The rapid development of global trade relies heavily on efficient positioning-based technologies. Passive Positioning based on Low Earth Orbit (LEO) constellation becomes the attractive approach to tackle the problem with its strong anti-interference, high concealment, and low-cost advantages. In the case of a wide area monitoring range of LEO constellation, massive data are obtained, which is a challenge for the traditional solution to simultaneously meet the positioning speed and accuracy. To efficiently process the massive data, many researchers have studied the Passive Positioning method supported by deep learning. However, the current studies were one-side, such as assuming the number of emitters and receivers being the fixed value and only considering the signal data of Passive Positioning problem, which are not universal and need to be more conducive to the application in the LEO constellation positioning scene. To address these issues, we first propose an end-to-end Deep neural network for the Passive Positioning of emitters and receivers with variable numbers called DPV. In addition, we publish a passive positioning dataset for emitters and receivers with the variable number in a multi-signal aliasing condition. In this dataset, our proposed method realizes effective positioning. Compared with the traditional methods, the proposed method achieves state-of-the-art performance under simultaneous positioning accuracy and positioning speed requirements.
The dataset will be made public very soon.