You are on page 1of 5

An IoT solution for load monitoring and tracking of

garbage-truck fleets
1st Jorge Maia 2nd Jones Yudi
IoT Division Automation and Control Group
CrayzTechLabs University of Brasilia
Brasilia, Brazil Brasilia, Brazil
jm@crazytechlabs.com.br jonesyudi@unb.br

Abstract—One of the most significant problems in the modern Mondays, vacation periods, festive days like carnival have
world is the amount of waste generated by the growing urban different behaviors from citizens, and on the formula, we can
population. The destination of this garbage can be recycling, add weather, economic factors, and many other components.
burning or disposal in landfills. A family of four people can
produce, on average, 3 kilograms of garbage per day. The The income level of the neighborhood can also define quantity
collection of this waste is mostly done by trucks that go through and quality of garbage [2].
the streets of the cities collecting the garbage from door to The density of the garbage collected will vary depending
door, following predetermined routes. When a truck reaches its on its composition, and there are also physical and regulation
maximum load, it goes to the unloading site, returning to the limits for the maximum allowed truck load. If a truck reaches
route at the same exit point in order to finish that route. The
logistical problems involved are several: fines for overloading its load limit it must travel to the disposal place and back to
the truck, late collection times, labor lawsuits, among others. the scheduled route to finish the collection on time. Due to
This project proposes a solution based on Internet Things to the unpredictability of the waste composition, there is a need
monitor the location of the truck and its cargo generating maps for tracking the truck load during the collection to allow for
in near real-time, to allow the dynamic optimization of the routes optimized operation.
followed by the various paths of the fleet. Our system consists of
an on-board system that measures the truckload and determines Telemetry on trucks and service cars is a commodity in-
its position, sending this data to a cloud computing solution. stalled from the factory as an optional. On the waste collection
Several tests were carried out in the field, with pathways from service in developing countries, sometimes we are talking
a real fleet, operating on an ordinary working day. The results about old trucks, modified to apply the compression body,
show that the system developed is viable, successfully meeting tilt systems, and other components needed for this heavy-duty
the requirements of the application.
Index Terms—Internet of Things, Azure, IoT Hub, Vehicular work. The available fleets in such places are generally not
Telemetry, Edge Computing, Cloud Computing equipped with telemetry or tracking systems.
In this work, the environment of study is Recife, a coastal
I. I NTRODUCTION city located at the northeast of Brazil. We had access to a real
solid waste collector fleet, where the challenge is to increase
Every city is a huge solid waste generator in potential; the operational performance using Internet of Things concepts.
challenge of waste management is a real issue everywhere Figure 1 shows a waste truck collector with the weight sensors
in the world, and different strategies are applied overall the used and the IoT Device developed in this project. In the
world, depending on many factors (culture, local regulation, next sections, we describe the main issues to be solved, the
economics, and so on). Data from the World Bank shows drawbacks, the proposed solution, the design decisions we
that in 2050 more than 3 billion tons of solid waste will have made, as well as the results obtained.
be generated worldwide each year, and the waste industry is
estimated at US$ 433 billion annually. Almost 70% of them II. P ROBLEM S TATEMENT
have the final destination to sanitary landfills and dumpsites This project aims to improve in the efficiency in collecting
[4]. route, decreasing costs with fuel, maintenance, fines, and
The most common waste collection service is based on people’s salaries, as well as avoiding accidents caused by
curbside collections, where a vehicle travels along the city vehicle’s overload [6].
collecting the waste from household, commercial, and indus- In Recife and some federal roads in Brazil, the law limits
trial activities [3]. The routes are predefined for each region of the maximum weight on the vehicle, in case of overweight
the city, and scheduled within time-boxed operational hours. fines are applied. Every truck has the baseline weight on the
In some cases, more than one truck is working on the area to vehicle documentation and also the maximum duty charge.
accomplish on-time service and finish the collection schedule. Every journey begins on the garage with a driver and three
Another issue of this public service is that the garbage is dy- operators, fueling the truck, forwarding to collect and finish on
namically generated and sometimes non-predicted. Holidays, sanitary landfill, where a scale weights the vehicle, heading to

