Navatek proposes to develop a fault prediction and detection solution that improves NASA’s ability to reveal latent, unknown conditions while also improving its detection time and reducing the rate of false positive and negative detections of known conditions that would lead to failure of the life sustainment system. Our approach feeds historical and real-time sensor data to a digital twin of the life sustainment system, which is a digital simulation of the entire functioning system and its environment. This digital twin is used by a reinforcement learning adversarial agent (RLAA) to simulate many possible scenarios into the future. The RLAA autonomously learns the environmental and system perturbations that lead to faults in its simulations, thus providing a method for prediction. These predictions are continuously compared against new incoming data to detect faults and further improve the digital twin’s accuracy.
Navatek’s experience in physics-based modeling and simulation of environmental conditions and power flows is directly relevant to building the digital twins needed to manage the system health of space habitats. For Phase I and II, we will limit the scope of our digital twins to structural integrity and HVAC systems, to include their power supplies, leveraging data from pressure, flowrate, UV, temperature, and current/voltage sensors. Navatek is also uniquely capable of designing and rapidly prototyping low-cost inflatable structures. If selected for Phase II, we propose to construct a surrogate inflatable space habitat as an experimental apparatus with which to perform rapid, low-cost validation of the digital twin’s fault prediction and detection ability.
The proposed effort will lead to valuable contributions to active fault detection in hazardous environments. The reinforcement learning adversarial agent architecture we are proposing would significantly expand the operational envelope of NASA space environment research by enabling faults to be accurately predicted and prevented, saving lives and infrastructure in the following applications.
♦ Habitation Systems
♦ Power Plant Operations
♦ Flight Controllers
♦ Spaceflight Missions
Reinforcement learning adversarial agents have non-NASA commercial applicability for fault detection in advanced automated systems. Such systems are necessarily complex, the volumes of sensor data are large and not well-suited for human-only monitoring, and the consequences of system failure are severe. Examples:
♦ Robot Exploration in Hazardous Environments
♦ Unmanned Underwater Vehicles