Enhancement of RSSI-Based Localization Using an Extended Weighted Centroid Method With Virtual Reference Node Information

In this paper, an extended weighted centroid localization (EWCL) method for a received signal strength indicator (RSSI)-based localization system is proposed. In the traditional WCL method, an unknown target position is estimated using the actual position of each reference node and RSSI information collected from each reference node based on the distance between the target and each reference node. The estimation accuracy is significantly increased when more reference nodes are applied in the localization system. In addition, in the traditional WCL method, four reference nodes deployed in a square area are initialized and tested. Such a configuration produces good localization results. By this information, in this work, the EWCL method is developed to estimate a target's position taking into consideration information from four reference nodes, where only three actual reference nodes are deployed in a real test area. Here, the fourth reference node position as the virtual reference node position is defined, and a new technique involving a guideline solution is developed to estimate the distance between the target and the fourth reference node using distance information from the three actual reference nodes. By our purpose, the hardware cost and the complexity of the experimental setup can be reduced, while the estimation accuracy can be increased. To verify the EWCL method, both theoretical and experimental studies are performed. The theoretical study first demonstrates the EWCL performance without the effect of radio signals. The experimental study using low-cost low-power 2.4 GHz nodes deployed in a semi-outdoor test area reveals more realistic results. The results indicate that the EWCL method significantly outperforms the WCL method in all test cases. The localization error is reduced by 49.151% in the theoretical test. In the experimental test, the localization errors are reduced by 33.708 and 28.351% for scenarios without and with a human presence, respectively.