AI-Powered Civic Issue Reporting and Resolution System

Abstract

Urban local bodies and municipal corporations frequently face challenges in efficiently managing citizen complaints related to public infrastructure such as potholes, garbage accumulation, broken streetlights, fallen trees, and damaged road signs. Traditional complaint systems relying on telephone hotlines or manual web portal submissions are often slow, error-prone, and opaque. This paper presents an AI-powered web-based civic issue reporting and resolution system designed to simplify and automate the entire complaint lifecycle. Citizens upload an image of an urban problem through a user-friendly web interface; the system applies a custom-trained YOLOv8s convolutional object detection model to automatically identify and classify the issue, and uses the Qwen/Qwen2.5-7B-Instruct large language model to generate a professionally worded maintenance description, eliminating the need for citizens to compose reports manually. On the administrative side, municipal officials access a role-based management dashboard displaying all reported issues with GPS coordinates on an interactive map, and can acknowledge, assign, and resolve complaints. Citizens track complaint status in real time through three lifecycle stages—Reported, Accepted, and Resolved—ensuring full transparency and accountability. The YOLOv8s model was trained on a custom dataset of 32,960 annotated images across five urban issue categories and achieved an overall test-set mean Average Precision at IoU 0.50 (mAP50) of 0.787. The platform is built on a React.js frontend, Spring Boot REST API backend, PostgreSQL database, and Python AI microservices, accessible from any browser-capable device without installation.

Introduction

This paper presents an AI-powered Smart Civic Issue Reporting and Management Platform designed to improve urban infrastructure maintenance and citizen complaint handling. Rapid urbanization has increased problems such as potholes, fallen trees, illegal garbage dumping, damaged road signs, and non-functional streetlights, while existing complaint systems remain slow, manual, and inefficient.

The proposed platform simplifies reporting by allowing citizens to upload a photo, confirm GPS location, and submit a complaint. Using YOLOv8s object detection and the Qwen2.5-7B-Instruct language model, the system automatically identifies the issue type and generates a professional maintenance description. Municipal officials access complaints through a structured dashboard that provides issue categorization, map-based location information, and complete lifecycle tracking from submission to resolution.

The system is implemented as a full-stack web application using React.js, Spring Boot, PostgreSQL, and AI microservices developed in Python. Security is ensured through JWT-based authentication and role-based access control. The platform supports both citizens and municipal officials while remaining accessible through any web browser.

The AI component uses a custom-trained YOLOv8s model trained on over 26,000 urban infrastructure images across five categories: potholes, garbage, fallen trees, electrical poles, and road signs. The model performs real-time object detection, while the Qwen language model automatically generates concise maintenance reports. Parallel AI processing reduces response time and improves user experience.

A review of existing civic complaint systems revealed several limitations, including restricted issue categories, lack of automated description generation, platform dependence on Android applications, and limited complaint tracking. The proposed system addresses these gaps through multi-class object detection, AI-generated descriptions, cross-platform accessibility, and transparent issue lifecycle management.

The platform follows a three-tier architecture enhanced with a dedicated AI services layer and was developed using the Agile methodology. Overall, the system aims to improve municipal responsiveness, reduce manual workload, enhance transparency, and strengthen citizen engagement in urban governance through AI-driven automation and smart city technologies.

