Geolocation Neural Network for Heterogeneous Antenna Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing geolocation methods are inadequate for accurately locating objects transmitting radio signals to networks with heterogeneous antenna density, particularly outdoors, as they fail to provide precise positioning due to sparse antenna coverage and are not optimized for mobile objects.
Innovation Solution
A method involving a learning phase where the space is divided into zones, with a database recording signal information from known positions to each antenna, and a neural network is trained to determine the probability of location in each zone, allowing for precise object positioning using a classifier and neural network analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional geolocation methods (filtering by extended Kalman filter, fingerprinting) are used, then the system is simple to implement, but the measurement precision is insufficient for heterogeneous antenna density networks
Solution Approach 1:
The patent divides the service area into multiple zones, each associated with a specific antenna. This segmentation allows the system to handle heterogeneous antenna density by treating each zone independently, improving measurement precision without requiring a complete redesign of the geolocation system.
Solution Approach 2:
The patent implements a learning phase before actual geolocation operations, where the system pre-processes signal data from multiple antennas and stores it in a database. This preliminary action enables the neural network to be trained in advance, improving geolocation precision during operational phase without adding real-time computational complexity.
2Adaptability or versatility
If a dense mesh network is used (as in FR 3068141), then indoor location precision is improved, but the system becomes unsuitable for outdoor locations with sparse antenna coverage
Solution Approach 1:
The patent applies different processing strategies to different zones based on local antenna density characteristics. Each zone is analyzed independently with its own signal strength patterns, allowing the system to adapt to varying antenna densities across different geographic areas rather than requiring uniform dense coverage throughout.
Solution Approach 2:
The neural network-based geolocation system is designed to handle both indoor and outdoor environments universally. By training the neural network on diverse signal patterns from multiple antennas with heterogeneous density, the same system can effectively locate objects in both dense indoor environments and sparse outdoor environments.
3Measurement precision
If fingerprinting method is used with database of received signal levels, then the method is simple to implement, but the location precision is limited to pre-defined positions in the database
Solution Approach 1:
The patent replaces the traditional fingerprinting method's discrete position matching approach with a neural network-based continuous position estimation system. The neural network processes signal strength patterns from multiple antennas and outputs precise coordinate positions, eliminating the limitation of being constrained to pre-defined database positions while maintaining implementation feasibility.
Data Source
AI summary
A method for locating an object in a plane that includes a first so-called learning phase for informing a database, and a second so-called location phase, the location phase using a classifier operating using the database comprising information representing the transmission of signals between the object and antennas, for determining a probability of location of the object in zones of the plane, the location phase further using a neural network, for determining, for the zone of the plane for which the location probability is the highest, the position of the object in the form of coordinates in the plane. The invention also relates to a location device implementing the method.