A Three-Layer Parallel Implementation of the DPE Receiver Based on GPU

The Popular of Global Navigation Satellite System (GNSS) is attributed to its good performance. However, the performance of a conventional GNSS receiver is severely degraded or even failed to provide services in complex environments. Direct Position Estimation (DPE) combines all the visible satellite signals for energy accumulation in the navigation domain first, and then the navigation solution is estimated directly from the accumulation results. Hence the navigation performance of GNSS can be effectively improved in complex environments by DPE. However, the signal accumulation of DPE requires calculating the correlation values of all the visible satellite signals at each candidate grid in the navigation domain, the computation load of which limits its application. To alleviate this problem, a GPU-based parallel implementation of the DPE receiver is designed, which decomposes the operations involved in the joint accumulation into three levels of parallel processing. The effectiveness of the designed parallel implementation of the DPE receiver is verified by experiment results. With the available hardware in our laboratory, the designed DPE implementation improves the efficiency to approximately 146 times on average over the conventional DPE implementation.