Reconfigurable LQR Framework for Optimality-Preserving Fault-Tolerant Control

When actuators lose control effectiveness due to<br/>mechanical damage or reduced motor efficiency, rotary-wing<br/>unmanned aerial vehicles (UAVs) become more prone to<br/>accidents. Fault-tolerant control strategies for addressing<br/>such faults are broadly classified into Passive Fault-Tolerant<br/>Control Systems (PFTCS) and Active Fault-Tolerant Control<br/>Systems (AFTCS). While PFTCS are designed to handle<br/>predefined fault scenarios, AFTCS actively estimate faults<br/>online and reconfigure the control system to recover desired<br/>performance. Linear Quadratic Regulator (LQR) control<br/>is widely employed within AFTCS frameworks. However,<br/>experimental evidence reported in the literature indicates<br/>that, following actuator loss of effectiveness, reconfiguring<br/>the system model using estimated fault parameters alone is<br/>insufficient for restoring pre-fault performance. This limitation<br/>arises because the LQR control-weighting matrix R, designed<br/>for the healthy system, no longer represents the true cost<br/>of actuator usage under fault conditions. Although several<br/>studies have proposed reconfiguring the R-matrix to improve<br/>fault accommodation, these approaches do not guarantee<br/>preservation of the pre-fault LQR optimality. In this work,<br/>a systematic method for reconfiguring the LQR R-matrix<br/>that preserves pre-fault optimality is proposed. In addition,<br/>a reconfigurable feedforward control scheme is introduced to<br/>enhance reference-tracking accuracy under actuator faults.<br/>The proposed approach is validated on a two-degree-offreedom helicopter platform through both simulation and<br/>hardware-in-the-loop experiments, demonstrating satisfactory<br/>fault-tolerant performance.