arXiv:2609.08527v1 [cs.NI] 8 Sep 2026
CleanCity-BinSense: An IoT-Enabled Smart Waste Management System with Configurable Real-Time Fill Monitoring and Nearest-Neighbor Route Optimization Mohammad Adnan Kabir∗
Intifad Muhammad Sayeed
[email protected] Islamic University of Technology Gazipur, Bangladesh
[email protected] Jahangirnagar University Dhaka, Bangladesh
Abstract
CCS Concepts
Urban waste management in developing cities is often hindered by inefficient fixed-schedule collection, overflowing waste bins, and unnecessary fuel consumption caused by non-optimized collection routes. This paper presents CleanCity-BinSense, a low-cost, endto-end IoT-enabled smart waste management system designed to support scalable real-time waste monitoring and demand-driven collection. The proposed system integrates a solar-powered sensor node equipped with an ultrasonic sensor for real-time bin fill-level monitoring. A key contribution of the system is a configurable sensing model based on two calibration parameters, FULL_DISTANCE and EMPTY_DISTANCE, enabling deployment across bins of varying sizes and geometries without requiring firmware modification. Fill percentage is computed using a geometry-configurable linear normalization algorithm validated through hardware experiments with a mean absolute error (MAE) of 0.38 cm, remaining within the manufacturer-specified tolerance of the sensor hardware. Sensor readings are transmitted via Wi-Fi to a centralized web platform that provides role-based dashboards for administrators, operators, and drivers, along with a public real-time bin-status map. The system further incorporates a lightweight nearest-neighbor route planning algorithm using SQL Server’s spatial function to generate proximity-based collection routes with low computational overhead. Experimental evaluation demonstrates an average endto-end system latency of 5.3 seconds, dominated primarily by the configurable sensing interval rather than network overhead, while route generation for typical urban collection zones completes in under 100 ms. These results demonstrate the feasibility of deploying a low-cost, configurable, and infrastructure-light smart waste management system capable of supporting heterogeneous urban waste networks in resource-constrained environments such as Dhaka, Bangladesh.
• Computer systems organization → Embedded systems; Sensor networks; • Applied computing → Environmental sciences.
∗ Both authors contributed equally to this research.
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected]. ICCA 2026, Dhaka, Bangladesh © 2026 Copyright held by the owner/author(s). Publication rights licensed to ACM. ACM ISBN 979-8-XXXX-XXXX-X/26/10 https://doi.org/10.1145/XXXXXXX.XXXXXXX
Keywords Internet of Things, Smart Waste Management, ESP8266, Ultrasonic Sensor, Smart City, Route Optimization, Real-Time Monitoring, SolarPowered IoT, Nearest-Neighbor Algorithm, Configurable Sensing ACM Reference Format: Mohammad Adnan Kabir and Intifad Muhammad Sayeed. 2026. CleanCityBinSense: An IoT-Enabled Smart Waste Management System with Configurable Real-Time Fill Monitoring and Nearest-Neighbor Route Optimization. In Proceedings of 4th International Conference on Computing Advancements (ICCA 2026). ACM, New York, NY, USA, 8 pages. https://doi.org/10.1 145/XXXXXXX.XXXXXXX
1
Introduction
Rapid urbanization in developing countries has intensified the challenges of municipal solid waste (MSW) management. In cities such as Dhaka, Bangladesh, one of the most densely populated cities in the world, waste collection vehicles follow fixed schedules largely independent of actual bin occupancy [1]. This creates two major inefficiencies: bins overflow before collection arrives, creating public health hazards, while vehicles are dispatched to largely empty bins, wasting fuel and driver time. The Internet of Things (IoT) paradigm offers a compelling solution. By embedding sensors in bins and connecting them to a centralized platform, city authorities can shift from time-based to demand-based collection. Previous studies have demonstrated that such systems can reduce collection costs by 20–40% in pilot deployments [2, 3]. However, most existing systems share a critical practical limitation: sensing ranges are hardcoded for a specific bin geometry, limiting deployment across heterogeneous bin infrastructures without modifying and re-flashing firmware. This severely limits real-world scalability where bins of varying heights and capacities coexist on the same city network. This paper presents CleanCity-BinSense, an integrated IoTenabled waste management system that addresses these limitations while improving operational efficiency and scalability. The key contributions are: • A solar-powered IoT sensor node in a custom 3D-printed enclosure with a configurable sensing range, operators set two firmware parameters, FULL_DISTANCE (the sensor reading when the bin is at full capacity) and EMPTY_DISTANCE (the reading when the bin is empty) per bin to match any
ICCA 2026, October 15–16, 2026, Dhaka, Bangladesh
bin geometry, enabling uniform deployment across heterogeneous bin infrastructure • A calibrated, geometry-independent linear fill-percentage algorithm validated with real hardware (MAE 0.38 cm). • A full-stack web platform with role-based dashboards (Admin, Operator, Driver) and a public bin map, operating without mandatory cloud dependency. • An automated nearest-neighbor route optimization algorithm using SQL Server’s STDistance() geodesic distance function for accurate proximity-based stop ordering, with sub100 ms computation time enabling real-time responsive planning. • A complete, physically built and tested prototype demonstrating end-to-end integration from sensor to optimized driver route. The remainder of this paper is organized as follows. Sections 2–6 present the related work, system design, implementation, and experimental evaluation, while Sections 7 and 8 discuss limitations and conclude the paper.
