The document describes research into improving indoor localization using Gaussian processes (GP) and different mean offset models. It presents the problem of indoor navigation and existing solutions using Wi-Fi/Bluetooth fingerprinting and sensor data. Constant, linear, and log-distance mean offset models for GP signal prediction are introduced and evaluated. The log-distance model provides the most accurate predictions, visibility area estimation, and access point positioning. Evaluation shows the log-distance model outperforms other methods, particularly with reduced reference point data.