| Abstract |
Mobile Ad-hoc Networks (MANETs) encounter challenges related to dynamic topologies, energy constraints, and link instability. To enhance performance, a hybrid model integrates Ad-hoc On-demand Multipath Distance Vector (AOMDV), Power-Efficient Gathering in Sensor Information System (PEGASIS), and Deep Q-Network (DQN). OMNeT++, a discrete event simulation framework, facilitates comprehensive data collection, capturing key network parameters such as node density, signal strength, packet transmission rate, residual energy, routing paths, link breakage frequency, and packet delivery status. Stored in scalar and vector files, the dataset enables energy-efficient communication modeling and adaptive routing optimization. Data preprocessing involves cleaning, normalizing, and handling missing values using interpolation techniques while filtering noisy data through a sliding window averaging method. Feature extraction focuses on hop count, queue length, residual energy, and link expiration time, ensuring consistency and facilitating faster convergence. The hybrid model architecture integrates AOMDV for reliable multipath routing, PEGASIS for energy-efficient communication, and DQN for adaptive path optimization. Simulation results indicate a 15% improvement in throughput, a 10% rise in packet delivery ratio, a 20% reduction in end-to-end delay, and a 30% decrease in average energy consumption. Applications extend to military operations, disaster recovery, vehicular networks, IoT sensor networks, and smart city infrastructure, ensuring efficient and adaptive communication. © 2025 IEEE. |
