DDNSAS: Deep reinforcement learning based deep Q-learning network for smart agriculture system

As the global population continues to grow and environmental conditions become increasingly unpredictable, meeting the demands for food becomes increasingly difficult. To overcome these challenges, smart agriculture has emerged as the key technology. Deep Learning model with Internet of Things (IoT), unmanned aerial vehicle (UAV), and edge–fog–cloud architecture enabled smart agriculture as a key component for next agriculture revolution. In this work, we present two stage end-to-end DRL based smart agricultural system. In stage one, we proposed ACO enabled DQN (MACO-DQN) model to offload task including fire detection, pest detection, crop growth monitoring, irrigation scheduling, soil monitoring, climate monitoring, field monitoring etc. MACO-DQN model offload the task to either edge, fog or cloud networking devices based on latency, energy consumption and computing power. Once the task offloaded to computing devices (edge, fog or cloud), task of prediction and monitoring various agriculture activities is performed at stage two. In stage two, we proposed DRL based DQN (RL-DQN) model for prediction and monitoring agricultural task activities. Finally, we demonstrate experimental findings of our proposed model that represent a marked enhancement in terms of convergence speed, planning success rate, and path accuracy. To evaluate its performance, the method presented in this paper was compared to traditional deep Q-networks-based intensive learning method under consistent experimental conditions. Overall, 98.5% precision, 99.1% recall, 98.1% F-measure, and 98.5% accuracy is obtained when using our proposed methodology and based on the performance results the model outperforms other existing methodologies.