Neural Network Location Detection for Scalable RTLS
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Real-Time Location Systems (RTLS) face challenges in accurately tracking objects in real-time due to multipath effects and processing bottlenecks, leading to inaccurate and stale location indications, especially when dealing with large quantities of objects.
Innovation Solution
The implementation of interrelated neural networks with adaptive error monitoring and processing techniques, such as Centroid k-means and median filters, to handle missing readings and extreme variations in received signal strength indicator (RSSI) values, allowing for scalable real-time location determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If known tags and communication standards such as Wi-Fi are used for tracking, then full-scale deployment potential is achieved, but accuracy is impaired by multipath effects
Solution Approach 1:
The patent introduces neural networks as intermediary components that process raw RSSI measurements and transform them into accurate location estimates. The neural networks act as mediators between the wireless communication system and the location determination process, learning to compensate for multipath effects and environmental variations through training data, thereby maintaining accuracy while enabling large-scale deployment
2Quantity of substance
If realistic quantities of objects are tracked, then tracking coverage is improved, but time delays from processing bottlenecks occur
Solution Approach 1:
The system performs preliminary action by pre-training neural networks offline with extensive training data collected from the environment. This preprocessing step creates ready-to-use location determination models that can quickly process real-time RSSI data without requiring complex computations during actual tracking operations, thereby enabling scalable tracking with minimal processing delays
Solution Approach 2:
The patent replaces traditional mechanical signal processing methods with neural network-based computational models. Instead of using complex mathematical algorithms and signal processing techniques that require significant computational resources in real-time, the system uses pre-trained neural networks that have already learned the environmental characteristics, enabling fast and efficient location determination for multiple objects simultaneously
3Device complexity
If traditional location determination methods are used, then system simplicity is maintained, but location indications become stale and inaccurate
Solution Approach 1:
The system implements dynamics by continuously updating and adapting neural network models based on new training data. The neural networks can be retrained periodically or adaptively to account for changes in the environment, such as new obstacles, changing signal patterns, or additional tracking objects. This dynamic adaptation maintains high location accuracy without requiring complete system redesign, balancing complexity with performance
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for real-time location detection consists of three groups of components. Mobile subjects to be tracked are equipped with wireless transceivers capable of sending and optionally for receiving data over pre-determined radio frequency (RF) band(s). Router/base station access point devices are equipped with wireless transceivers capable of sending and receiving data over pre-determined radio frequency (RF) band(s) in order to communicate with mobile units. Routers are combined into specific overlapping router groups, with each group forming a spatial sub-network. System central processing and command station(s) perform data processing and implementation of computational models that determine the mobile unit location. System deployment consists of three phases: collection of training and testing data, network training and testing, and network adaptive maintenance.