Blockchain Integrated Hierarchical Double Deep Q‐Learning for Optimized Aerial Base Station Deployment in 6G Networks for Smart Cities

ABSTRACT The advent of 6G networks introduces new challenges in the deployment and management of Aerial Base Stations (ABSs) requiring advanced techniques to optimize network performance across multiple conflicting objectives such as coverage, energy efficiency, interference management, and security. This paper presents a novel Blockchain integrated Hierarchical Double Deep Q‐Learning (BH‐DDQL) framework designed to address these challenges by leveraging reinforcement learning for dynamic ABS positioning and operational management. Its hierarchical structure enables a multi‐layered decision‐making process breaking down the ABS deployment problem into manageable sub‐problems, each optimized through reinforcement learning. At the top level, BH‐DDQL handles strategic decisions such as ABS positioning to maximize coverage and minimize interference. At the lower level, it fine‐tunes parameters like power control and resource allocation to enhance energy efficiency and service quality. The BH‐DDQL framework solves the challenge of optimal positioning by utilizing a dynamic swarm‐enhanced mutated slime mold optimization (DSEMSMO) algorithm, which finds effective initial placements for ABSs. To manage energy constraints, the framework incorporates energy‐aware decision‐making within the reinforcement learning process, ensuring that ABSs operate efficiently over extended periods. Interference is minimized through a multi‐objective reward function that balances coverage and interference, ensuring that the ABSs provide optimal service without causing excessive interference to other network elements. Security concerns are addressed by integrating blockchain technology, which secures ABS operations and data through a tamper‐proof log and transparent decision‐making process. The BH‐DDQL framework employs Pareto optimality for deployment configuration, ensuring that no single performance metric is improved at the expense of others. The iterative learning process allows the ABSs to adjust to changing network conditions and environmental factors, securing optimal performance over time. The simulation is conducted in NS‐3 and the numerical results show that the proposed framework achieves a coverage probability of 96.8%, an energy consumption of 500 joules, and a throughput of 10.5 gigabits per second, representing significant advancements over existing systems.