Ultra-wideband (UWB) positioning systems have been extensively studied and implemented in dense urban and indoor environments. Assuming that the positions of UWB anchors are given, the position of a UWB tag on a moving platform can be estimated by measuring its distances to the anchors. However, precisely determining anchor positions in advance is labor-intensive and may be impractical in emergency situations, such as firefighting or search and rescue missions. To address this challenge, this paper presents a method for simultaneously localizing both the tag and fixed anchors in real-time without any prior knowledge of the anchor positions. Multiple UWB tags (MUT) are placed on an autonomous mobile robot (AMR), which allows the anchor initialization at the first observation without delay. In addition, a novel protocol for the MUT measurement is proposed to minimize measurement time. The geometric structure of the MUT placement is incorporated as a constraint in the estimation algorithm. In this paper, a tightly coupled approach is proposed for the estimation, which enhances the convergence and accuracy of localization. The proposed method is rigorously evaluated through simulations and real-world experiments, with results demonstrating its superior performance in the UWB localization, confirming its effectiveness in practical applications. Specifically, the proposed method achieves root mean square errors of 0.076 m in AMR's position, 0.017 rad in AMR's heading angle, and 0.131 m in anchor position, representing reductions of 24.05%, 45.4%, and 56.36%, respectively, compared with a state-of-the-art algorithm.