Edge-Node Compressive Sensing with Adaptive Signal-Energy Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed computing systems face challenges in accurately and efficiently analyzing sensor data at resource-constrained edge nodes due to limited computation resources and low power bandwidth, leading to data degradation during compression and transmission.
Innovation Solution
The method involves computing a signal energy indicator to dynamically adjust compression parameters such as sampling frequency and window length, allowing the system to autonomously adapt to changing conditions while maintaining accuracy and minimizing bandwidth and computing resources, using adaptive window compression and decompression schemes based on energy content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If data is increasingly compressed at edge nodes to reduce transmission bandwidth and computational resources, then resource efficiency is improved, but data accuracy after decompression deteriorates
Solution Approach 1:
The patent implements dynamic compression parameter adjustment based on signal characteristics. The system continuously monitors signal properties and adapts compression parameters (such as transformation type, quantization levels, and sampling rates) in real-time, allowing optimal balance between compression ratio and reconstruction accuracy for different data conditions.
Solution Approach 2:
The system changes compression parameters based on signal energy and characteristics. Different compression algorithms and parameter sets are selected dynamically - for example, using DCT with specific quantization matrices for certain signal types, or switching between different transformation domains based on the measured signal properties, thereby maintaining accuracy while achieving compression.
2Quantity of substance
If compression is performed at resource-constrained edge nodes, then transmission bandwidth is reduced, but computational complexity at edge nodes increases
Solution Approach 1:
The patent extracts and implements only the essential compression functionality at edge nodes, performing lightweight signal transformation and parameter optimization locally, while delegating more complex reconstruction and analysis tasks to centralized servers with greater computational resources. This distribution of computational complexity resolves the contradiction between local compression needs and edge node capabilities.
Solution Approach 2:
The system introduces an intermediary compression layer at edge nodes that prepares data for efficient transmission without requiring full decompression capability at the edge. The intermediary performs optimized preprocessing and compression using adapted algorithms that are computationally feasible for resource-constrained devices, while maintaining compatibility with centralized reconstruction systems.
3Speed
If faster compression is implemented to meet timely transmission requirements, then transmission speed is improved, but compression accuracy deteriorates
Solution Approach 1:
The system implements periodic adjustment of compression parameters based on signal characteristics and transmission requirements. It uses periodic signal analysis to determine optimal compression settings, switching between different compression strategies at appropriate intervals to maintain accuracy while achieving timely transmission.
Solution Approach 2:
The patent applies partial compression strategies where not all data requires the same level of compression. Critical signal components are compressed with higher fidelity while less important components use more aggressive compression. This selective approach achieves timely transmission of essential information without sacrificing overall reconstruction accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system and method for compressive sensing using edge nodes of a distributed computing network. The method includes collecting a raw data signal continuously by a sensor of the edge node. A signal energy indicator is dynamically updated that quantifies an energy distortion in the raw data signal. One or more compression characteristics are determined as a function of the signal energy indicator as the signal energy indicator is updated. The raw data signal is subsampled in accordance with current values of the one or more compression characteristics to create a compressed data signal. An output is transmitted that includes the compressed data signal to a centralized node.