Intelligent Waste Sorting and Routing System for Smart Cities

Abstract

Waste management issues have gotten worse due to rapid urbanization, which has resulted in resource waste, inefficiency, and pollution of the environment. To solve these problems, the Intelligent Waste Sorting and Routing System for Smart Cities provides an automated, data-driven solution. While machine learning algorithms categorize waste into recyclable, organic, and non-recyclable categories, IoT-enabled smart bins identify the type of waste and fill levels. A central system receives real-time data and uses it to determine the best collection routes for waste management trucks, reducing carbon emissions, fuel consumption, and operating expenses. Additionally, the system gives city officials analytics to improve their recycling plans and environmentally friendly operations. In contemporary smart cities, this solution enhances productivity, lessens environmental impact, and encourages sustainable urban living by combining intelligent sorting, IoT sensing, and optimized routing.

Introduction

1. Problem Overview

• Rapid urbanization has led to a significant increase in municipal solid waste, overwhelming traditional fixed-schedule waste collection systems.

• Issues with current systems include:

  • Collection of half-filled bins

  • Excessive fuel use, labor, and emissions

  • Poor source-level sorting, contaminating recyclables

• There is a critical need for real-time, intelligent, and sustainable waste management solutions in smart cities.

2. Proposed Solution

An integrated, data-driven system combining:

• IoT-enabled smart bins (monitor fill level, waste type)

• Machine learning (ML) for on-device waste classification

• Dynamic vehicle routing to optimize collection paths

3. System Components

A. Waste Classification

• Uses a Convolutional Neural Network (CNN) (MobileNetV2) to classify waste into:

  • Recyclable, Organic, Non-Recyclable

• Image preprocessing (resizing, augmentation) and training performed with >10,000 images

• Deployed on ESP32-CAM microcontroller using TensorFlow Lite

• Achieved 93% accuracy, inference time <150 ms

B. IoT Sensing and Communication

• Smart bin hardware:

  • Ultrasonic sensor for fill level

  • Load cell for weight

  • ESP32-CAM for image capture

• LoRaWAN used for long-range, low-power communication

• Edge device transmits data when fill level >70%

C. Routing Optimization

• Modeled as a Dynamic Vehicle Routing Problem (DVRP)

• Uses a Genetic Algorithm (GA) with:

  • Order crossover, swap mutation, and tournament selection

• Optimizes for:

  • Distance, time, fuel consumption, and load balancing

• Reacts in real time to bin status updates

4. System Workflow

1. Bins collect real-time sensor and image data

2. ML model classifies waste type

3. Routing engine computes optimal paths

4. Route info sent to drivers via a mobile app

5. Data logged in a central database for monitoring and retraining

5. Implementation Details

Hardware:

• ESP32-based smart bins powered by solar + lithium-ion batteries

• Weatherproof enclosures built using 3D printing

• Raspberry Pi as the LoRaWAN gateway

Software:

• Arduino C++ on ESP32 devices

• Flask + MQTT backend in Python

• React.js dashboard with Google Maps API

• PostgreSQL for data storage

Machine Learning:

• Trained and fine-tuned MobileNetV2 on combined dataset

• On-device inference using TensorFlow Lite

Routing Engine:

• Implemented in Python with dynamic re-routing based on real-time bin data

6. Testing and Results

Simulations:

• Modeled 100 bins across 5 zones using SUMO traffic simulator

• Metrics:

  • Fuel consumption ↓ by ~29%

  • CO? emissions ↓ by ~27%

  • Collection time ↓ by ~35%

Real-world Pilot:

• Deployed 5 smart bins in a test neighborhood for 2 weeks

• Validated:

  • Reliable data transmission

  • On-device ML inference

  • Dynamic routing responsiveness

7. Key Innovations

• Unified system combining classification, sensing, and routing

• Edge ML deployment for low-latency, real-time sorting

• Dynamic route optimization responding to real-world waste levels

Conclusion

This work demonstrates an intelligent waste sorting and routing system that couples IoT sensing, lightweight on-device classification, and dynamic VRP optimization. In simulations and a small field pilot, we observed substantial operational gains—fuel (?29%), CO? (?27%), and time (?35%)—while maintaining high classification reliability. The approach is practical to deploy in modern smart cities and offers a pathway to measurable environmental and cost benefits. Future work will focus on city-scale deployments, predictive scheduling, and resilience features (fault tolerance, connectivity fallbacks), along with richer policy analytics for municipal decision-makers.

References

[1] M. A. Hannan, M. Akhtar, R. A. Begum, A. Hussain, and H. Basri, “IoT-based smart waste bin monitoring and municipal solid waste management system,” IEEE Access, vol. 6, pp. 70216–70223, 2018. [2] A. Kumar, M. Bhatnagar, and A. Singh, “Deep learning based waste classification for smart cities,” in Proc. IEEE Int. Conf. Smart Cities, 2020, pp. 45–50. [3] S. Idwan, M. M. Abdallah, and K. M. Al-Taani, “GA-based optimization for smart waste collection routing in urban areas,” in Proc. IEEE Int. Conf. Internet of Things (iThings), 2021, pp. 560–565. [4] S. Ahmad, M. S. Sarfraz, and H. Kim, “PSO-based vehicle routing optimization for solid waste management,” in Proc. IEEE Conf. Sustainable Computing, 2020, pp. 250–256. [5] A. A. H. Ali, A. M. Abdullah, and M. Ismail, “Design and implementation of IoT-based solid waste monitoring system,” in Proc. IEEE Int. Conf. Control System, Computing and Engineering (ICCSCE), 2019, pp. 246–251. [6] J. K. Thakker and S. Shah, “An IoT based solid waste management system for smart cities,” in Proc. IEEE Int. Conf. Innovative Mechanisms for Industry Applications (ICIMIA), 2020, pp. 310–315. [7] C. J. Deepan, N. Senthilkumar, and S. Manikandan, “Implementation of AI and IoT for smart waste management,” International Journal of Recent Technology and Engineering, vol. 8, no. 5, pp. 1008–1012, 2020. [8] A. A. Wahab, A. Kadir, and M. H. Badiozaman, “Dynamic scheduling in solid waste collection using Internet of Things,” in Proc. IEEE Int. Conf. Smart City Innovations, 2019, pp. 87–92. [9] M. R. Alam and S. V. U. Kumar, “Image-based trash classification using deep learning,” in Proc. IEEE Int. Conf. Data Science and Engineering (ICDSE), 2019, pp. 194–199. [10] S. P. Singh and R. K. Sharma, “Smart bin implementation for smart cities,” in Proc. IEEE Int. Conf. Power, Control, Signals and Instrumentation Engineering (ICPCSI), 2017, pp. 99–104.

Copyright

Copyright © 2025 Praveen Kumar Naidu Rayanki, Bhavishya G. 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.