Convolutional Neural Network Time-of-Flight Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Wi-Fi-based positioning systems face challenges in accurately estimating the position of mobile devices indoors due to multipath issues and signal blockage, where GPS and GLONASS technologies perform poorly.
Innovation Solution
Implementing a wireless device equipped with a pretrained convolutional neural network (CNN) to perform coarse time of arrival (TOA) estimation and calculate a line of sight (LOS) estimate using impulse response, enabling accurate range determination between devices through one-way or round-trip communication methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPS or GLONASS technology is used for position estimation, then positioning can be performed in open environments, but positioning accuracy deteriorates significantly indoors due to signal blockage and multipath issues
Solution Approach 1:
The patent introduces Wi-Fi signals as an intermediary positioning method that works effectively in indoor environments where GPS fails. The system uses Wi-Fi-based time-of-flight estimation to determine device position, providing a complementary positioning solution that operates in environments where satellite-based systems are blocked by buildings and structures.
2Reliability
If Wi-Fi-based positioning is used to overcome signal blockage indoors, then positioning can be performed in indoor environments, but accuracy deteriorates due to multipath issues
Solution Approach 1:
The system performs preliminary coarse time-of-arrival estimation to identify the arrival time of the direct line-of-sight signal before processing the full signal. This preliminary action allows the system to isolate and analyze only the relevant signal portion, reducing the impact of multipath reflections that arrive later and would otherwise corrupt the ranging measurement.
Solution Approach 2:
The patent segments the received signal processing into distinct stages: coarse TOA estimation to identify signal arrival, followed by extraction of the impulse response, and finally CNN-based LOS estimation. This segmentation allows each processing stage to focus on specific aspects of the signal, improving overall accuracy by handling different signal characteristics appropriately at each stage.
3Device complexity
If traditional signal processing methods are used for time-of-flight estimation, then system complexity remains low, but positioning accuracy deteriorates in multipath environments
Solution Approach 1:
The patent replaces traditional mechanical signal processing methods with a neural network-based approach. The convolutional neural network learns to identify line-of-sight signals from impulse response characteristics, substituting complex algorithmic processing with a trained model that can accurately distinguish direct signals from multipath reflections without requiring explicit mathematical modeling of the propagation environment.
Data Source
AI summary
Embodiments herein relate to using a convolutional neural network (CNN) for time-of-flight estimation in a wireless communication system. A wireless device may receive, from a remote device, wireless communications including a first transmission time value associated with the transmission of the wireless communications. The wireless device may perform a coarse time-of-arrival (TOA) estimation on the wireless communications received from the remote device. The coarse TOA estimation may be used to generate an estimated impulse response, which may be input to a CNN associated with the wireless device to calculate a line-of-sight estimate. The wireless device may determine a range between the wireless device and the remote device based on the transmission time value and the line-of-sight estimate.