978-1-7281-6389-5/20/$31.00 ©2020 IEEE 406

Authorized licensed use limited to: University of Gothenburg. Downloaded on December 21,2020 at 16:53:30 UTC from IEEE Xplore. Restrictions apply.
based on distance sensors, measuring the relative displacement
of the chassis with the wheel axles, for example, capacitive
[9] and ultrasonics [10].
Another important factor is the acquisition of load data with
the vehicle in motion (dynamic load) or stopped (static load)
[6]. To facilitate the determination of the amount of garbage
on the road, in this project the measurement of the load is
done when the vehicle is stationary.
Another critical aspect of the solution is the near-realtime
need for information and the confidence of it. The weight
should be indexed by a position and a timestamp, producing
a map in near-realtime.
Internet of Things is a term coined in 1999 by Kevin
Fig. 1. Modifications on garbage truck. Ashton [5] to determine things connected to the Internet,
nowadays it is used to almost all devices on the field, with
or without intelligence. Communication technologies emerged
discharges the solid waste, then crossing into another scale at
to accomplish the new goals of this growing market of devices,
the end of this process. Some fines are generated at this step
and on the software and cloud side, big players like Microsoft,
if the weight exceeds the maximum allowed. Transcending
Amazon, Google, and IBM have their public platform to en-
the fines, overweight generate unpredictable maintenance costs
sure that security and costs of development are accomplished,
and an unsafe motorist and operator’s journey.
producing a cost-effective time-to-market. [8]
The journey of one travel still have some other several items
Against an old telemetric solution that only sends messages
that can measure the productivity, but the main goal is to
by GSM or SMS, the advantage of an IoT device is that it
collect all the garbage in the time window defined for the
is monitorable, intelligent, more secure, and actions executed
region.
on the field controlled by a robust platform on the cloud.
Every day, some trucks arrive in the sanitary landfill with
There are some communication protocols and technologies
different weights and loads, from small load to overloaded
for metropolitan areas such as LoRaWan, Sigfox, Narrowband
based on visually understanding of operators. Density, type
IoT, CAT-M, specifically to IoT communication, based on low
of material collected, compression ratio are not accurately
power consumption and long-range communication; those are
measured, sometimes generating the first travel overloaded and
more valuable to new solutions instead of the traditional 2G
the second one or the last executed with a small load.
or 3G. Each one has properties that fit a specific application
TABLE I
need. [7]
M EASUREMENTS OF REGION A, FOR SAME TRUCK
III. T HE P ROOF OF C ONCEPT
Date Time Start Duracao Baseline Garbage (Kg) The solution to acquire the weight of the truck should start
10/10/2019 8:45 am 2h48min 11.290 11.543 on a small scale to prove the concepts and then start growing
10/10/2019 12:20pm 1h10min 11.190 790 with a rollout to all fleet in a maximum time of 2 years.
11/10/2019 8:45am 3h24min 11.245 10.320 The technology should attend requisites of security, scalability,
11/10/2019 12:10pm 2h45min 11.273 7.878 costs, reliability, automated deploy, and maintenance capabil-
ities.
Table 1 shows the discrepancies about the weights in travels The information should be at rest, and on travel, devices
to finish the area. The first travel is overweight, and the second communication should be monitored to prevent data lacks.
one has less than a quarter of capacity, presenting the problem The detection of anomalous data should alarm, and immediate
about the weight not perceived during the way to the landfill. actions should be taken, dashboards, and integrate with legacy
systems is also needed.
A. Related Work
There are several ways to determine the actual load of a A. Why choose a cloud provider instead keep it on-premises?
vehicle. The most common system is the one used by traffic The use of a public provider gives some benefits instead
authorities on fixed road sections: a balance that measures the of a private cloud. Different services in models like IaaS,
total weight of the loaded truck and, by simple subtraction Infrastructure as a Service, PaaS, Platform as a Service, and
from the known weight of the empty truck, determines the SaaS, Software as a Service, can increase the possibility
load. This system, however, is not feasible for our application, of grow the utilization of them without upfront costs. The
since it is not possible to distribute the balancers throughout main difference when contracting a cloud provider instead
a city [1]. of maintaining internal infrastructure, on-premises environ-
Vehicle-mounted solutions are the most suitable to allow ment, is the commitment to physical infrastructure, upfront
refined monitoring of the road load. Some of these systems are investments, and benchmark of the services. IoT solutions and