2
Background & Prior Research
Sensor-Based Fill Detection: Gutierrez et al. [1] proposed one of the earliest IoT waste bin systems using infrared sensors and GSM data transmission. While effective for proof-of-concept, GSM introduces recurring operational costs that are unsuitable for largescale city deployments. Khan et al. [8] proposed a knapsack-based IoT-assisted waste collection system that prioritizes bins by toxicity and fill level, demonstrating reduced collection visits in simulation, but without per-bin sensing configuration. None of these systems expose a configurable sensing range, a critical limitation where bins of varying heights coexist on the same city network. CleanCity-BinSense addresses this gap by making the sensing range fully configurable per bin and integrating it with a complete rolebased management web platform. Advanced Optimization Techniques: Jerbi et al. [6] proposed a sophisticated optimization framework that combines IoT sensors with a Dynamic Bald Eagle Search Optimization algorithm (DBESO) and a Kernel-based Extreme Learning Machine (KELM) for accurate waste status prediction. Although their approach delivers strong predictive performance in simulation, it is evaluated without physical hardware deployment and requires specialized machine learning expertise and computational infrastructure that may be impractical for resource-constrained city environments [7]. Ahmed et al. [7] conducted a comprehensive review of AI and IoT architectures for municipality waste management, identifying real-time sensor integration, role-based access, and demand-based scheduling as the most impactful practical requirements for developing city deployments. CleanCity-BinSense directly targets these requirements with a lightweight, physically deployed architecture. Route Optimization: Anagnostopoulos et al. [3] surveyed a dynamic routing algorithm based on real-time fill data, demonstrating significant travel distance reductions. Longhi et al. [2] integrated fill data with external GIS software for route planning. Addas et al. [9] deployed an IoT and cloud analytics system across 10 locations in Lahore with dynamic route optimization, achieving
Kabir and Sayeed
measurable improvements in route efficiency and fuel consumption. However, these approaches rely on external GIS infrastructure [1, 2] or third-party cloud analytics platforms [9], introducing additional infrastructure dependencies that increase deployment complexity and cost. CleanCity-BinSense embeds proximity-based route optimization directly into the operator web platform using SQL Server spatial functions, requiring no external GIS infrastructure or cloud dependency. Platform and Dashboard Design: Folianto et al. [4] presented SmartBin for Singapore, focusing on sensor hardware and a wireless mesh network; their system does not provide role-differentiated access for operators and drivers, nor route optimization capability. Chowdhury et al. [5] specifically targeted Dhaka using NodeMCU and the ThingSpeak cloud platform for data storage and visualization. However, ThingSpeak dependency means that the system requires continuous internet connectivity and cannot operate if the cloud service is unavailable, a significant risk in environments with unreliable internet infrastructure. CleanCity-BinSense operates on a locally-hosted server, requiring only a site-local network between bins and the server, and provides role-based dashboards absent from both prior systems. Solar-Powered IoT: Solar-powered smart waste monitoring has been explored as a practical direction for smart city planning [16]. Our system integrates solar power into the bin sensor unit for sustainable autonomous operation across diverse deployment sites. Table 1 compares CleanCity-BinSense with key prior works across six practical deployment dimensions. Table 1: Comparison with related smart waste management systems. HW = hardware prototype, Dash = web dashboard, Route = route optimization, Cfg. Range = configurable sensing range per bin. System
HW
Solar
Dash
Roles
Route
Cfg.
Gutierrez et al. [1] Folianto et al. [4] Chowdhury et al. [5] Khan et al. [8] Jerbi et al. [6] Addas et al. [9] CleanCity-BinSense
✓ ✓ ✓ × × ✓ ✓
× × × × × × ✓
× ✓ ✓ × × × ✓
× × × × × × ✓
× × × ✓ ✓ ✓ ✓
× × × × × × ✓
Unlike previous studies that focus on individual aspects such as fill-level detection, route optimization, or cloud-based monitoring, CleanCity-BinSense integrates these capabilities into a single deployable smart waste management platform. The system introduces several practical innovations: • Configurable per-bin sensing range for bins of varying dimensions. • Role-based web dashboards for administrators, operators, and drivers. • Built-in SQL Server spatial route optimization without external GIS or cloud services. • Self-hosted, solar-powered architecture for reliable and sustainable municipal deployment. Together, these features provide a practical, scalable solution that addresses several deployment limitations identified in existing smart waste management systems.
