Data-Driven Dynamic Active Node Selection for Event Localization in IoT Applications - A Case Study of Radiation Localization

• Main Authors: Ahmed Alagha, Shakti Singh, Rabeb Mizouni, Anis Ouali, Hadi Otrok
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: IEEE Access
• Subjects: IoT; localization; active node selection; data-driven; radiation detection
• Online Access: https://ieeexplore.ieee.org/document/8625410/

In this paper, the problem of active node selection for localization tasks, on the Internet of Things (IoT) sensing applications, is addressed. IoT plays a significant role in realizing the concept of smart environments, such as in environmental, infrastructural, industrial, disaster, or threat monitoring. Several IoT sensing nodes can be deployed within an area to collect regional information for the purpose of achieving a common contextual goal. Active node selection proves useful in mitigating common IoT-related issues like resource allocation, network lifetime, and the confidence in the collected data, by having the right sensors active at a given time. Current active node selection schemes prove inefficient when adapted to localization tasks, as they- 1) are usually designed for general monitoring, not localization, 2) do not dynamically exploit data readings in the selection process, and 3) are mostly designed for systems with nodes having sensing ranges. To address these challenges, we propose a novel Data-driven active node selection approach that- 1) dynamically uses data readings from current active nodes to select future ones, 2) assesses the area coverage achieved by a group of nodes while considering range-free sensors, 3) considers parameters like residual energy, power cost, and data confidence levels in the selection process, and 4) combines group-based and individual-based selection mechanisms to enhance the localization process in terms of time and power consumption. These considerations are integrated into a two-phase active node selection mechanism that uses genetic and greedy algorithms to select optimum groups for localization tasks. The efficacy of the proposed approach is validated through an example of radioactive source localization by using real-life and synthetic datasets, and by comparing the proposed approach to existing benchmarks. The results demonstrate the ability of the proposed approach to performing faster localization at low energy cost, even with a smaller number of active nodes.