AI-Driven Smart City Infrastructure: An Implementation Framework for Sustainable Urban Development in Emerging Economies

Abstract

Smart city initiatives in emerging economies face distinctive challenges including limited infrastructure, rapid urbanization, fiscal constraints, and the need to deliver services to diverse and frequently underserved populations. This study proposes an artificial intelligence (AI)-driven smart city implementation framework tailored to emerging-economy contexts and evaluates it through pilot deployments in Pune (India), Surabaya (Indonesia), and Penang (Malaysia). The proposed framework consists of five layers: (i) sensing and Internet of Things (IoT); (ii) network and edge computing; (iii) data management; (iv) AI analytics; and (v) application and citizen interface. Specific AI components include YOLOv8-based computer vision for traffic and pedestrian flow analysis (mAP@0.5 = 0.892), LSTM with attention for short-term energy load forecasting (MAPE = 4.7%), BERT-derived sentiment analysis for citizen feedback (F1 = 0.879), and a double deep Q-network (DDQN) for adaptive traffic signal control. Pilot evaluations demonstrate traffic flow improvements of 23.4 per cent, average travel time reductions of 18.2 per cent, energy consumption reductions of 17.8 per cent, and net positive citizen feedback (NPS +14.3 points). The system maintained P95 latency of 187 ms under operational load with horizontal scalability tested up to four times the baseline workload. The study contributes a deployable architectural framework, comprehensive empirical evaluation, and detailed implementation guidance for emerging-economy smart cities