Abstract: Root architecture can be widely used for genotype selection as the high nitrogen fixation for leguminous plants. The root traits can be quantified to optimize the soybean cultivar with the high nitrogen potential. This study aims to explore the root architecture and symbiotic nitrogen fixation potential of different soybean-rhizobia symbiotic combinations. The soybean was inoculated with the different rhizobia. A modular rhizobox was reconstructed and then assembled with multiple grids of stainless steel using the root system architecture (RSA). Among them, each slice of steel grid was held in a supporting rack, including 4 hundred small cells. The soybean root architecture was reconstructed to connect some points on the grid layer by layer. The root system architecture also included the lateral root angle, the root length distribution in depth, the root distribution on the top view, and nodule production in soil. The nitrogen potential was evaluated in the different combinations of soybean and rhizobia. The rhizobox was set as 1 cm grid size and 3 cm interval, according to the minimum compaction of grid size. Small grid size and interval greatly contributed to the large soil compactness. The results showed that there was a significant increase in the lateral root angle, the root volume, and the root length of the root system architecture after the inoculation with rhizobia. There was no outstanding effect on the circumferential distribution of the roots. The soybean roots were symmetrically distributed in four columns as usual. The shallow root system was also found in the symbiotic combination with the highest nitrogen fixation potential (Huning 95-1 and USDA3l1b110). Specifically, many lateral roots often appeared in the shallow layer. The root system architecture after the inoculation with rhizobia under soil conditions was provided for some evidence to screen and select the rhizobia and soybean cultivar with the high modulation potential. As such, the original backbone and topology were maintained more suitable for nutrition, breeding, and ecology, compared with the 3D point cloud. The finding can also be used for the multiple purposes of root system architecture.