IoT Data Transmission Optimization Using SVR Attribute Pruning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data transmission methods in IoT networks are inefficient in spectral usage and storage optimization, leading to computational delays and increased susceptibility to errors in real-time data transmission over wireless channels, particularly in time-critical applications.
Innovation Solution
A system and method that utilize Support Vector Regression (SVR) models to construct sets of attributes based on correlation, with error bounds for predicting non-base attributes, and dynamic real-time data pruning, optimizing data transmission by transmitting only necessary model parameters and retraining data based on error thresholds and SINR awareness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If data compression frameworks through statistical methods are used to compress offline or storage data, then data volume is reduced, but real-time data compression over wireless channel is not addressed
Solution Approach 1:
The patent segments data into base attributes and non-base attributes, where base attributes are transmitted and non-base attributes are reconstructed at the receiver using SVR models. This segmentation enables real-time processing by transmitting only essential data while reconstructing the rest, thus reducing data volume without significant delay.
Solution Approach 2:
The patent performs preliminary action by training SVR models offline using historical data. The trained models are then deployed at the receiver end to reconstruct non-base attributes in real-time. This preliminary training phase separates the computationally intensive model training from the real-time transmission phase, enabling efficient real-time data compression.
2Quantity of substance
If wavelet compression is used to compress and re-construct transmitted data, then data volume is reduced, but significant delays are introduced in the system
Solution Approach 1:
The patent changes the fundamental parameter of data representation by using Support Vector Regression models instead of traditional wavelet transformation. This parameter change in the compression approach allows for faster reconstruction times while maintaining data fidelity, thus reducing system delay compared to wavelet-based methods.
3Device complexity
If principal component analysis and discrete cosine transform are used for compression of IoT data, then data dimensionality is reduced, but real-time compression is not achieved
Solution Approach 1:
The patent extracts only the most important features (base attributes) from the IoT data and transmits them, while the remaining attributes (non-base) are reconstructed at the receiver. This extraction approach reduces data dimensionality and enables real-time compression by focusing computational resources on identifying and transmitting only the essential data components.
4Quantity of substance
If only model parameters are transmitted to reduce data redundancy, then transmission bandwidth is saved, but susceptibility to channel errors increases
Solution Approach 1:
The patent implements beforehand cushioning by transmitting both model parameters and a subset of actual training data. This provides a safety buffer against channel errors, as the receiver can use the transmitted training data to retrain or refine the SVR models if reconstruction accuracy deteriorates, thus cushioning against the harmful effects of wireless channel errors.
Solution Approach 2:
The patent incorporates feedback mechanisms where the receiver evaluates reconstruction accuracy and can request retransmission of training data or model parameters if errors are detected. This feedback loop enables dynamic adaptation to channel conditions, maintaining reliability by adjusting the amount of transmitted data based on actual reconstruction performance.
Data Source
AI summary
Disclosed are system (100) and method (300) for optimizing data transmission in communication network. System comprises Internet of Thing (IoT) device node (102) having sensors for capturing sensor data and controller (106) configured at the IoT device node for constructing set of attributes from the sensor data, such that each set of attribute comprises sensor data based on correlation. Support Vector Regression (SVR) models for pruning the sensor data at IoT device node (102) is defined. Values for each attribute in set of attributes predicted based on predicting run-time errors in values according to the SVR models defined for each of base attribute and the non-base attribute. The run-time errors for each of base attributes and non-base attributes gets compared with threshold error values. The IoT device node (102) transmits either model parameters for attributes or training data along with the set of attributes by transceiver (108) to the data collector node (104).