CleanCity-BinSense
3
ICCA 2026, October 15–16, 2026, Dhaka, Bangladesh
System Architecture
• ESP8266 Microcontroller: Serves as the primary processing unit, providing Wi-Fi connectivity, GPIO control, and embedded networking capability for transmitting sensor readings to the web platform [12, 13]. • HC-SR04 Ultrasonic Distance Sensor: Measures the distance between the bin lid and the waste surface. The sensor operates within a range of 2–400 cm, with an approximate resolution of 0.3 cm and an operating frequency of 40 kHz. The effective sensing range is calibrated per bin according to its physical dimensions. • 4-Digit 7-Segment LED Display: Displays the current fill percentage locally, or “FULL” when the bin reaches capacity, without requiring network connectivity. • Solar Panel with Charging Circuit: Provides autonomous power supply, eliminating the need for wired electrical infrastructure and enabling flexible deployment. • Custom 3D-Printed Enclosure: A PLA-printed two-part enclosure designed with dedicated cutouts for the ultrasonic sensor, LED display, and cable routing.
CleanCity-BinSense follows the canonical three-layer IoT architecture [10], the most widely adopted foundational model in IoT system design, as shown in Figure 1: the Perception Layer (physical sensing), the Network Layer (Wi-Fi communication), and the Application Layer (web platform and route optimization). Each bin is registered in the system with its GPS coordinates, physical capacity specifications, and its own calibrated FULL_DISTANCE and EMPTY_DISTANCE values. This per-bin configuration enables the web platform to correctly interpret fill percentages from bins of different heights, making the system deployable across heterogeneous urban bin infrastructure without any hardware changes.
4.2
3D-Printed Enclosure Design
The enclosure was designed in CAD software and 3D printed in PLA. Figure 2 shows the three-part CAD design: (a) the lid featuring a recessed sliding channel for tool-free assembly, (b) the main body with two circular ports for the ultrasonic transducers and internal PCB mounting clips, and (c) the rear view with keyhole mounting slots for wall or bin-lid installation.
Figure 1: Three-layer IoT architecture of CleanCityBinSense showing perception, network, and application layers with the route optimization module. Perception Layer: Each smart bin is fitted with a sensor unit at the lid. The HC-SR04 ultrasonic distance sensor [11] emits a pulse and measures the echo return time from the waste surface. The microcontroller processes this signal, computes the fill percentage using its calibrated range parameters, and displays the status on a 7-segment LED display. Network Layer: The microcontroller connects to the site-local Wi-Fi infrastructure (e.g., a municipal access point or premises network) and transmits sensor readings to the web platform server via HTTP POST every 5 seconds. The payload includes distance, fill percentage, and bin identifier. Application Layer: The web platform server stores readings in SQL Server, serves role-specific dashboards (Admin, Operator, Driver), runs route optimization on demand, and serves a public bin map. A dedicated /data endpoint receives all hardware readings.
4 Hardware Design 4.1 Components The sensor unit comprises:
Figure 2: 3D CAD design of the sensor enclosure: (a) lid with sliding channel, (b) main body with ultrasonic sensor ports and internal PCB clips, (c) rear view showing keyhole mounting slots and internal cable management channel.
4.3
Assembled Prototype
Figure 3 shows the fully assembled prototype. The front face exposes both ultrasonic transducers and the LED display showing live fill percentage or “FULL” at maximum capacity.
4.4
Configurable Fill Level Calculation
A key design principle of CleanCity-BinSense is that the sensing range is not hardcoded. The firmware exposes two operator-configurable parameters: • FULL_DISTANCE (𝐷 𝑓 𝑢𝑙𝑙 ): the distance reading (cm) when the bin is completely full. Set to match the bin’s full-state geometry (e.g., waste level just below the sensor).
