A Deep Inverse Reinforcement Learning for Adaptive 6G Network Optimization and Disaster Response in Smart Cities

Abstract

It is anticipated that the quick development of 6G mobile communication networks will revolutionise smart cities by making intelligent transportation, healthcare, manufacturing, and residential services possible. Resilient 6G Network Optimisation is necessary to guarantee the continuous operation of vital services in the event of technological malfunctions, human mistake, natural disasters, or terrorist attacks, as societies become more reliant on digital infrastructure. The suggested MaximumEnt-InverseRL system uses a thorough data processing pipeline that includes feature engineering, categorical encoding, feature scaling, outlier detection and removal, and managing missing values in order to overcome this difficulty. To enhance prediction performance, a hybrid feature selection approach that combines ensemble learning and CFS-RF is used. Additionally, to facilitate intelligent network decision-making, the Maximum Entropy Principle (MEP) is combined with Reinforcement Learning (RL) and a Deep Neural Network (DNN)-based reward function within an Inverse Reinforcement Learning (IRL) framework. The suggested MaximumEnt-InverseRL model, which achieves an exceptional accuracy of 95.03%, greatly improves 6G Network Optimisation while enhancing Disaster Response in Smart Cities, according to simulation findings. These results show that the suggested strategy provides a strong and flexible way to preserve resilient 6G infrastructure and guarantee the continuation of vital urban services in the event of unexpected disruptions.