Handling Node Discovery Problem in Fog Computing using Categorical51 Algorithm With Explainability
Abstract
Node discovery problem deals with the issue of determining the addition of computing nodes to the fog computing platform. It is a difficult issue since the fog computing system leverages services at the proximity of computing devices with diverse resource capability. Categorical51 (C51) algorithm based on deep-Q-network is used to perform computation over discrete action space environment. Here an explainable C51framework is designed for solving the node discovery problem. The fog nodes are classified either as reliable or as unreliable. The reasons are identified for this classification to enhance the transparency and explainability of the framework. The policies generated are of high quality as the policy is updated less frequently. Mathematical modeling reveals that the performance of proposed C51framework is good in terms of Quality of Service (QoS) and Quality of Security (QoSec) parameters. The performance of the proposed framework is tested against one of the existing works using iFogSim simulator. The performance is found to be promising with respect to performance parameters such as fault tolerance, energy consumption, speedup, and node placement accuracy.
Publication Title
2023 IEEE World AI IoT Congress, AIIoT 2023
Recommended Citation
Krishnamurthy, B., Shiva, S., & Das, S. (2023). Handling Node Discovery Problem in Fog Computing using Categorical51 Algorithm With Explainability. 2023 IEEE World AI IoT Congress, AIIoT 2023, 210-215. https://doi.org/10.1109/AIIoT58121.2023.10174564
