Distributed Sensor Decision Engine for Low-Traffic Wireless Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wireless sensor networks (WSNs) face challenges in large-scale optimization, scalability, and efficient data processing, particularly in real-time applications, due to the lack of integrated connectivity solutions and redundant data transmission, which hinders efficient edge computing and central processing.
Innovation Solution
A sensing system with a distributed decision engine and gateway that configures sensors in multiple transmission modes, applies intentional time delays, and selectively controls data transmission to optimize bandwidth, power, and computational resources, using a distributed neural network for intelligent data processing and synchronization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If sensors transmit all raw data to the gateway and computing network, then data completeness is improved, but network bandwidth consumption and power usage increase significantly
Solution Approach 1:
The patent segments the data processing function into two parts: edge processing at sensors (filtering and preliminary analysis) and central processing at the gateway. This segmentation allows only essential data to be transmitted, reducing power consumption while maintaining data completeness for critical events.
Solution Approach 2:
The patent applies preliminary action by performing data filtering, aggregation, and preliminary analysis at the sensor edge before transmission. This preliminary processing eliminates redundant data early in the pipeline, reducing network bandwidth consumption and power usage while preserving complete information about significant events.
2Loss of information
If sensors transmit all raw data to the gateway, then data completeness is improved, but network traffic volume increases
Solution Approach 1:
The patent extracts only the essential and relevant data features from raw sensor data at the edge devices before transmission. By taking out only the necessary information (filtered data, aggregated statistics, critical event markers), the system maintains data completeness for decision-making while dramatically reducing network traffic volume.
Solution Approach 2:
Preliminary data filtering, aggregation, and feature extraction are performed at the sensor edge before transmission to the gateway. This preliminary action removes redundant information early, ensuring that only essential data occupies network bandwidth, thus reducing overall traffic while preserving completeness of meaningful information.
3Area of stationary object
If a large number of sensors are deployed to improve coverage, then system coverage is improved, but training the neural network becomes more difficult and computationally demanding
Solution Approach 1:
The patent segments the neural network into distributed edge models at sensors and a centralized gateway model. Each sensor performs local inference with a lightweight model, while the gateway handles more complex processing. This segmentation allows large-scale deployment with extensive coverage while keeping individual training tasks manageable and enabling federated learning approaches.
Solution Approach 2:
The patent enables self-service through distributed edge computing where each sensor performs local data processing and preliminary analysis independently. This self-service capability at the edge reduces the computational burden on the central gateway, allowing large-scale sensor deployment with manageable training complexity through localized intelligence.
4Use of energy by moving object
If edge processing is implemented at sensors to reduce bandwidth usage, then power consumption is reduced, but it may hinder central processing in the cloud
Solution Approach 1:
The patent segments processing into edge functions (filtering, aggregation, preliminary analysis) and central functions (comprehensive analysis, long-term storage, complex inference). This segmentation enables power-efficient edge processing that reduces bandwidth usage while preserving complete raw data and essential features for central processing, maintaining cloud capability through selective data transmission.
Solution Approach 2:
Preliminary data processing at the edge prepares data for efficient transmission to the cloud without losing information needed for central analysis. The edge performs filtering and feature extraction that highlights important patterns, enabling power-efficient transmission while ensuring the cloud receives pre-processed data that maintains its full analytical potential.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A sensing system comprises a plurality of sensors; a gateway device configured to exchange data with the sensors; and a distributed decision engine comprising one or more sensor portions provided at one or more of the sensors and a gateway portion provided at the gateway device. Each of the sensors is configurable to operate in a plurality of different transmission modes wherein the sensor transmits primary sensor data and/or outputs of the sensor portion of the decision engine provided at said sensor; and the gateway is arranged to selectively control the operation of the sensors and to configure the sensor transmission mode.