Fault detection, isolation, and recovery for autonomous parafoils

Abstract

Autonomous precision airdrop systems are widely used to deliver supplies to remote locations. This aerial delivery method provides a safety and logistical advantage over traditional ground- or helicopter-based payload transportation methods. The occurrence of a fault during a flight can severely degrade vehicle performance, effectively nullifying the value of the guided system, or worse. Quickly detecting and identifying faults enables the choice of an appropriate recovery strategy, potentially mitigating the consequences of an out-of-control vehicle and recovering performance. This thesis presents a fault detection, isolation, and recovery (FDIR) method for an autonomous parafoil system. The detection and isolation processes use residual signals generated from observers and other system models. Statistical methods are applied to evaluate these residuals and determine whether a fault has occurred, given a priori knowledge of how the system behaves in the presence of faults. This work develops fault recovery strategies that are designed to mitigate the effects of several common faults and allow for a successful mission even with severe loss of control authority. An extensive, high-fidelity, Monte Carlo simulation study is used to assess the eectiveness of FDIR, including the probability of correctly isolating a fault as well as the target miss distance improvement resulting from the implementation of fault recovery strategies. The integrated FDIR method demonstrates a very high percentage of successful isolation as well as a substantial decrease in miss distance for cases in which a common fault occurs. Flight test results consistent with simulations show successful detection and isolation of faults as well as implementation of recovery strategies that result in miss distances comparable to those from healthy flights.