CheckBin: Intelligent Trash Bin Monitoring
17.05.2022
CheckBin is a project designed for the Internet of Things course to monitor the fill level of trash bins in real time through a network of intelligent devices. From the beginning, we followed a structured methodology aimed at addressing the problem through rigorous analysis, careful design choices, and systematic experimentation.
The project originates from an observed inefficiency in waste management: bins overflowing due to rigid pickup routes not tailored to real needs. Instead of approaching the problem with ad hoc solutions, we structured the work into distinct phases: identification of requirements, detailed system analysis, careful choice of technologies, prototype development, performance measurements, and scalability evaluation.
Starting with requirements, we defined clear and measurable targets:
- Maximum 10% error in the fill level estimation;
- Dashboard update within 2 hours of a change;
- Energy independence for at least a year;
- Less than 5% loss of bin capacity;
- Compliance with LoRaWAN duty cycle restrictions;
- Error rate of fill level measurements below 5%.
Only after setting these goals we moved to the system design phase. For the sensing part, we selected ultrasonic sensors to estimate occupied volume and a load cell to detect anomalies by cross-validating weight and volume information. Their integration was carefully planned to ensure reliability even in non-ideal operating conditions.
To guarantee robustness, we designed a logic where each sensing event includes multiple ultrasonic measurements (with timeouts to avoid errors due to bin opening) and cross-checks with the load cell reading. The actuator logic dynamically locks or unlocks the bin based on both volume and weight thresholds.
Cloud architecture was designed with the same rigor. We structured the data path from LoRaWAN gateways (through The Things Network) to AWS IoT Core, Lambda, DynamoDB, and Amplify, ensuring scalability and modularity. Every technology was chosen based on a comparative analysis of alternatives, considering scalability, latency, and cost.
In the experimental phase, we carefully measured the energy consumption of every system component using multimeters and INA219 sensors. Sampling and transmission strategies were optimized based on energy profiling:
- Sensing frequency set to once per hour to balance accuracy and energy savings;
- Data transmission triggered only on fill level change or anomaly detection.
This approach allowed us to remain compliant with LoRa duty cycle constraints while maximizing system autonomy.
Extensive measurements were conducted on the fill level estimation accuracy, anomaly detection effectiveness, and network usage efficiency. Testing showed a 92.5% accuracy in fill level detection, a value expected to improve with larger datasets. Scalability tests were conducted in a real-world setting, reserving IoT-LAB resources to simulate a dense deployment of bins. The system managed the load effectively, although dashboard update mechanisms were tuned to refresh every 10 minutes to avoid unnecessary cloud stress.
The structured methodology enabled us not only to develop a working prototype but also to anticipate and solve problems that could arise in a real city-scale deployment, such as energy supply limitations and cloud infrastructure load.
Future technological improvements were identified through the same structured approach. Solar panels were proposed as a viable solution to reduce battery size and weight. Other ideas, like harvesting mechanical energy from user interactions, were evaluated but discarded after quantitative analysis showed their limitations.
Through the checkBin project, we demonstrated how a structured and rigorous engineering process, supported by continuous analysis and validation at every stage, can lead to practical, scalable solutions even in complex real-world scenarios.
The complete code and documentation are available in the public repository.