AI/ML RF Sensing for Target Detection in Multipath Blockage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face challenges in accurately detecting and determining information about target objects in complex propagation environments, such as those with line of sight blockage and multipath with weak line of sight blockage, due to the limitations of current technologies in accurately using machine learning based sensing technologies, such as 5G, especially in scenarios with complicated propagation environments.
Innovation Solution
Implementing artificial intelligence (AI)/machine learning (ML) models for radio frequency (RF) sensing, where sensing nodes and network entities configure and utilize AI/ML models to analyze received waveforms, extract features, and determine target object information, including position, speed, and quantity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sensing technologies are used in complex propagation environments, then device complexity is reduced, but measurement precision deteriorates due to inability to accurately detect target objects in line of sight blockage and multipath scenarios
Solution Approach 1:
The patent introduces AI/ML models as intermediary components between the sensing measurements and target object information determination. These models process raw sensing data and extract meaningful information about target objects, enabling accurate detection in complex environments without requiring direct complex signal processing in the sensing node itself.
Solution Approach 2:
The patent replaces traditional mechanical signal processing methods with AI/ML-based processing. Instead of using conventional algorithms to analyze sensing measurements, the system employs machine learning models that have been trained to recognize target object patterns in complex propagation environments, thereby improving measurement precision.
2Measurement precision
If AI/ML models are implemented for sensing, then measurement precision improves in complex propagation environments, but device complexity increases due to model configuration and processing requirements
Solution Approach 1:
The patent applies preliminary action by pre-configuring AI/ML models with appropriate architectures and parameters before deployment. The network entity configures the sensing node with pre-trained models that are already optimized for specific sensing tasks, eliminating the need for real-time model training and reducing operational complexity.
Solution Approach 2:
The patent implements universal AI/ML models that can handle multiple sensing scenarios and target object types. These models are designed to be versatile and can process different types of sensing measurements (e.g., RF signals, acoustic signals) and identify various target objects, thereby reducing the need for multiple specialized models.
3Reliability
If AI/ML models are used for sensing, then reliability improves in detecting target objects with line of sight blockage, but loss of time increases due to model configuration and processing
Solution Approach 1:
The patent uses preliminary action by pre-configuring and pre-training AI/ML models before actual sensing operations. The models are prepared in advance with appropriate architectures, parameters, and training data, so that during operation they can quickly process sensing measurements and provide reliable detection results without requiring time-consuming real-time training or configuration.
Solution Approach 2:
The patent optimizes model processing time by adjusting parameters such as model complexity, input data dimensions, and processing thresholds. By carefully selecting and tuning these parameters, the system achieves a balance between detection reliability and processing speed, ensuring that AI/ML-based sensing meets real-time requirements.
Data Source
AI summary
Disclosed are techniques for artificial intelligence/machine learning (AI/ML) based sensing. In an aspect, a sensing node may configure an AI/ML model to be used for sensing. The sensing node may obtain sensing measurements. The sensing node may determine sensing target information by applying the AI/ML model to the sensing measurements. Example sensing nodes include user equipment and base stations. In another aspect, a network entity or network node may receive first information indicating a capability of a sensing node to support an AI/ML model for sensing. The network entity or network node may configure, based on the first information, an AI/ML model to be used by the sensing node for sensing, and send AI/ML model configuration information to the sensing node. Example network entities or network nodes include base stations, sensing servers, and AI servers.