AI Routing Energy Optimization Framework for Adaptive WSN by Using Contiki-Cooja

This research study introduces AI routing, a symbolic AI-based multi-criteria routing protocol designed for energy optimization in wireless sensor networks (WSNs). AI routing, executed in Contiki-Cooja 2.7 with Sky motes and a mobile sink, has a streamlined fuzzy inference technique that calculates a next-hop score based on residual energy, connection quality, expected transmission count (ETX), and data priority. This lightweight method ensures long-term applicability on resource-constrained devices. Simulations involving 100 sensor nodes indicate that AI routing surpasses both traditional and AI-enhanced benchmarks. In comparison to RPL, CRT-LEACH, EEERP-RL, and PSO-based clustering, AI routing enhances network longevity by as much as 38%, attains an average Packet Delivery Ratio (PDR) of 95.1%, and decreases energy expenditure to 0.016 J per delivered packet. The control overhead is reduced, with AI routing incurring around 38.9 KB, which is lower than EEERP-RL's 60.2 KB. AI routing achieves balanced energy consumption, scalability, and resilience by merging symbolic AI rules with mobile sink awareness, eliminating the need for expensive on-device learning. A reproducibility package with code, simulation parameters, and random seeds is included. Here, ‘fuzzy link’ denotes a composite metric of received signal strength indicator (RSSI), ETX, and packet delivery ratio, while ‘energy threshold’ is the minimum 20% residual energy cutoff for forwarding. Future expansions will investigate adaptive weight calibration and hybrid reinforcement learning to enhance flexibility in various WSN applications.