For crowd sensing applications, the inaccuracy and conflict of reported samples, commonly caused by untrained participants and heterogeneous mobile devices, have received more and more attention. Given this, constructing a high-quality indoor bluetooth signal map and motion state map via crowd sensing is very challenging. In this paper, we design a robust indoor bluetooth signal map construction scheme, named by BiMCS. This is accomplished through a family of robust estimation and enhances the quality of a bluetooth signal map construction via crowd sensing. For heterogeneous devices, we use a linear robust estimation technique to map the signal strength scanned by different users into a uniform space. We also alleviate the negative influence of outliers and detect the abnormal participants by utilizing the multivariate robust statistics on the signal information uploaded by different users. While relying on the self-learning and self-judging to improve the quality of the indoor motion state map. Specifically, we translate the continuous motion information into a scalar value about the motion state through the internal comparison, which also effectively resolves the problem of diversity in mobile devices. Finally, we implement an Android-based system for constructing the indoor bluetooth signal map and motion state map as well as conduct extensive experiments under an indoor environment. The experimental results validate the enhancement of the constructed bluetooth signal map by the proposed BiMCS, additionally verifying the high accuracy in estimating the motion state of participants.