Fig. 1. Process of Vision/Inertial/UWB fusion positioning algorithm optimized based on factor graph

Fig. 2. Global factor diagram of VIUFPA algorithm

Fig. 3. Comparison of UWB ranging values before and after filtering

Fig. 4. Trajectory estimation results generated by each sequence of PL-VIO algorithm

Fig. 5. Trajectory estimation results generated by each sequence of VIUFPA algorithm

Fig. 6. Overall distribution of absolute pose errors of each sequence

Fig. 7. Experimental platform and scene configuration

Fig. 8. Comparison of experimental results of the four methods in the office scene

Fig. 9. Test configuration and preset trajectory of underground parking

Fig. 10. Comparison of experimental results of the four methods in the underground parking scene