Integrated Navigation Algorithm for Intelligent USVs with Unstable GPS

In complex marine or unpredictable combat environments, there may be unstable and even loss of GPS for the integrated navigation system of unmanned surface vehicles (USVs). This paper analyzes the error model of loose combination navigation, and proposes a method of the recurrent neural network (RNN) aided inertial navigation system (INS). When the GPS signal is interrupted, the RNN is used to correct the navigation error of the INS. Simulation results show that in case of unstable GPS, the RNN-aided navigation system can ensure rapid convergence of error, and the east position error can be kept within a satisfactory scope. Compared with the single inertial navigation system, the navigation accuracy is effectively improved.