Reinforcement Metalearning for Interception of Maneuvering Exoatmospheric Targets with Parasitic Attitude Loop

We use Reinforcement Meta-Learning to optimize an adaptive integrated guidance, navigation, and control system suitable for exoatmospheric interception of a maneuvering target. The maps observations consisting strapdown seeker angles rate gyro measurements directly thruster on-off commands. Using high fidelity six degree-of-freedom simulator, we demonstrate that the optimized policy can adapt parasitic effects including angle measurement lag, attitude loop resulting from scale factor errors Gaussian noise on rotational velocity measurements, time varying center mass caused by fuel consumption slosh. Importantly, gives good performance over wide range challenging target maneuvers. Unlike previous work enhances observability inducing line sight oscillations, our is only available gyros. Through extensive Monte Carlo simulation randomized scenarios, close augmented proportional navigation with perfect knowledge full engagement state. computationally efficient requires minimal memory, should be compatible today's flight processors.