407

Authorized licensed use limited to: University of Gothenburg. Downloaded on December 21,2020 at 16:53:30 UTC from IEEE Xplore. Restrictions apply.
Fig. 2. Overview of full system architecture.

scale concepts run together, some times elasticity is needed All raw data is saved on one Storage Account service for
for the use of services; without a cloud provider, the costs further use if needed, on this PoC this data is not used, but
should come upfront, and the time to upgrade or grow the we decide to maintain it as it came from devices to future
infrastructure can create a lack of solution’s needs. implementations.
C. Edge Details
B. Azure IoT Platform On the edge side, every truck has a device installed on
Based on our previous experience and extensive documenta- the front, composed by an ESP32 Microcontroller, connected
tion available, for this project, we decided to use the Microsoft to communications modules to GSM and GPS, also have
Azure IoT Suite as our public cloud services provider. a connection to a CAN transceiver, an RS485 module and
The present solution is based on Azure IoT. All devices an SDCard module where all raw data is saved for backup
managed by Azure IoT Hub, the message broker, and pro- purposes. In some versions, we also installed a TFT display,
visioned securely using the device provisioning service. The but it is not mandatory, as an RGB LED provides by color
messages come from the devices to the IoT Hub then routed to and a blink code the working status and module fail as well.
the Time Series Insights services or alarms based on properties An overview of the idea is presented in Fig.3.
generated based on a small intelligence embedded on the
device code. Figure 2 shows an overview of our architecture.
The Time Series Insights (TSI) generates heatmaps of
devices communications and promote all queries using an API
to the data ingested. This service also has a Javascript SDK
for interactive charts, facilitating the development of end-user
interfaces and dashboards. The heatmap of communication
was used on the PoC to track the truck’s communication in
realtime.
Fig. 3. Edge Components - Hardware
All Alarm messages received are routed from IoT Hub to an
instance of Service Bus Queue to connect and trigger an action On ESP32, we have a small amount of internal memory
on a service called Logic Apps. The Logic Apps is a service and two cores of processors, besides a couple of I/O pins
that implements a visual workflow model of activities, based and connectivity ports. Embedded code programmed using
on a trigger and binding to some other services. This solution Espressif IoT Development Frameworks (ESP-IDF), where we
is using this service to create a comfortable and enriched added the IoT Hub C SDK to manage the device properties,
experience on alarms actions, with Logic Apps, the final user secure communication, and payload send to the Azure IoT
(non-technical) can define the workflow and change it at any Hub.One of the critical points about data generated from
time without effort. sensors is the concern about the position of measure within a
The frontend is provided by a web app, behind some rules timeframe. To promote this, we have to get the actual position
of a firewall, to make the application only accessible by users of the device and combine a timestamp. Figure 4, presents the
involved in the proof of concept. The Web App is running on sequence of all actions taken from the device. The two cores
an App Service with small hardware. of ESP32 are used in parallel to provide a continuous sensor

408

Authorized licensed use limited to: University of Gothenburg. Downloaded on December 21,2020 at 16:53:30 UTC from IEEE Xplore. Restrictions apply.
acquisition in one core, as some of the sensors are active on the data on internal memory when received from each connection
generation data without a range of defined time. The second to the device from the modules or sensors. This process is a
core is responsible for positioning and message sending. guarantee in case of a reboot by software, hardware, or power
failure. Any kind of error or reboot generates a dump of data
to be sent in the following operational status to the cloud. All
data acquired is reduced and take the form of a message like in
Listing 1, using the serialization in JSON (JavaScript Object
Notation) format.
Listing 1
B ODY OF THE MAIN MESSAGE IN OUR SOLUTION

