Minimum Description Feature Selection for Complexity Reduction in Machine Learning-based Wireless Positioning
Myeung Suk Oh, Anindya Bijoy Das, Taejoon Kim, David J. Love, Christopher G. Brinton
TL;DR
This work addresses the high computational burden of deep WP methods that rely on high-dimensional power delay profiles. It introduces MDL-based minimum description features—largest power measurements and their times—and a multi-channel P-NN that jointly processes a sparse PDP image and raw measurement matrices, augmented by self-attention to improve spatial learning. An adaptive feature-size selection framework based on log-likelihood, information-acquisition probability, and KL divergence enables scenario-specific tuning without online retraining, delivering strong performance at substantially reduced feature dimensions, especially in low-SNR or NLOS conditions. Collectively, these elements yield a practical, scalable WP solution with favorable performance-complexity tradeoffs suitable for mobile and resource-constrained environments.
Abstract
Recently, deep learning approaches have provided solutions to difficult problems in wireless positioning (WP). Although these WP algorithms have attained excellent and consistent performance against complex channel environments, the computational complexity coming from processing high-dimensional features can be prohibitive for mobile applications. In this work, we design a novel positioning neural network (P-NN) that utilizes the minimum description features to substantially reduce the complexity of deep learning-based WP. P-NN's feature selection strategy is based on maximum power measurements and their temporal locations to convey information needed to conduct WP. We improve P-NN's learning ability by intelligently processing two different types of inputs: sparse image and measurement matrices. Specifically, we implement a self-attention layer to reinforce the training ability of our network. We also develop a technique to adapt feature space size, optimizing over the expected information gain and the classification capability quantified with information-theoretic measures on signal bin selection. Numerical results show that P-NN achieves a significant advantage in performance-complexity tradeoff over deep learning baselines that leverage the full power delay profile (PDP). In particular, we find that P-NN achieves a large improvement in performance for low SNR, as unnecessary measurements are discarded in our minimum description features.