Applying Neural Networks in Aerial Vehicle Guidance to Simplify Navigation Systems

Abstract

The Guidance, Navigation and Control (GNC) of air and space vehicles has been one of the spearheads of research in the aerospace field in recent times. Using Global Navigation Satellite Systems (GNSS) and inertial navigation systems, accuracy may be detached from range. However, these sensor-based GNC systems may cause significant errors in determining attitude and position. These effects can be ameliorated using additional sensors, independent of cumulative errors. The quadrant photodetector semiactive laser is a good candidate for such a purpose. However, GNC systems’ development and construction costs are high. Reducing costs, while maintaining safety and accuracy standards, is key for development in aerospace engineering. Advanced algorithms for getting such standards while eliminating sensors are cornerstone. The development and application of machine learning techniques to GNC poses an innovative path for reducing complexity and costs. Here, a new nonlinear hybridization algorithm, which is based on neural networks, to estimate the gravity vector is presented. Using a neural network means that once it is trained, the physical-mathematical foundations of flight are not relevant; it is the network that returns dynamics to be fed to the GNC algorithm. The gravity vector, which can be accurately predicted, is used to determine vehicle attitude without calling for gyroscopes. Nonlinear simulations based on real flight dynamics are used to train the neural networks. Then, the approach is tested and simulated together with a GNC system. Monte Carlo analysis is conducted to determine performance when uncertainty arises. Simulation results prove that the performance of the presented approach is robust and precise in a six-degree-of-freedom simulation environment.