Multi-Stage Cascaded Algorithm for Localization with Integrated BEacon Ranking (CALIBER) for Bluetooth-Based Indoor Localization

Accurate indoor localization using Bluetooth Low Energy (BLE) remains challenging due to high variability and noise in Received Signal Strength Indicator (RSSI) measurements. This paper presents a multi-stage Cascaded Algorithm for Localization with Integrated BEacon Ranking (CALIBER) designed to enhance the accuracy of distance estimation and position inference in BLE-based systems. The proposed approach combines four successive processing stages: (1) beacon selection using a Minimum Composite Score criterion to filter unstable transmitters; (2) outlier removal through Median Absolute Deviation and Temporal Consistency Check; (3) residual noise reduction using Gaussian Kernel Smoothing; and (4) adaptive signal refinement via an Extended Kalman Filter. Filtered RSSI values are converted to distance estimates using a free-space path loss model, followed by trilateration to compute two-dimensional positions. Experimental evaluation in a corridor testbed demonstrates that the proposed CALIBER algorithm achieves a 56.9% improvement in localization accuracy compared to unfiltered RSSI input, reducing the mean localization error from 4.58 m to 1.97 m. Furthermore, by excluding extreme static points that exhibit uncharacteristic signal behavior, the mean error is further reduced to 1.12 m. The system requires only BLE beacons as infrastructure, making it easy to deploy and a low-cost solution for indoor localization in various built environments.