Integrated Data Registration for Networked Aircraft

An integrated data registration solution using neural networks is proposed. Previous data registration solutions typically involve a fixed model of biases and noise. The biases are then estimated and the estimates used to offset future sensor measurements. These models are unable to solve more complex situations where the errors do not fit into a given model. Given a changed operating environment, a new model would need to be developed. A new approach is needed that can respond flexibly to highly variable operating environments. Neural networks are a solution that can offer this flexibility. We propose to extend the work of Haim Karniely's"Sensor Registration using Neural Networks"to include sensor navigation and clock synchronization. The solution is algorithmic in nature, requiring the identification of a method for training the NN to remove sensor, navigation and clock biases. Customized error and learning rate functions will be developed to train the neural net. Existing Heron Systems products will be leveraged to create a test environment for training and evaluating the neural network. The test environment will include a scenario generator for developing training and test cases and an unclassified E-2 radar model.