Due　to　variance　problems　of　the　Received　Signal　Strength　Indicator　(RSSI),　the　Wi-Fi　fingerprinting　(WF)　positioning　has　great　volatility.　Although　several　calibration　algorithms　have　been　proposed,　the　effect　is　unsatisfied.　In　this　paper,　we　presented　a　new　algorithm　which　combines　WF　and　Precision　Depth　Recorder　(PDR)　based　on　Unscented　Kalman　Filter　(UKF)　to　optimize　the　positioning　results.　Moreover,　an　adaptive-step-length-estimation　(ASLE)　algorithm,　which　can　gain　good　performance　while　tracking　the　pedestrians,　was　designed　to　reduce　the　cumulative　error　of　PDR　positioning　system.　Experiments　show　that　all　error　distances　are　less　than　4.46m,　of　which　79%　are　within　2m.　The　combination　of　WF　and　PDR　leads　to　a　more　stable　indoor　positioning　over　the　long　term.　©　2018　Association　for　Computing　Machinery.