In an IoT network, routing protocol refers to a single path routing protocol, and each node transmits its traffic directly to a preferred parent selected exclusively based on the adopted Objective Function (OF). Initially, RPL is designed for low traffic scenarios; due to congestion problems in nodes, many performance issues arise, such as load imbalance, high packet loss, and fast energy depletion of bottleneck nodes under heavy traffic load, and also it degrades the network performance.
   Thus, congestion-aware dynamic routing protocol in IoT network is necessary. Moreover, incorporating a load balancing mechanism in RPL routing is essential for overcoming the poor quality of service (QoS) and maintaining efficient network performance in terms of delay and reliability. Thus, an effective solution for overcoming the congestion issue in RPL, the deep learning model, has emerged. Deep Reinforcement Learning-based routing strategy effectively tackles the congestion problem in RPL networks. Each node adopts Q-learning to select its preferred parent based on a composite feedback function.