Optimizing Smart City Infrastructure Using 5G Edge AI with Adaptive Multi-Agent Reinforcement Learning

Rapid urban expansion in current cities creates new challenges regarding effective management of infrastructure combined with resource handling and public service delivery. This research introduces an improved framework to maximize smart city infrastructure through joint use of 5G edge artificial intelligence with Adaptive Multi-Agent Reinforcement Learning (AMARL). A new system connects 5G edge computing to adaptive reinforcement learning agents which optimizes real-time choices for various urban domains including traffic regulation and energy networks alongside environmental detection and public security operations. The placement of AI models at 5G edge nodes cuts latency levels down by more than 80% thereby providing essential ultra-low response times for dynamic urban environments. The implementation of the AMARL framework yielded a performance increase of resource allocation efficiency which enabled an improved utilization rate up to 31 % during disaster relief logistics operations. The decision-making accuracy reached between 13 % and 17% improvement across various scenarios where traffic congestion reached 92<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> accuracy followed by power outage management at 90% accuracy and emergency response reaching 89%. The anomaly detection system based on AI technology proved better at identifying threats in comparison to normal intrusion detection systems while achieving 20% better detection accuracy. The framework demonstrates capability to transform urban infrastructure through its delivery of scalable and efficient as well as secure solutions for smart cities. The investigation delivers important findings to smart city optimization research through its presentation of a flexible and reliable system for urban management.