ICCA 2026, October 15–16, 2026, Dhaka, Bangladesh
Kabir and Sayeed
Figure 3: Assembled prototype: normal operation showing fill percentage (left); “FULL” alert when bin reaches capacity (right). • EMPTY_DISTANCE (𝐷𝑒𝑚𝑝𝑡 𝑦 ): the distance reading (cm) when the bin is completely empty. Set to match the bin’s internal depth. These two values can be changed in the firmware configuration and uploaded once per bin during installation, requiring no hardware modifications. The same sensor unit can therefore serve a shallow 30 cm public litter bin or a deep 120 cm industrial waste container by simply updating these two values. The fill percentage 𝐹 is computed by the following geometryindependent linear normalization formula: 𝐹 = constrain
𝐷𝑒𝑚𝑝𝑡 𝑦 − 𝑑 × 100, 0, 100 𝐷𝑒𝑚𝑝𝑡 𝑦 − 𝐷 𝑓 𝑢𝑙𝑙
(1)
where constrain(𝑥, 0, 100) clamps the result to [0, 100] to handle edge cases (e.g., sensor noise beyond calibrated bounds). Worked example: For a bin calibrated with 𝐷 𝑓 𝑢𝑙𝑙 = 4.0 cm and 𝐷𝑒𝑚𝑝𝑡 𝑦 = 73.0 cm, a reading of 𝑑 = 16.4 cm gives:
Algorithm 1: Distance Measurement and Fill-Level Computation (ESP8266 Firmware). Only 𝐷 𝑓 𝑢𝑙𝑙 and 𝐷𝑒𝑚𝑝𝑡 𝑦 differ between bin deployments; both functions are identical for every bin geometry. Input: Calibration constants 𝐷 𝑓 𝑢𝑙𝑙 , 𝐷𝑒𝑚𝑝𝑡 𝑦 (set once per bin at installation) Output: Distance reading 𝑑 (cm); Fill percentage 𝐹 (%) 1 Function getDistanceCM(): 2 Trigger ultrasonic pulse on TRIG_PIN 3 duration ← echo pulse width on ECHO_PIN (timeout 30,000 𝜇s) 4 if duration = 0 then 5 return −1 // invalid reading 6 end 7 return (duration × 0.0343)/2 // cm, using speed of sound 8 Function getFillPercent(𝑑): 9 if 𝑑 ≤ 𝐷 𝑓 𝑢𝑙𝑙 then 10 return 100 11 end 12 if 𝑑 ≥ 𝐷𝑒𝑚𝑝𝑡 𝑦 then 13 return 0 14 end 𝐷𝑒𝑚𝑝𝑡 𝑦 − 𝑑 × 100 15 𝐹← 𝐷𝑒𝑚𝑝𝑡 𝑦 − 𝐷 𝑓 𝑢𝑙𝑙 16 return constrain(𝐹, 0, 100)
4.5
Data Transmission
The ESP8266 posts sensor data every 5 seconds. The transmitted payload consists of four fields: measured distance, computed fill percentage, bin identifier (bin_id), and device MAC address (device_mac), encoded as a URL-form HTTP POST request to the server’s This matches the live reading of 81% observed during hardware /data endpoint. testing (within rounding tolerance), confirming the formula’s vaAlgorithm 2 shows the corresponding firmware transmission lidity. loop. The sensor reading and fill computation are invoked every Boundary behavior: When 𝑑 ≤ 𝐷 𝑓 𝑢𝑙𝑙 , the formula returns 5 seconds; the result is encoded as a URL-form payload and dis100% (bin full). When 𝑑 ≥ 𝐷𝑒𝑚𝑝𝑡 𝑦 , it returns 0% (bin empty). The patched via HTTP POST. Invalid readings (𝑑 = −1) are discarded constrain call ensures stable output in both cases regardless of before transmission to prevent corrupt data reaching the server. sensor noise. The bin_id uniquely identifies the bin; device_mac allows the Firmware implementation. Algorithm 1 shows the two firmware server to match readings to registered hardware. If a transmission functions that implement the sensing pipeline on the ESP8266. Only fails, the firmware retries every 5 seconds for up to 5 minutes bethe two calibration constants, FULL_DISTANCE (𝐷 𝑓 𝑢𝑙𝑙 ) and EMPTY_DISTANCE fore discarding the reading. (𝐷𝑒𝑚𝑝𝑡 𝑦 ), differ between bin deployments; the getDistanceCM() 𝐹=
73.0 − 16.4 56.6 × 100 = × 100 ≈ 82% 73.0 − 4.0 69.0
(2)
and getFillPercent() functions are identical for every bin regardless of its physical geometry. getDistanceCM() triggers the HC-SR04 pulse sequence and converts the echo duration to centimeters using the speed of sound (0.0343 cm/𝜇s). getFillPercent() applies the linear normalization of Eq. 1 with constrain clamping, producing a stable 0–100% output for any calibrated range. The echo timeout of 30,000 𝜇s bounds the maximum measurable distance to approximately 515 cm, safely exceeding the calibrated 𝐷𝑒𝑚𝑝𝑡 𝑦 range used in this deployment. Changing only 𝐷 𝑓 𝑢𝑙𝑙 and 𝐷𝑒𝑚𝑝𝑡 𝑦 adapts the same firmware to any bin geometry, requiring no further code modifications.
5 Software Platform 5.1 Technology Stack The backend is built on ASP.NET Core MVC (.NET 9.0) [18] with Entity Framework Core for data access [19] and SQL Server for storage. The frontend uses Razor Views with Bootstrap for responsive UI and Leaflet.js [20] for interactive map visualization. REST API endpoints enable hardware communication and future mobile app integration.
