Q-learning-based energy-aware parent selection and adaptive Trickle control for RPL in smart city IoT networks

Abstract Smart city applications such as garbage management, water management, traffic control, and streetlight automation demand adaptive, energy-efficient, and scalable communication protocols. RPL (Routing Protocol for Low-Power and Lossy Networks) is an extensively used routing protocol, especially for IoT-based networks that contain constrained nodes with limited energy, memory, and processing power. This work proposes an enhanced RPL-based routing algorithm that combines Q-learning-based energy-aware parent selection with heuristic adaptive Trickle interval control. The Q-learning component selects rank-valid parent nodes using link quality, residual energy, packet delivery, delay, and routing overhead, while a Q-value hysteresis threshold is used to limit unnecessary parent changes. The adaptive Trickle mechanism adjusts the routing-control interval according to short-term network conditions while retaining the standard Trickle suppression rule. The proposed algorithm was evaluated in the Cooja simulator for node densities ranging from 80 to 400 nodes, Tx/Rx ranges of 60–100 m, and Random Walk and Random Waypoint mobility models. The performance was compared with MRHOF, RPL*, and Q-RPL using PDR, latency, average power consumption, energy efficiency, routing overhead, parent switching rate, and end-to-end delay. At 400 nodes and a 60 m Tx/Rx range, the proposed algorithm improved PDR over Q-RPL by 4.2% under RWM and 4.3% under RWPM. Energy efficiency improved by 12.7% and 12.5%, while routing overhead was reduced by 23.5% and 21.4%, parent switching rate by 45.5% and 44.3%, and end-to-end delay by 15.3% and 15.1%, under RWM and RWPM, respectively. These results indicate that combining learned parent selection with adaptive Trickle control improves routing reliability, stability, and energy utilization in dynamic smart city IoT networks.