Autonomous Data Acquisition in the Hierarchical Edge-Based MCS Ecosystem

Abstract

Mobile crowdsensing (MCS) is a human-driven sensing paradigm that empowers ordinary citizens to use their mobile devices and become active observers of the environment. Due to the large number of devices participating in MCS tasks, MCS services generate a huge amount of data which needs to be transmitted over the network, while the inherent mobility of users can quickly make information obsolete, and requires efficient data processing. Since the traditional cloud-based architecture may increase the data propagation latency and network traffic, novel solutions are needed to optimize the amount of data which is transmitted over the network. In our previous work we have shown that edge computing is a promising technology to decentralize MCS services and reduce the complexity of data processing by moving computation in the proximity of mobile users. In this paper, we introduce a novel approach to reduce the amount of redundant data in the hierarchical edge-based MCS ecosystem. In particular, we propose the usage of Bloom filter data structure on mobile devices and edge servers to enable users participating in MCS tasks to make autonomous informed decisions on whether to contribute data to the edge servers or not. Bloom filter proves to be an efficient technique to obviate redundant sensor activity on collocated mobile devices, reduce the complexity of data processing and network traffic, while in the same time gives useful indication whether MCS data is valuable at a certain location and point in time. We evaluate Bloom filter with respect to filter size and probability of false positives, and analyze the number of lost data readings in relation to expected number of different elements. Our analysis shows that both filter size and error rate are sufficiently small to be used in MCS.