Family of Special Operations Vehicles (FOSOV)

Most of the existing navigation algorithms for SOVs are based on the measurements from INS/GNSS which closely depends on the availability of GPS. An alternative way is to utilize inertial measurement sensors for navigation. However, SOVs might have large inertial measurement errors even with the help of state-of-the-art technologies such as Kalman Filter. Inspired by the emerging machine learning technologies as well as advanced adaptive filtering algorithms, Intelligent Fusion Technology, Inc. and University of Nevada, Reno propose to develop a novel hybrid learning enhanced hierarchical multi-layer adaptive filter based IMU error pattern identification and predictive error correction algorithm for SOV applications. Key innovation includes (1) adopting hierarchical multi-layer hybrid filter bank to reduce the complexity and maintain the accuracy of navigation state estimation model, (2) leveraging hybrid reinforcement learning and time series long short-term memory neural network based online IMU predictive error correction, (3) utilizing General Adversarial Network (GAN) based Transfer Learning to effectively adapt the model to various environments and missions, (4) integrating a real-time GPS jamming/interference detection and alerting system with team awareness kit Integration.