CSI Based Indoor Localization Using Ensemble Neural Networks

Abstract

Indoor localization has attracted much attention due to its indispensable applications e.g. autonomous driving, Internet-Of-Things (IOT), and routing, etc. Received Signal Strength Indicator (RSSI) was used intensively to achieve localization. However, due to its temporal instability, the focus has shifted towards the use of Channel State Information (CSI) aka channel response. In this paper, we propose a deep learning solution for the indoor localization problem using the CSI of a 2 × 8 Multiple Input Multiple Output (MIMO) antenna. The variation of the magnitude component of the CSI is chosen as the input for a multilayer Perceptron (MLP) neural network. Data augmentation is used to improve the learning process. Finally, various MLP neural networks are constructed using different portions of the training set and different hyperparameters. An ensemble neural network technique is then used to process the predictions of the MLPs in order to enhance the position estimation. Our method is compared with two other deep learning solutions; one that uses Convolutional Neural Network (CNN), and the other uses MLP. The proposed method yields higher accuracy than its counterparts.

CSI-MIMO: K-nearest Neighbor applied to Indoor Localization

“…However, when the number of training samples is sufficient, it is able to localize with lower error. When all the 16 antennas are used, the k-nearest neighbor solution achieves a 2.4 cm MSE compared to 3.1 cm for the Ensemble NN technique [16]. This represents a 16% improvement over the closest error.…”

“…We propose a computationally lighter method to reduce noise and dimensionality with a negligible loss in polynomial estimation accuracy. Another approach tested on the same dataset uses the difference between adjacent magnitude values [16] as the input to a Neural Network Ensemble. In addition, a data augmentation step is used to improve estimation accuracy.…”

“…2) NDR [15]: The magnitude component is reduced using a polynomial regression and used as an input to an MLP. 3) Ensemble [16]: The differences between magnitude values are fed into an MLP Neural Network ensemble with data augmentation. Figure 8 presents the position estimation accuracy of the proposed k-nearest neighbors solution and the previ-ously mentioned comparable methods.…”

Application of Reinforcement Learning and Deep Learning in Multiple-Input and Multiple-Output (MIMO) Systems

“…Indoor localization based on DL and CSI for 28 MIMO antenna is considered in [ 181 ]. The input to the multi-layer perceptron NN is the change in the magnitude component of the CSI and the learning process is improved using data augmentation.…”

CSI-Based Localization with CNNs Exploiting Phase Information

“…More complex neural network configurations were used in [3] using as input the time-domain channel impulse response. The authors of [4] were able to achieve sub-centimeter accuracy by employing a denoising technique and an ensemble of neural networks. They considered only the magnitude of the channel, since they identified that phase measurements at the same position can change over time.…”