Adaptive Markov Inference Game Optimization
AMIGO is a game theory enabled machine learning solution. The current prototype has been designed and trained for a specific ground space surveillance asset: Lockheed Martin Space Fence. AMIGO’s data includes 90 objects performing continuous thrust maneuvers within the Space Fence field of regard (FoR). The acceleration magnitudes are 0.0g (normal behavior), 0.001g, 0.25g, 0.5g, 1.0g, 2.0g, and 3.0g. Excellent satellite behavioral classification results with a false alarm rate of 1.5% can be obtained. AMIGO can also be applicable to other assets.
METHOD: Space Domain Awareness
KEYWORDS: Dynamic Game Theory, ML/AI
TECHNICAL SPECS: Uncertainty Modeling and Propagation, Game Theoretic optimization
RESEARCH: Shen, D., Sheaff, C., Lu, J., Chen, G., Blasch, E., & Pham, K. (2019, July). Adaptive Markov inference game optimization (AMIGO) for rapid Discovery of satellite behaviors. In Sensors and Systems for Space Applications XII (Vol. 11017, p. 57). SPIE.