Resilience-enhanced control reconfiguration for autonomous systems
MetadataShow full item record
Unmanned systems keep replacing manned systems as a paradigm shift. According to the Unmanned Autonomous Systems (UAS) market forecast reports, the UAS market value is expected to grow two to three times higher in ten years. Considering the economic impacts of UAS application in job markets and component manufacturing industries, the UAS market value may very well exceed, which is predicted in the reports. However, regulations have limited the effective utilization of UAS due to safety concerns. These restrictive regulations significantly delay the potential usefulness of civilian and commercial UAVs. According to the Unmanned Aerial Vehicle (UAV) incidence reports, mechanical failures come out to be one of the top reasons for the incidents except for human errors. Technically, it is impossible to avoid any fault or failure in any systems. However, it can be possible to save the faulty system if the faults are treated properly. In this regard, this research has reviewed the state-of-the-art techniques regarding system safety improvement in the presence of a critical fault mode. Promising concepts are resilience engineering and Active Fault Tolerant Control (AFTCS) systems. Resilience engineering has been more focus on system design and resilience assessment methods. AFTCS mainly contributes to the fast and stable operating point recovery without the consideration of long-term system performances or mission success. Prognostics-enhanced reconfigurable control frameworks have proposed the online prognosis for a Remaining Useful Life (RUL) prediction within the control scheme but do not address comprehensive mission capability trade-offs. The objective of this study is to design a resilience-enhanced reconfigurable control framework for unmanned autonomous systems in the presence of a critical fault mode during the operation. The proposed resilience-enhanced reconfigurable control framework is composed of three fundamental modules: 1) immediate performance recovery by Model Predictive Control (MPC) and Differential Dynamic Programming (DDP) approaches, 2) long-term mission capability trade-offs by an optimization routine, and 3) situational awareness by a particle filtering-based fault diagnosis and Case-Based Reasoning (CBR). Critical development of this thesis is an introduction of an adaptation parameter in an MPC formulation (Module 1) and optimization process to find an optimal value for the adaptation parameter (Module 2). Module 3 enables long-term mission capability reasoning when a new fault growth pattern is observed. In order to test the efficacy of the proposed framework, under-actuated hovercraft as a testbed and an insulation degradation of an electrical thrust motor as a critical fault mode are introduced. The experiments explore the effect of the adaptation parameter on long-term mission capabilities and identify the necessity of the proper trade-offs. Further experiments investigate the efficacy of each module and the integrated framework. The experiment results show that the adaptation parameter adjusts a control strategy, so that mission capabilities are optimized while vulnerable long-term mission capabilities are recovered. The integrated framework presents the improvement to the probability of mission success in the presence of a critical fault mode. Lastly, as a generalization of the design process for the resilience-enhanced reconfigurable control framework, a design methodology suggests a step-by-step design procedure. Assumptions of the research have guided the required steps and limitations of the proposed framework.