Urban farming, specifically aquaponics, offers a sustainable approach to food production; however, novice practitioners often struggle due to the complexity of the system and a reliance on stable internet connectivity for real-time monitoring. This paper proposes an interactive Augmented Reality (AR) and Digital Twin (DT) framework for IoT-based aquaponics to address these limitations. Developed through a rigorous five-phase research methodology, the system integrates a resilient WebSocket communication architecture with edge-level data buffering, successfully recovering 92.38% of sensor records during simulated network outages. To sustain continuous environmental visualisation during hardware or network failures, a lightweight slope-only linear regression fallback mechanism was implemented, achieving high predictive accuracy for dynamic parameters like temperature and humidity. Furthermore, to accommodate non-technical urbanites (aged 18 to 35), the mobile AR interface incorporates a rule-based mistouch detection algorithm using spatial and temporal thresholds, successfully intercepting 37.25% of interactions as accidental inputs to prevent physical actuation errors. A User Acceptance Test (UAT) with 20 participants yielded a System Usability Scale (SUS) score of 56.63, reflecting the initial cognitive load of spatial AR navigation, yet recorded a high overall user satisfaction rate of 85%. This paper contributes a resilient, fault-tolerant, and user-centric management solution that advances smart agriculture for urban farming initiatives.