CleanCity-BinSense
ICCA 2026, October 15–16, 2026, Dhaka, Bangladesh
Algorithm 2: Main Firmware Loop: Sensor Polling and HTTP POST Transmission Input: Wi-Fi connection status; calibration constants 𝐷 𝑓 𝑢𝑙𝑙 , 𝐷𝑒𝑚𝑝𝑡 𝑦 ; server URL Output: Periodic HTTP POST transmission of sensor readings 1 while true do 2 if Wi-Fi status = CONNECTED then 3 𝑑 ← getDistanceCM() 4 𝐹 ← getFillPercent(𝑑) 5 if 𝑑 > 0 then 6 payload ← {distance: 𝑑, fill: 𝐹 } 7 Send HTTP POST payload to server /data endpoint 8 end 9 end 10 Wait 5 seconds 11 end
that returns bins within a selected radius, along with their current fill level, status, last reading time, and distance from the selected point. Bins are color-coded by status to improve readability: green for normal bins, orange for bins that need pickup, and red for overflowing bins. Citizens can use this map to understand the current condition of bins in their area without accessing the authenticated dashboards.
5.2
Figure 4: Public bin map with color-coded fill status: green (Normal), orange (Needs Pickup), red (Overflowing/Urgent).
Per-Bin Configuration Management
When registering a new bin in the system, the Admin records the bin’s FULL_DISTANCE and EMPTY_DISTANCE values alongside its GPS coordinates and capacity. The server stores these per-bin parameters and uses them to validate incoming sensor readings and display fill-level history correctly for bins of different geometries. This design means that a city deploying CleanCity-BinSense across hundreds of bins of varying sizes does not require any custom firmware builds, only a one-time calibration entry at installation.
5.3
Role-Based Access Control
The web platform provides three differentiated user roles: • Admin: Responsible for route planning and system oversight. The Admin dashboard presents summary metrics, recent pickups, and route status information. Admin users can generate new collection routes by selecting a zone, date, truck, driver, and minimum fill threshold. The system then creates a route with ordered stops using the configured planning procedure. • Operator: Responsible for bin and reading management. Operator users can register new bins, view all bins with their latest fill levels, and submit new sensor readings. The operator dashboard highlights bins that need attention, including bins marked as needing pickup or overflowing. • Driver: Responsible for route execution in the field. Driver users view only the routes assigned to them, inspect route details on an interactive map, and record pickup completion for each stop. The interface is optimized for quick use during field operations and shows route progress clearly.
5.4
Public Bin Map
The platform includes a publicly accessible map that displays nearby bins with live status information. The map is centered around the user’s location when available, or around a default city location otherwise. Bins are retrieved through a dedicated API endpoint
5.5
Route Optimization Algorithm
When an Admin creates a collection route, the system applies a nearest-neighbor heuristic implemented in C# within the ASP.NET Core backend, supported by SQL Server’s spatial geography::STDistance() function for accurate geodesic distance measurement [14]. The algorithm is summarized in Algorithm 3. The nearest-neighbor heuristic is a classical approximation approach for route ordering problems related to the traveling salesman problem [15]. Algorithm 3: Nearest-Neighbor Route Optimization with Spatial Distance Input: Set 𝐵 of bins requiring collection (each with GPS coordinates); zone reference point 𝑧; minimum fill threshold 𝜏 Output: Ordered stop list 𝑅; estimated total route distance 𝐷 1 Filter 𝐵: retain only bins where fill ≥ 𝜏 2 current ← 𝑧 3 𝑅 ← [] 4 𝐷 ← 0 5 while 𝐵 ≠ ∅ do 6 𝑏 ∗ ← arg min STDistance(current, 𝑏) 𝑏 ∈𝐵
Append 𝑏 ∗ to 𝑅 8 𝐷 ← 𝐷 + STDistance(current, 𝑏 ∗ ) 9 current ← 𝑏 ∗ 10 𝐵 ← 𝐵 \ {𝑏 ∗ } 11 end 12 return 𝑅, 𝐷 7
Spatial distance measurement. Inter-bin distances are computed using SQL Server’s geography::STDistance() function, which
ICCA 2026, October 15–16, 2026, Dhaka, Bangladesh
calculates the geodesic (great-circle) distance in meters between two WGS-84 latitude/longitude points on an ellipsoidal Earth model. This is more accurate than planar Euclidean distance for urbanscale routes and requires no external mapping API. Each bin’s coordinates are stored as DECIMAL(9,6) latitude/longitude fields in the Bins table and queried as SQL Server geography objects at route-planning time. The distance from a reference point 𝑝 to a bin at coordinates (𝜙, 𝜆) is computed as:
Kabir and Sayeed
volume collected (litres), bin condition, and optional photo. The system tracks completion progress and generates a collection report (total waste, time, distance) upon route completion.
