Complexity Reduction in Machine Learning-Based Wireless Positioning: Minimum Description Features

2024-02-14

ICC 2024 - IEEE International Conference on Communications

AI Summary

Deep learning models for wireless localization suffer from high computational overhead and poor deployability on mobile devices due to high-dimensional channel features. Method: This paper proposes P-NN, a lightweight neural network for localization based on minimal descriptive features. We introduce a novel information-theoretic, adaptive mechanism for selecting optimal feature space dimensionality, quantifying feature discriminability via mutual information and entropy. Instead of using the full power-delay profile (PDP), P-NN extracts only two time-domain features—the peak power value and its arrival time—from the time-series power spectrum. Contribution/Results: The resulting architecture achieves high localization accuracy while drastically reducing feature dimensionality and computational complexity. Experiments show that P-NN matches the accuracy of PDP-based baselines while compressing feature dimensionality by over 90% and reducing inference latency by an order of magnitude—yielding a superior accuracy–efficiency trade-off.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Intelligent Robots: Localization, Mapping, and Navigation

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

A recent line of research has been investigating deep learning approaches to wireless positioning (WP). Although these WP algorithms have demonstrated high accuracy and robust performance against diverse channel conditions, they also have a major drawback: they require processing high-dimensional features, which can be prohibitive for mobile applications. In this work, we design a positioning neural network (P-NN) that substantially reduces the complexity of deep learning-based WP through carefully crafted minimum description features. Our feature selection is based on maximum power measurements and their temporal locations to convey information needed to conduct WP. We also develop a novel methodology for adaptively selecting the size of feature space, which optimizes over balancing the expected amount of useful information and classification capability, quantified using information-theoretic measures on the 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).

Problem

• Reduces complexity in wireless positioning using minimum description features
• Selects features based on maximum power measurements and temporal locations
• Optimizes performance-complexity tradeoff versus full power delay profile methods

Innovation

• Uses minimum description features for complexity reduction
• Selects features based on power measurements and timing
• Adaptively optimizes feature space size using information theory