Fog-GMFA-DRL: Enhanced deep reinforcement learning with hybrid grey wolf and modified moth flame optimization to enhance the load balancing in the fog-IoT environment

• Published in: Advances in engineering software (1992) Vol. 174; p. 103295
• Main Authors: Gupta, Shally, Singh, Nanhay
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.12.2022
• Subjects: GMFA; GWO; IoT; Load balancing; Reinforcement learning; Resource allocation; GMFA; Load balancing; GWO; Resource allocation; Reinforcement learning; IoT
• ISSN: 0965-9978
• DOI: 10.1016/j.advengsoft.2022.103295

•Internet of Things (IoT) can facilitate a plethora of data transactions among various servers.•Due to the immense growth of the IoT, many heterogeneous, wireless or wired, sensing and control devices are being used in different real-world scenarios.•The proposed model is based on dynamic load balancing in fog-IOT environment by utilizing a novel hybrid Grey Wolf Optimization with modified Moth Flame algorithm (GMFA).•The DQN based deep reinforcement learning approach performance is enhanced with GMFA algorithm, and this combined strategy is named as GMFA-DQN. Internet of Things (IoT) can facilitate a plethora of data transactions among various servers. In the IoT, fog servers are utilized to achieve effective data transactions from dynamic devices. However, load balancing is still a significant task researcher mainly focus on mitigating the load balancing issue. Some virtual machines may be overloaded when other virtual machines are idle due to a bad scheduling policy. Therefore, the proposed model is based on dynamic load balancing in a fog-IOT environment by utilizing a novel hybrid Grey Wolf Optimization (GWO) with the Modified Moth Flame algorithm (MMFA). In addition, the GMFA mainly helps to enhance Deep reinforcement learning (DRL). The performance of the actor-critic based deep reinforcement learning (DRL) approach is enhanced with the GMFA algorithm, and this combined strategy is named GMFA-DRL. RL offers several advantages regarding resource allocation issues, and simulations demonstrate that it performs better than reactive techniques. The proposed GMFA-DRL approach is implemented through the Python-based platform-Jupyter. The performance is evaluated using performance matrices such as Throughput, Latency, Makespan, Load Balancing Level (LBL), and Energy Consumption. The simulation results illustrate that the proposed model achieves high Throughput, low energy consumption, minimum Latency, minimum Makespan, and load balancing results. Therefore, the proposed approach can be proven more effective than the existing technique.