In complex geomorphic environments where satellite signals are severely disturbed, satellite positioning and navigation accuracy is low owing to fewer visible satellites and poor satellite signal quality. To solve this problem, a GPS/BDS/inertial navigation deep positioning algorithm is proposed. Based on GPS/BDS pseudo-range combined positioning, this algorithm introduces inertial device measurement values for inertial navigation assisted combined positioning. With the help of fade-out Kalman filtering, it estimates the measurement noise and system noise online to further reduce positioning errors and output high accuracy positioning results. Experimental results show that the algorithm described in this paper has higher theoretical accuracy and higher computational efficiency than traditional combined localization algorithms, and has certain theoretical significance and practical value.