Conclusion

This paper presented an AI-powered civic issue reporting and resolution platform that addresses the well-documented inefficiencies in traditional manual complaint management systems used by urban local bodies and municipal corporations. All six primary objectives established in the introduction have been fully met, and the results obtained exceed the performance benchmarks reported by comparable prior systems in the literature. The core technical contribution of this work is a custom-trained YOLOv8s object detection model capable of identifying five categories of urban infrastructure problems—potholes, fallen trees, electrical poles, road signs, and garbage—achieving an overall test-set mAP50 of 0.787 with stable convergence during training across 60 epochs without overfitting. Compared with prior systems such as CitySolution, which rely on single-label image classification, the proposed approach performs object detection, enabling both classification and localization of multiple infrastructure issues within a scene. The second major contribution is the integration of the Qwen/Qwen2.5-7B-Instruct large language model for automated maintenance report generation. No prior work in the surveyed literature combines real-time visual detection with natural language generation in a civic reporting context. By dispatching both the YOLO classification request and the LLM description request in parallel using JavaScript\'s Promise.all, the system delivers a fully populated report form within the time a citizen would otherwise spend typing a description manually. The platform delivers a complete and transparent issue lifecycle. From the moment a citizen submits a report, the complaint is publicly visible on the dashboard with a red status badge. When a municipal official accepts the issue, the status transitions to accepted with a yellow badge. Upon resolution, the complaint is marked resolved with a green badge and a timestamp is recorded. After 10 minutes the resolved issue is removed from the public dashboard, keeping the feed focused on active problems while preserving the full history in the citizen\'s personal reporting log. The role-based access control system built on JSON Web Tokens and Spring Security ensures that citizens and officials each interact only with the functionality appropriate to their role. The three-step OTP-based registration flow adds a layer of identity verification that reduces the risk of false reports. Functional testing confirmed correct behavior across all critical user journeys. The system performed consistently across Chrome, Firefox, and Safari on both desktop and mobile viewports. Future work can extend the system in several directions: (1) development of a mobile application with real-time camera capture and push notifications; (2) expansion of the AI detection model to cover additional urban issue categories such as drainage blockages, dead animals, and mosquito-breeding sites; (3) integration with IoT sensors embedded in street infrastructure for automatic issue detection and reporting; (4) implementation of data analytics dashboards for predictive maintenance and hotspot identification; and (5) large-scale deployment across multiple municipal corporations with a multi-tenant architecture to support concurrent cities.

References

[1] F. Shama, A. Aziz, and L. B. M. Deya, \"CitySolution: A Complaining Task Distributive Mobile Application for Smart City Corporation Using Deep Learning,\" arXiv preprint arXiv:2410.12882, Oct. 2024. [2] X. Kong, X. Liu, B. Jedari, M. Li, L. Wan, and F. Xia, \"Mobile Crowdsourcing in Smart Cities: Technologies, Applications, and Future Challenges,\" IEEE Internet of Things Journal, vol. 6, no. 5, pp. 8095–8113, Oct. 2019. [3] D. Walwadkar, J. Patil, M. Hussain, and S. Yadav, \"Smart Civic Issue Reporting System,\" International Journal of Advanced Research in Science, Communication and Technology (IJARSCT), vol. 2, issue 1, pp. 248–254, Mar. 2022. [4] Dinesh Kumar, \"Smart City App for Citizen Complaint Management,\" Indian Journal of Engineering Research Networking and Development, vol. 2, issue 05, May 2025. [5] N. Thoutam, R. Gujar, K. Patil, P. Sananse, and A. Shah, \"Android Based Disaster Management Application Using Machine Learning,\" International Journal of Scientific Research in Engineering and Management (IJSREM), vol. 7, issue 11, DOI: 10.55041/IJSREM26814, Nov. 2023. [6] G. Jocher et al., \"Ultralytics YOLOv8 — Object Detection, Segmentation and Classification,\" GitHub Repository, https://github.com/ultralytics/ultralytics, 2023. [7] Alibaba Cloud Research, \"Qwen2.5: A Large Language Model Series,\" Qwen Technical Report, Alibaba Cloud, 2024. [8] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, \"You Only Look Once: Unified, Real-Time Object Detection,\" Proc. IEEE CVPR, pp. 779–788, 2016. [9] A. Vaswani et al., \"Attention Is All You Need,\" Advances in Neural Information Processing Systems (NeurIPS), 2017. [10] T. Brown et al., \"Language Models are Few-Shot Learners,\" NeurIPS, vol. 33, pp. 1877–1901, 2020. [11] A. Howard et al., \"MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications,\" arXiv:1704.04861, 2017. [12] G. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, \"YOLOv4: Optimal Speed and Accuracy of Object Detection,\" arXiv:2004.10934, 2020. [13] H. Touvron et al., \"Llama 2: Open Foundation and Fine-Tuned Chat Models,\" arXiv:2307.09288, 2023. [14] R. Banks and A. Lee, \"Spring Boot Reference Documentation,\" VMware Pivotal, https://spring.io/projects/spring-boot, 2023. [15] J. Walnes, \"React: A JavaScript Library for Building User Interfaces,\" Meta Open Source, https://reactjs.org, 2023.

Copyright

Copyright © 2026 K. Sindhu Abhirami, K. Krishna Sai, V. Pardha Siva Sitha Rama Reddy, P. Srinivasa Reddi. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.