Distributed Data Collection and Storage Algorithms for Collaborative Learning Vision Sensor Devices with Applications to Pilgrimage

TL;DR

This paper introduces two distributed data storage algorithms for wireless sensor networks to efficiently collect and store data for collaborative learning, with practical deployment in pilgrimage-related monitoring.

Contribution

It proposes novel distributed data storage algorithms (DSA-I and DSA-II) that optimize data dissemination and encoding in resource-constrained sensor networks.

Findings

DSA-I is efficient with known total node count, using flooding with O(n) mixing time.

DSA-II operates without knowing total node count, with encoding complexity O(Cμ^2).

Algorithms perform well in real-world deployment for monitoring in pilgrimage settings.

Abstract

This work presents novel distributed data collection systems and storage algorithms for collaborative learning wireless sensor networks (WSNs). In a large WSN, consider $n$ collaborative sensor devices distributed randomly to acquire information and learn about a certain field. Such sensors have less power, small bandwidth, and short memory, and they might disappear from the network after certain time of operations. The goal of this work is to design efficient strategies to learn about the field by collecting sensed data from these $n$ sensors with less computational overhead and efficient storage encoding operations. In this data collection system, we propose two distributed data storage algorithms (DSA's) to solve this problem with the means of network flooding and connectivity among sensor devices. In the first algorithm denoted, DSA-I, it's assumed that the total number of nodes is known for each node in the network. We show that this algorithm is efficient in terms of the encoding/decoding operations. Furthermore, every node uses network flooding to disseminate its data throughout the network using mixing time approximately O(n). In the second algorithm denoted, DSA-II, it's assumed that the total number of nodes is not known for each learning sensor, hence dissemination of the data does not depend on the value of $n$. In this case we show that the encoding operations take $O(C\mu^2)$, where $\mu$ is the mean degree of the network graph and $C$ is a system parameter. Performance of these two algorithms match the derived theoretical results. Finally, we show how to deploy these algorithms for monitoring and measuring certain phenomenons in American-made camp tents located in Minna field in south-east side of Makkah.