LOADng-AT: a novel practical implementation of hybrid AHP-TOPSIS algorithm in reactive routing protocol for intelligent IoT-based networks.

Despite that IoT technology provides a promising future for human life, some significant challenges such as routing, security, low-cost equipment, energy consumption, privacy, and reliability can considerably affect its performance. In recent studies, routing has been considered as one of the most critical challenges in IoT due to many existing IoT devices in a network. Selecting a non-optimal path increases collision probability and packet latency which sharply reduces the network performance. Hence, a novel flexible, scalable, and efficient routing protocol named LOADng-AT has been introduced in this study to overcome this challenge. Using LOADng-AT, firstly, taking advantage of the hello message, information about link quality parameters up to two hops is exchanged among nodes. This information includes various parameters such as bi-/unidirectional links, ETX, number of neighbors, link stability, received and residual energy, and SINR, gathered by hello message. Secondly, considering those parameters simultaneously, LOADng-AT efficiently chooses the best possible route based on the hybrid AHP-TOPSIS algorithm. The proposed protocol considerably decreases the RREQ broadcast storm. Moreover, it supports an error recovery path without rerunning a new routing process that noticeably decreases the network routing delay. LOADng-AT can be easily adapted to any QoS parameter with very low complexity which is very important for a delay-sensitive IoT-based network. Finally, in case of noisy condition, the proposed protocol can be efficiently used in low-quality links. Simulation results based on several scenarios in the OMNET ++ simulator show that end-to-end delay (EED) and packet delivery ratio (PDR) parameters significantly improved in the proposed method compared to other similar methods. In other words, using LOADng-AT, the mean values of PDR parameter were 98.64 and 82 percent for network sizes 400 × 200 and 800 × 400 m2, respectively, while the mean values of EED were 0.001 and 0.018 for network sizes 400 × 200 and 800 × 400 m2, respectively. The overall improvement for PDR and EED parameters in all 60 considered scenarios compared to similar methods were 82.26 and 16.6 percent, respectively. [ABSTRACT FROM AUTHOR]