Extraterrestrial habitats involve a tightly coupled combination of hardware, software, and humans while operating in an unforgiving environment that poses many risks, both anticipated and unanticipated. Traditional approaches with such systems of systems focus on reliability, robustness, and redundancy. These approaches seek to avoid failure rather than reduce overall risk. However, faults are inevitable, and understanding and managing the complex and emergent behavior and cascading events of such a complex system is critical. This study describes the development of HabSim, a computational simulation environment intended to support research to establish the know-how to design and operate resilient and autonomous SmartHabs. HabSim is a modular virtual testbed composed of many of the coupled dynamic systems expected in a typical SmartHab. A heterogeneous set of interconnected physics-based and phenomenological models is used to represent the essential functions of a SmartHab. HabSim further considers disruptions and models damage and repair of certain components. This paper discusses a) system and subsystem requirements of the deep space habitat included in the HabSim platform; b) architectural choices made in response to the requirements; c) technical considerations for developing, verifying, configuring, and executing HabSim; and d) illustrative sample results from a simulation of a representative disruption scenario.
Deep-space habitation presents unique challenges in operating complex control-based systems in distant and disruptive environments far from Earth. The capability to handle unforeseen situations and manage system health is crucial for resilient operations in deep space. Trusted health management strategies operational on the International Space Station (ISS) heavily rely on ground control support. These strategies must adapt to cope with sizable communication delays imposed by the infrastructure limitations or light distances that separate ground control and the habitat. Alternative approaches that increase reliance on artificial intelligence-based onboard autonomy are constrained by limited onboard computing and risks associated with human-system integration. There is an increased need to study the design and architecture of the health management system (HMS) of such complex systems to address autonomy challenges for deep-space habitation. This paper presents a model-based HMS testbed developed at the Resilient Extra-Terrestrial Habitats Institute that is part of the HabSim, a system-of-systems simulator of a smart habitat. The testbed is designed to facilitate performance assessment of critical health management functions essential to enable fault-management autonomy, situational awareness, and interface-based mission control. In this paper, the HMS testbed is motivated and placed within HabSim, the architecture of its components is detailed, and the testbed features are demonstrated through an illustrative example. The lessons learned in developing this testbed that may be useful for others seeking to solve similar problems are also discussed.
Various types of sensors are needed to monitor the health state of smart deep-space habitats. However, measured data can be affected by sensor faults, which influence the health management system and consequently the decision-making. In this paper, an unsupervised learning approach based on convolutional autoencoders (CAEs) is developed to detect anomalies in temperature and pressure sensors. The proposed method is systematically investigated using a habitat simulator (HabSim). Several illustrative examples are demonstrated in the nominal and hazardous states of the habitat, including micrometeorite impact and fire scenarios. The performance of the proposed method using CAEs is compared with that of existing methods using auto-associative neural networks (AANNs) and variational autoencoders. This comparison is based on typical evaluation metrics, including precision, recall, F1 score, training time, and testing time. The effect of temperature–pressure coupling on the detection performance of CAEs and AANNs is explored by training different data-driven models, including one with temperature sensors, one with pressure sensors, and one with both temperature and pressure sensors. The effect of the number of faulty sensors on the performance of CAEs is studied, as with an increase in the number of faulty sensors, redundant information among the sensors is reduced. The capability of CAEs to change the number of sensors without redesigning the network architecture and retraining the neural network is investigated and demonstrated. The capabilities and limitations of the proposed solution are discussed.