Listing 1: Geodesic distance computation using STDistance() d = geography :: Point ( p ) . STDistance ( 3 geography :: Point (𝜙 , 𝜆 , 4326) ) 1 2
Here, SRID 4326 denotes the standard WGS-84 coordinate reference system [17]. Time complexity is O (𝑛 2 ) where 𝑛 is the number of bins; for typical routes of 15–20 stops this completes in under 100 ms, enabling real-time responsive optimization. Figure 5 shows the route planning interface and Figure 6 shows the resulting optimized route map.
Figure 7: Driver route interface showing 7 numbered stops in optimized sequence with individual “Log Pickup” action buttons.
6 Experimental Results 6.1 Sensor Accuracy Across Calibrated Range To validate the configurable sensing approach, accuracy was tested at five representative distance points within an experimentally calibrated range (𝐷 𝑓 𝑢𝑙𝑙 = 4.0 cm, 𝐷𝑒𝑚𝑝𝑡 𝑦 = 73.0 cm for this test bin). The same formula (Eq. 1) applies without modification for any other calibrated range. Table 2: Sensor Accuracy at Representative Fill Levels
Figure 5: Route planning interface: admin configures target zone, date, assigned vehicle and driver, and minimum fill threshold before triggering automated route generation.
Figure 6: Optimized route map showing a 7-stop collection path with sequential numbered stops, route summary panel, and timestamped stop list.
5.6
Driver Navigation and Collection Logging
The driver interface (Figure 7) displays the optimized route as numbered stops on a Leaflet map. At each stop, the driver logs: waste
Actual (cm)
Measured (cm)
Error (cm)
Fill (%)
4.0 16.0 34.0 50.0 73.0
4.2 16.4 34.8 50.5 73.0
0.2 0.4 0.8 0.5 0.0
100% 81% 55% 33% 0%
Mean Absolute Error (MAE)
0.38
The MAE of 0.38 cm is within the rated hardware tolerance of the HC-SR04 ultrasonic sensor (±3 mm) [11], confirming that measurement error is bounded by sensor physics rather than software calibration. This translates to a fill percentage error of less than 1%, negligible for practical collection decisions. The constrain clamping ensures stable 100% and 0% readings at the calibrated boundaries regardless of minor sensor noise. The same measurement accuracy is preserved for any other calibrated range since the formula is geometry-independent.
6.2
System Latency
End-to-end latency was measured across 20 consecutive readings. Average latency from sensor measurement to web platform dashboard update was 5.3 seconds, dominated by the fixed 5-second firmware polling interval. Network transmission itself averaged under 300 ms on the site-local Wi-Fi infrastructure, confirming
CleanCity-BinSense
ICCA 2026, October 15–16, 2026, Dhaka, Bangladesh
that system response is limited by design polling frequency rather than network overhead.
6.3
Power Consumption
The ESP8266 unit with active Wi-Fi and sensor polling draws 120– 160 mA at 3.3V, giving an average power consumption of ≈0.5W. The integrated 5V/1W solar panel sustains continuous operation under normal daylight and charges a backup battery for nighttime or overcast operation.
6.4
rather than a planar approximation ensures that stop ordering remains accurate across all urban-scale deployments regardless of geographic location.
Mobile Web Interface
Figure 8 shows the local web interface hosted by the ESP8266 at 192.168.4.1, accessible from any device on the bin’s Wi-Fi access point. It displays fill level as a visual bar (81% in this reading) and the raw distance measurement (16.4 cm), allowing on-site inspection without any external tools.
6.6
Comparison with Existing Smart Waste Management Systems
Table 1 compares CleanCity-BinSense with representative systems. Compared with previous work, the proposed platform combines configurable per-bin sensing, embedded route optimization, rolebased management, and self-hosted deployment within a single architecture. Unlike cloud-dependent or GIS-assisted systems, it achieves 0.38 cm sensing MAE, 5.3 s end-to-end latency, and sub100 ms route generation on a complete hardware-software prototype using commercially available components, demonstrating its practicality for municipal deployment.
6.7
Network Reliability
During experimental evaluation, data transmission over the sitelocal Wi-Fi network remained stable, with successful delivery of sensor readings throughout repeated testing. The observed average end-to-end latency of 5.3 seconds was primarily determined by the fixed sensing interval rather than communication delays, while network transmission typically required less than 300 ms. These results indicate that, under reliable local network conditions, communication overhead has minimal impact on overall system responsiveness.
7 Figure 8: ESP8266-hosted local web interface showing realtime fill level (81%) and distance (16.4 cm) for on-site inspection.
