Impact of Preprocessing on Neural Network-Based RSS/AoA Positioning

TL;DR

This work addresses the nonlinearities and geometry-dependent noise in hybrid RSS-AoA positioning by using a multilayer perceptron (MLP) to map measurements to the 3D target position $\hat{\mathbf{t}} \in \mathbb{R}^3$. It compares two input regimes: preprocessed features derived from a model-based linearization (WLS) and raw RSS/AoA measurements, keeping a consistent network architecture and training. Results show that the DL model with geometry-aware preprocessing consistently surpasses traditional linear estimators under RSS noise and matches or exceeds state-of-the-art performance under AoA noise, whereas raw-data learning yields no significant gains. The findings highlight the value of combining geometry-informed preprocessing with neural networks for robust and accurate 3D localization in practical wireless systems.

Abstract

Hybrid received signal strength (RSS)-angle of arrival (AoA)-based positioning offers low-cost distance estimation and high-resolution angular measurements. Still, it comes at a cost of inherent nonlinearities, geometry-dependent noise, and suboptimal weighting in conventional linear estimators that might limit accuracy. In this paper, we propose a neural network-based approach using a multilayer perceptron (MLP) to directly map RSS-AoA measurements to 3D positions, capturing nonlinear relationships that are difficult to model with traditional methods. We evaluate the impact of input representation by comparing networks trained on raw measurements versus preprocessed features derived from a linearization method. Simulation results show that the learning-based approach consistently outperforms existing linear methods under RSS noise across all noise levels, and matches or surpasses state-of-the-art performance under increasing AoA noise. Furthermore, preprocessing measurements using the linearization method provides a clear advantage over raw data, demonstrating the benefit of geometry-aware feature extraction.

Figures (5)

• Figure 1: 3d system model
• Figure 2: Illustration of the MLP-based positioning network. The model can operate on two types of inputs: preprocessed features derived from the WLS formulation ($\text{vec}(\mathbf{A}_w), \text{vec}(\mathbf{b}_w)$) or raw measurements ($\mathbf{p}, \boldsymbol{\phi}, \boldsymbol{\alpha}$). The network consists of a linear layer, layer normalization, ReLU activation, and a final linear layer to predict the 3D target position $\hat{\mathbf{t}} = [\hat{t}_x, \hat{t}_y, \hat{t}_z]^T$. Training is performed using the MSE loss with the Adam optimizer.
• Figure 3: RMSE of the target position versus noise standard deviation $\sigma_{n_i}$.
• Figure 4: RMSE of the target position versus azimuth noise standard deviation $\sigma_{m_i}$.
• Figure 5: RMSE of the target position versus elevation noise standard deviation $\sigma_{\nu_i}$.