ARADISS: Adaptive Real-Time Anomaly Detection & Identification For Space Systems

Space missions continue to increase in number, complexity, and time cost constraints. To lead technological advancements and successfully execute these missions, NASA desires new robust onboard automated fault management technologies that address the full range of hardware software faults, are transparent reusable across platforms. This will lead to reduced costs and improved autonomy, resilience, mission quality especially in missions that cannot afford comprehensive fault management and have a higher mission risk tolerance. Global Technology Connection, Inc., (GTC) proposes to automate generation of ARADISS (Adaptive Real-time Anomaly Detection and Identification for Space Systems) models tuned to any given system. Feasibility of this framework was successfully demonstrated in Phase I on multi-rotor UAVs, electric propulsion testbed, and a preliminary CubeSat dataset. In Phase II, GTC plans to use a CubeSat platform used on NASA missions to steer the development, evaluation, and hardware integration planning. ARADISS models include simple machine learning norm models to estimate system operational variables, and self-adapting anomaly detection and identification models. ARADISS framework is applicable to virtually all electrically powered systems. Independent one-one norm maps make this approach extremely reusable, transparent and explainable. ARADISS models can be rapidly adapted to any platform and are ideal for space platforms with limited processing power. Applicable future space mission platforms include NISAR, SWOT, Dragonfly, and SPHEREx. GTC has received interest from several space technology firms, and has partnered with CubeSat OEM and a space systems integration for Phase II for rapid technology transition and space deployment in Phase III. Aggressive commercialization activities are being carried out for technology transition to NASA and other commercial space applications. Space missions continue to increase in number, complexity, and time & cost constraints. To successfully deploy these missions, new & robust onboard automated fault management technologies that address the full range of hardware & software faults are needed that are transparent & reusable across platforms. This will lead to reduced costs, prevent delays, & improve autonomy, resilience, & mission quality especially in missions that cannot afford comprehensive fault management and have a higher mission risk tolerance. To fill this need, Global Technology Connection has successfully demonstrated ARADISS framework on Satellites, UAVs, and electric propulsion testbeds in Phase I. ARADISS models include simple machine learning norm models to estimate system operational variables, and self learning anomaly detection and identification models. Independent one-one norm maps make this approach extremely reusable, transparent and explainable. ARADISS models can be rapidly adapted to any platform and are extremely suitable systems with limited processing power. The overall objective for the SBIR Phase II is to demonstrate the automated generation of ARADISS models for anomaly detection and identification on any given cyber-physical system. Specific technical objectives include: Automated System Abstraction: Demonstrate construction of system digital twin as correlation graphs that identifies all correlated variables and their correlations strengths Battery Metrics Based ML Prediction: Generate ML models trained on battery data for real-time prediction of operational variables Anomaly Detection: Identity system appropriate anomaly detection model and its optimal key configuration parameters to achieve best performance metrics Anomaly Verification: Utilize intra-layer parameter correlations identified during system abstraction to verify any detected anomalies to reduce false positives Anomaly Identification: Identify root cause of anomalies by checking inter and intralayer correlations Evaluation on Space Platforms: Implement and evaluate ARADISS models on real CubeSat datasets, hardware, and on NPAS as an easy path towards deployment on space platforms GTC proposes the following deliverables during Phase II: Quarterly reports and one final report Prototype demonstration at the end of Phase II
Benefits: The US accounts for 74% of the global investment in space exploration. It is expected to grow to over $20B by 2027 from $14.6B in 2017. 63% of the planned 80 space missions over the next decade are government missions. NASA JPL missions that could potentially benefit from automated fault management technology are satellites, spacecraft, rovers/landers, and space instruments. Specific applicable missions are Polaris, NISAR, SWOT, and SPHEREx. Our tools are ideal for onboard implementation. Several US companies like Lockheed Martin, Northrop Grumman, Planet Labs, SpaceX, and Argotec, Inc. are active in the development of spacecraft, satellites, & instruments, and UAV systems. Our novel fault management technology will enable faster development, testing, and deployment timelines, and also increase resilience. This will reduce the overall development, testing, and mission costs.