Adaptive Scale Factor Compensation for Missiles with Strapdown Seekers via Predictive Coding

TL;DR

This paper introduces a deep learning-based adaptive method to correct scale factor errors in missile seekers, improving accuracy in realistic simulations and applicable to various missile and inertial measurement systems.

Contribution

It presents a novel predictive coding neural network approach for real-time scale factor error compensation in missile guidance systems.

Findings

Effective compensation of rotational velocity errors

Successful real-time adaptation in six degrees-of-freedom simulation

Applicable to various missile seeker configurations

Abstract

In this work we present a method to adaptively compensate for scale factor errors in both rotational velocity and seeker angle measurements. The adaptation scheme estimates the scale factor errors using a predictive coding model implemented as a deep neural network with recurrent layer, and then uses these estimates to compensate for the error. During training, the model learns over a wide range of scale factor errors that ideally bound the expected errors that can occur during deployment, allowing the deployed model to quickly adapt in real time to the ground truth error. We demonstrate in a realistic six degrees-of-freedom simulation of an exoatmospheric intercept that our method effectively compensates for concurrent rotational velocity and seeker angle scale factor errors. The compensation method is general in that it is independent of a given guidance, navigation, and control system implementation. Although demonstrated using an exoatmospheric missile with strapdown seeker, the method is also applicable to endoatmospheric missiles with both gimbaled and strapdown seekers, as well as general purpose inertial measurement unit rate gyro compensation.