6.5
Route Optimization Performance
Route optimization was evaluated by creating collection routes for bins distributed across a registered zone, with the nearest-neighbor algorithm ordering stops using STDistance()-based geodesic distances. Results are summarized in Table 3. Table 3: Route Optimization Performance Metrics Metric Algorithm execution time (avg.) Distance function used Coordinate reference system Admin-to-driver route delivery Bins filtered by fill threshold Stop ordering basis
Value <100 ms STDistance() (WGS-84) SRID 4326 <1.5 s ≥70% (configurable) Geodesic proximity
The sub-100 ms computation time enables responsive real-time optimization, and the operator can immediately reassign or regenerate routes if collection priorities change. Using STDistance()
Discussion
The experimental results demonstrate that CleanCity-BinSense enables practical real-time waste monitoring and collection optimization for resource-constrained environments. The configurable sensing approach achieved reliable measurements across the calibrated range, while the system maintained stable communication and generated optimized routes in sub-100 ms. Together, these results validate the proposed hardware, software, and optimization framework as a scalable solution for heterogeneous urban waste collection. Scalability: The most significant practical advantage of CleanCityBinSense is its configurable sensing range, which enables straightforward scaling across heterogeneous urban waste networks. In real deployments, bins of varying depths (e.g., 40 cm, 80 cm, and 120 cm) coexist on the same collection network. CleanCity-BinSense uses a single firmware with two per-bin calibration parameters set during installation. Combined with the modular architecture, each sensor node operates independently while communicating with a centralized server. This enables incremental expansion from small pilot deployments to city-wide implementations without requiring hardware or firmware modifications. Security: The current prototype operates within a trusted sitelocal network where communication occurs over HTTP between authenticated hardware and the server. User access to the management platform is protected through role-based authentication, restricting administrative, operational, and driver functions according to user privileges. For large-scale municipal deployment over
ICCA 2026, October 15–16, 2026, Dhaka, Bangladesh
public networks, additional security measures such as HTTPS encryption, device authentication, API tokens, secure credential management, and regular software updates should be incorporated to protect sensor data and prevent unauthorized access. Limitations: The nearest-neighbor heuristic provides practical routes but does not guarantee global optimality. For larger route sets (𝑛 > 50), algorithms such as Christofides or genetic algorithms may improve performance, while pre-computed distance matrices could further reduce planning latency. Large-scale deployment also requires one-time sensor calibration for different bin geometries and periodic maintenance to mitigate environmental effects. The current system relies on site-local Wi-Fi; deployments without existing network infrastructure would require 4G or NB-IoT connectivity. GPS coordinates are manually configured during installation, with autonomous location reporting left for future work. Long-Term Operation and Energy Considerations: The solarpowered ESP8266 prototype consumed approximately 0.5 W during normal operation, with the integrated charging system supporting continuous daytime operation and nighttime battery use. Future work will evaluate durability, battery longevity, sensor stability, and communication reliability through extended field deployments.
8
Conclusion
This paper presented CleanCity-BinSense, a complete IoT-enabled smart waste management system featuring a solar-powered sensor unit with a configurable sensing range for diverse bin geometries, a role-based web platform, and automated nearest-neighbor route optimization using SQL Server’s STDistance() function. Experimental results demonstrated a sensor MAE of 0.38 cm, within the ultrasonic sensor’s rated tolerance, 5.3 s end-to-end latency dominated by the sensing interval, and sub-100 ms route generation, demonstrating the system’s suitability for deployment in resourceconstrained urban environments such as Dhaka, Bangladesh. Future work includes (1) machine learning for predictive fill scheduling, (2) 4G/NB-IoT connectivity, (3) onboard GPS modules, (4) advanced route optimization (Christofides or genetic algorithms), (5) gas sensor-based waste classification, and (6) a 100-bin multizone pilot deployment.
