Hybrid Quantum-Classical ML for Wi-Fi Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless sensor networks face challenges in achieving accurate indoor localization and sensing without the need for expensive dedicated hardware, due to the instability and coarse granularity of RSSI measurements and the high computational power required for CSI measurements, especially in resource-limited hardware environments.
Innovation Solution
A hybrid quantum-classical machine learning system that leverages both classical deep neural networks (DNNs) and quantum neural networks (QNNs) for signal processing and sensing, utilizing remote quantum computing servers to distribute computational load and reduce hardware requirements, enabling low-power and low-cost signal processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If RSSI measurements are used for sensing, then the system is simple and low-cost, but the measurement accuracy and granularity are insufficient
Solution Approach 1:
The patent combines RSSI measurements from multiple communication links with machine learning algorithms to create a hybrid sensing system. By merging multiple low-precision measurements with computational processing, the system achieves high sensing accuracy while maintaining simplicity and low cost, resolving the contradiction between system complexity and measurement precision.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between raw RSSI measurements and final sensing results. The ML model processes and enhances the coarse-granularity RSSI data, extracting fine-grained sensing information without requiring expensive dedicated hardware, thus improving measurement precision while keeping the system simple.
2Measurement precision
If CSI measurements are used for sensing, then the sensing accuracy is improved, but the computational power requirement increases significantly
Solution Approach 1:
The patent extracts only the essential features from CSI measurements needed for sensing, rather than processing the entire high-dimensional CSI data. By taking out and processing only the relevant components, the system achieves high sensing accuracy while reducing the computational power requirement to levels suitable for resource-limited hardware.
Solution Approach 2:
The patent segments the sensing task into multiple stages: feature extraction from CSI, ML model training, and inference. By dividing the computational workload and using lightweight ML models for inference, the system maintains high sensing accuracy while reducing the real-time computational power requirement on resource-limited devices.
3Measurement precision
If deep neural networks are deployed on resource-limited hardware, then accurate sensing is achieved, but the power consumption and hardware requirements increase
Solution Approach 1:
The patent uses lightweight, simplified neural network models that can be deployed on resource-limited hardware with acceptable accuracy. These simplified models consume significantly less power than full DNNs while still achieving accurate sensing, making them suitable for battery-powered IoT devices.
Solution Approach 2:
The patent implements partial DNN functionality through lightweight models that perform only the essential sensing tasks. By implementing just enough computational capability to achieve accurate sensing without full DNN power, the system reduces energy consumption while maintaining measurement precision.
Data Source
AI summary
Communication-capable devices such as commercial Wi-Fi devices can be used for integrated sensing and communications (ISAC) systems to jointly exchange data and monitor environment. Such devices typically require diverse signal processing such as machine learning inference that demands high-power operations for real-time sensing and computing. The present invention provides a way to realize energy-efficient computing by exploiting the capability of data communications to access distributed computing resources including classical computers and quantum computers over networks. The system and method are based on the realization that computationally intensive processing is offloaded to networked hybrid classical-quantum computing to build dynamic computing graphs. Some embodiments use automated classical-quantum machine learning whose circuits and hyperparameters are automatically adjusted via gradient or heuristic optimization for Wi-Fi indoor monitoring and human tracking. For some embodiments, the system and method can reduce the power consumption and the number of trainable parameters by integrating classical and quantum neural networks. For some embodiments, signal processing such as denoising, filtering and detection is realized with hybrid classical-quantum processers over networks.