Ultra-wideband (UWB) technology is used for indoor positioning, but its positioning accuracy is usually degenerated by various obstacles in the indoor environment because of non-line-of-sight (NLOS). Facing the complex and changeable indoor environment, an indoor positioning system with UWB based on a digital twin is presented in this paper. The indoor positioning accuracy is improved with a perception–prediction feedback of cyber-physics space in this indoor positioning system. In addition, an anchor layout method with virtuality–reality interaction and an error mitigation method based on neural networks is put forward in this system. Finally, a case study is presented to validate this indoor positioning system with a significant improvement in positioning accuracy.