Abstract: An indoor localization method is proposed based on Wi-Fi fingerprint in the human-robot shared environment, in order to solve the problem of indoor global localization for service robots and pedestrians. Firstly, the device-independent robust position fingerprint is extracted from dual-band Wi-Fi signals by kernel principal component analysis (KPCA) for WiFi fingerprinting. Then, combining with the pedestrian dead reckoning (PDR) method, a Wi-Fi/PDR integrated positioning algorithm is presented based on selective update particle filter (SUPF) to improve the stability and the accuracy of pedestrian positioning. In the algorithm, the Wi-Fi localization results in the moving scene are preliminarily corrected using PDR, and the corrected results are evaluated by defining the trusted space of an adaptive size, so that untrusted Wi-Fi localization estimations are removed before the data fusion. Finally, localization experiments are carried out in a real scenario, and the average positioning error of the Wi-Fi/PDR integrated positioning algorithm is about 2 m. Experimental results demonstrate that the proposed method improves the accuracy and the robustness of the positioning system.