Machine-Learned Position Correction for Weak GPS Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing GPS systems face inaccuracies due to clock mismatches between satellites and receivers, leading to errors in position determination, especially when signals from fewer than four satellites are received, and can result in significant deviations from the desired location.
Innovation Solution
A system and method utilizing machine learning to analyze facility and environment information, including text, landmark, and environmental data, to provide precise position information by classifying and matching these items to GPS data, using a user's current GPS information and surrounding picture data for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPS satellite signals are used for position determination, then position information can be obtained globally, but position accuracy deteriorates when signals from fewer than four satellites are received or when clock mismatches occur
Solution Approach 1:
The patent introduces machine learning models as intermediary systems between GPS signal reception and final position determination. The learning model processes GPS coordinates along with facility information, environment information, and image data to compensate for GPS errors, thereby maintaining position accuracy even when direct GPS signals are weak or insufficient
Solution Approach 2:
The patent replaces the purely mechanical/mathematical GPS calculation system with an intelligent system that uses machine learning algorithms. Instead of relying solely on satellite signal geometry and time measurements, the system substitutes computational intelligence to learn and correct position errors based on patterns in facility and environment data
2Measurement precision
If machine learning models are trained with frequent updates to adapt to changing environments, then position accuracy improves, but system complexity and data processing requirements increase
Solution Approach 1:
The patent performs machine learning training in advance using pre-collected facility information, environment information, and image data. The learned models are then deployed for position determination, avoiding the need for real-time complex processing during actual position queries. This preliminary training approach simplifies the operational system while maintaining high accuracy
Solution Approach 2:
The patent uses pre-extracted facility information and environment information from maps and databases, rather than processing all possible environmental data in real-time. This partial action approach focuses computational resources on the most relevant features for position correction, reducing overall system complexity while maintaining effectiveness
Data Source
AI summary
The present invention relates to a system for providing precise position information on the basis of machine learning, and a provision method therefor, and, more specifically, to a system for providing precise position information on the basis of machine learning, comprising: an information provision unit, which performs machine-learning-based learning for precise position determination, receives position-related information from a user to analyze the position-related information, and thus determines a precise position, and provides the determined precise position information; and a user terminal unit, which receives the precise position information by using a pre-installed application, inputs position-related information about a desired position through the application and receives precise position information related to the inputted position-related information.