An Energy-Efficient and QoS-Preserving Hybrid Cross-Layer Protocol Design for Deep Learning-Based Air Quality Monitoring and Prediction

Cross-layer design countenances protocols belonging to diverse layers to collaborate and share network status information which ensures that contemporary route is optimal. Extremely reliable communication with least energy consumption can be achieved by adopting cross-layer protocol which is the main intention of this design. Due to its high reliability, cross-layer design is integrated with Internet of Things (IoT)-based real-time application for amplifying flexible layer techniques. However, amalgamation of inadequate layers and communication via interference affects the executing operations in environment. To overwhelm this, hybrid cross-layer protocol for air quality prediction (HCL-AQP) framework is proposed. Initially, the wireless sensor network (WSN) topology is constructed into dendrimer tree structure and then the sensor nodes are clustered by adaptive ball K-means clustering (AB-KC) algorithm for minimizing energy consumption thereby enhancing sensor coverage. After that, the sensor and UAV lifetime is ensured by performing two-level adaptive sleep scheduling using orthogonal frequency division multiple access (OFDMA). Besides, hybrid cross-layer protocol is designed for enhancing communication among several layers. Following that, parallel dual module deep Q-learning (PDMQL) and archerfish hunting optimization (AFHO) algorithm is adopted to implement congestion-aware routing thereby satisfying QoS. Furthermore, once the data are collected, the three stages of pre-processing is accomplished using outlier detection, batch normalization and dynamic interpolation using isolation forest (IF), cross-iteration batch normalization (CBN) and coarse-grained dynamic interpolation (CGDT), respectively. Finally, the significant features are extracted using prickling capsule network (PCapsNet) and air quality is predicted by attention based bi-directional long short-term memory (Att-BiLSTM) based on observed pollutants and meteorological data. The proposed work is conducted in Network Simulator-3.26 and the performance of proposed HCL-AQP framework is enumerated based on several performance metrics in terms of energy consumption, residual energy, end-to-end delay, throughput, packet delivery rate and accuracy.