findings

A convexity condition is derived for MIMO systems, generalizing earlier single-input results.

The framework handles actuator faults using effectiveness factors and information gain monitoring.

Simulations show the method balances fault detection, identification accuracy, and real-time performance.

Abstract

This paper revisits and extends the information-theoretic dual adaptive control framework initially developed by the author for single-input single-output systems to multiple-input multiple-output (MIMO) systems, with specific application to fault-tolerant control (FTC). The core contribution is a MIMO formulation that preserves the essential dual property, i.e., balancing control performance against parameter learning, while addressing the increased complexity of coupled multivariable systems. A convexity condition is derived for the MIMO optimization problem, generalizing the original SISO condition. The framework naturally handles actuator faults through a parameter vector that includes effectiveness factors, with fault detection achieved via monitoring of information gain. Control reconfiguration strategies ensure graceful performance degradation under faults. Simulation results demonstrate the effectiveness of this dual approach to FTC methods in balancing detection speed, identification accuracy, and tracking performance, while maintaining computational feasibility for real-time implementation.

Figures