MORE_LOC: MULTI-OUTPUT REGRESSION BASED LOCALIZATION FOR INDOOR APPLICATIONS

Abstract

Many indoor applications are developed and being used in the industry and homes. One of the key requirement in these applications is knowing the position of the target. In this paper a novel multi-output regression based localization (MORE_LOC) technique is proposed, which can estimate the location of target node by utilizing the received signal strength (RSS) from multiple anchor nodes. These anchor nodes are deployed in the same room or hall where target node is roaming. MORE_LOC utilizes machine learning algorithms for generation of prediction models. These models are helpful in estimation of current location of mobile target moving in the indoor environment. In this paper, datasets collected from different scenarios are also discussed. Experiments are conducted using real as well as synthetic dataset. The performance analysis of the proposed techniques is done by calculating mean absolute error (MAE), root mean square error (RMSE) and Mean Localization Error (MLE). It is observed that MORE_LOC gives best results when Decision Tree (DT) algorithm is utilized. Values of performance metrics in this case are: MAE 0.23m, RMSE 0.44m and MLE 0.36m. Experiments are also conducted by using MORE_LOC with kalman filter. In this case also proposed technique performs better with DT algorithm. And the values of performance metrics are: MAE 0.24m, RMSE 0.43m and MLE 0.38m.