References [1] Jose M. Gutierrez, Michael Jensen, Morten Henius, and Tahir Riaz. 2015. Smart waste collection system based on location intelligence. Procedia Computer Science 61 (2015), 120–127. doi:10.1016/j.procs.2015.09.170 [2] Sauro Longhi, Davide Marzioni, Emanuele Alidori, Gianluca Di Buo, Mario Prist, Massimo Grisostomi, and Matteo Pirro. 2012. Solid waste management architecture using wireless sensor network technology. In Proceedings of the 5th International Conference on New Technologies, Mobility and Security (NTMS). IEEE, 1–5. doi:10.1109/NTMS.2012.6208764 [3] Theodoros Anagnostopoulos, Arkady Zaslavsky, Kostas Kolomvatsos, Alexey Medvedev, Pouria Amirian, Jeremy Morley, and Stathes Hadjieftymiades. 2017. Challenges and opportunities of waste management in IoT-enabled smart cities: A survey. IEEE Transactions on Sustainable Computing 2, 3 (2017), 275–289. doi:10.1109/TSUSC.2017.2691049 [4] Fachmin Folianto, Yong Sheng Low, and Wai Leong Yeow. 2015. Smartbin: Smart waste management system. In Proceedings of the 2015 IEEE Tenth International Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP). IEEE, 1–2. doi:10.1109/ISSNIP.2015.7106974 [5] A. Chowdhury, M. R. Islam, and M. N. Islam. 2020. IoT based smart waste bin monitoring and municipal solid waste management system for developing countries. International Journal of Scientific and Engineering Research 11, 3 (2020), 1–7. Retrieved May 12, 2026 from https://www.ijser.org/researchpaper/IoT-based-
Kabir and Sayeed
Smart- Waste- Bin- Monitoring- and- Municipal- Solid- Waste- ManagementSystem-for-Developing-Countries.pdf [6] Houssem Jerbi, V. G. Anisha Gnana Vincy, Sondess Ben Aoun, Rabeh Abbassi, and Mourad Kchaou. 2025. Optimizing waste management in smart cities: An IoT-based approach using dynamic bald eagle search optimization algorithm (DBESO) and machine learning. Journal of Urban Management (2025). doi:10.1016/j.jum.2025.05.015 [7] Khalil Ahmed, Mithilesh Kumar Dubey, Ajay Kumar, and Sudha Dubey. 2024. Artificial intelligence and IoT driven system architecture for municipality waste management in smart cities: A review. Measurement: Sensors 36 (2024), Article 101395. doi:10.1016/j.measen.2024.101395 [8] Sangrez Khan, Bakhtiar Ali, Abeer A. K. Alharbi, Salihah Alotaibi, and Mohammed Alkhathami. 2024. Efficient IoT-assisted waste collection for urban smart cities. Sensors 24, 10 (2024), Article 3167. doi:10.3390/s24103167 [9] Abdullah Addas, Muhammad Nasir Khan, and Fawad Naseer. 2024. Waste management 2.0: Leveraging Internet of Things for an efficient and ecofriendly smart city solution. PLOS ONE 19, 7 (2024), Article e0307608. doi:10.1371/journal.pone.0307608 [10] Pallavi Sethi and Smruti R. Sarangi. 2017. Internet of Things: Architectures, protocols, and applications. Journal of Electrical and Computer Engineering 2017 (2017), Article 9324035. doi:10.1155/2017/9324035 [11] Elecfreaks. 2013. HC-SR04 Ultrasonic Ranging Module: Product Datasheet. Elecfreaks Technology. Retrieved May 12, 2026 from https://cdn.sparkfun.com/datas heets/Sensors/Proximity/HCSR04.pdf [12] Espressif Systems. 2023. ESP8266EX Datasheet, Version 7.1. Espressif Systems, Shanghai, China. Retrieved May 12, 2026 from https://www.espressif.com/site s/default/files/documentation/0a-esp8266ex_datasheet_en.pdf [13] Espressif Systems. 2023. ESP8266 Technical Reference, Version 1.7. Espressif Systems, Shanghai, China. Retrieved May 12, 2026 from https://www.espressif.co m/sites/default/files/documentation/esp8266-technical_reference_en.pdf [14] Microsoft. 2024. STDistance (geography Data Type) — SQL Server. Microsoft SQL Server T-SQL Spatial Reference Documentation. Microsoft Corporation, Redmond, WA, USA. Retrieved May 12, 2026 from https://learn.microsoft.com/en-us/sql/tsql/spatial-geography/stdistance-geography-data-type [15] Daniel J. Rosenkrantz, Richard E. Stearns, and Philip M. Lewis II. 1977. An analysis of several heuristics for the traveling salesman problem. SIAM Journal on Computing 6, 3 (1977), 563–581. doi:10.1137/0206041 [16] Md. Humaun Kabir, Sujit Roy, Md. Tofail Ahmed, and Mahmudul Alam. 2020. IoT based solar powered smart waste management system with real time monitoring: An advancement for smart city planning. Global Journal of Computer Science and Technology 20, H5 (2020), 1–10. Retrieved May 12, 2026 from https: //computerresearch.org/index.php/computer/article/view/1941 [17] National Imagery and Mapping Agency. 2000. Department of Defense World Geodetic System 1984: Its Definition and Relationships with Local Geodetic Systems. Technical Report NIMA TR 8350.2, Third Edition, Amendment 1. National Imagery and Mapping Agency, Bethesda, MD, USA. Retrieved May 12, 2026 from https://geodesy.unr.edu/hanspeterplag/library/geodesy/wgs84fin.pdf [18] Microsoft. 2024. ASP.NET Core Documentation. Microsoft Corporation, Redmond, WA, USA. Retrieved May 12, 2026 from https://learn.microsoft.com/enus/aspnet/core/ [19] Microsoft. 2024. Overview of Entity Framework Core. .NET Documentation. Microsoft Corporation, Redmond, WA, USA. Retrieved May 12, 2026 from https: //learn.microsoft.com/en-us/ef/core/ [20] Vladimir Agafonkin. 2010. Leaflet: An open-source JavaScript library for mobilefriendly interactive maps. Retrieved May 12, 2026 from https://leafletjs.com