Non-Linear Autoregressive Delay-Dependent INS/GPS Navigation System Using Neural Networks

Autoregressive neural network fusion architecture is presented for low-cost global positioning system (GPS) and inertial measurement unit (IMU) measurements integration. The proposed intelligent fusion architecture is a non-linear method that takes into account the variable delay between GPS measurement epochs. This delay is due to possible operation of the GPS/IMU integrated system in urban canyon environments. To verify the performance of the proposed method, a simulation environment is constructed. In the simulation environment, the vehicle's truth model is known and GPS/IMU measurements are simulated with a number of GPS measurements outages. The performance of the proposed fusion architecture is evaluated against the truth state of the vehicle. Subsequently, the proposed method is used in an experimental setup to estimate the state of a vehicle that is driven through a number of chosen paths. The performance of the fusion architecture is compared against a commercial off-the shelf solution.