{
” Modulo ” : 0 ,
” Timestamp ” : ” 2 0 1 9 / 1 0 / 1 1 T20 : 4 0 : 2 3 Z” ,
” IdCaminhao ” : ” AKF3377799 ” ,
” L a t i t u d e ” : −8.074555 ,
” Longitude ” : −34.916386 ,
” P o s i t i o n i n g ” : ”GPS | AGPS”
” Weight ” :
{
” HPressure ” : 2344.93 ,
” Weight ” : 1 3 4 5 3 . 3 3 3 4 5 ,
” WeightAcc ” : ” up ” ,
” t i m e s t a m p ” : ” 2 0 1 9 / 1 0 / 1 1 T20 : 4 0 : 2 3 Z” ,
” O r i g i n ” : ” EsqB ” ,
” L a t i t u d e ” : −8.074555 ,
” Longitude ” : −34.916386 ,
” P o s i t i o n i n g ” : ”GPS | AGPS”
}
}

IV. R ESULTS

On this Proof of Concept, we accomplished the challenge


of acquiring messages from multiple sources and formats on
the truck. We had Serial communication, CAN Bus queries,
and multiple forms of positioning data combined to a single
result.
As we propose at the beginning of this project, the goal was
Fig. 4. Sequence Diagram of Edge Side - Device
to provide the measurement on edge available on a platform
The module of GPS, Global Positioning System uses geo- to be integrated with other legacy systems and monitored.
stationary satellites to provide the position of a module, is We consider it the first stage of an IoT project, where
used for positioning acquisition, it takes some time to get the data is sent to the cloud to generate insights about the
an accurate position, and in some places, it does not work, operational data in near-realtime, in contrast with previous
gray areas, or commonly named shadow areas. As one of the operation where data was only accessible in the time frame of
concerns is about the positioning combined on the weight, we the next day.
also use the Assisted GPS, AGPS, that uses cellular towers to The data acquired from the truck with latitude, longitude
calculate the position of the module, It is more accurate when and weight can produce some initial value to the operation
we have more towers near and works analog as the GPS; the presenting the routes with their workload giving the ability to
AGPS can be combined with GPS to make the measure more change trucks during the day and optimizing the load of trucks
precise. as presented on Table II and on 5.
Messages are tagged with a property that describes how The use of the Azure IoT Platform was very successful
the position was acquired. If the payload of the sensors because it generates a gain in development time and opens
had any discrepancies or errors, it could generate an alarm the opportunity to integrate into any system on Azure Cloud
message or just another property to do not use it as a precise or not.
measurement. A non-commercial solution measures the truck The alarms programmed to work on the PoC were not used
load weight, providing a RS485 data stream to our platform because no clear rules exist to alarm any problem on the
within an acquisition rate of 5 minutes. Due to non-disclosure operation, just problems on the device are sent through an
agreement, more detailed information about the weight mea- SMS to the truck driver, helping us to maintain the system
surement system cannot be given in this work. A process saves working by a forced manual reboot triggered by the driver.

409

Authorized licensed use limited to: University of Gothenburg. Downloaded on December 21,2020 at 16:53:30 UTC from IEEE Xplore. Restrictions apply.
TABLE II on Mixed Design of Integrated Circuits and Systems, MIXDES 2014,
TABLE WITH WEIGHT BASED ON L ATITUDE AND L ONGITUDE FOR A pages 477–481, 2014.
SMALL ROUTE . [2] N. J. Bandara, J. P. A. Hettiaratchi, S. C. Wirasinghe, and S. Pilapiiya.
Relation of waste generation and composition to socio-economic factors:
Time Stamp Lat. Long. Weight A case study. Environmental Monitoring and Assessment, 135(1-3):31–
39, 2007.
2019/10/14T22:11:44Z -8.127057 -34.960377 11320.0 [3] S. S. Christian Ludwig, Stefanie Hellweg. Municipal Solid Waste
2019/10/14T22:30:51Z -8.123381 -34.944229 14519.6 Management: Strategies and Technologies for Sustainable Solutions.
Springer-Verlag Berlin Heidelberg, first edition, 2003.
2019/10/14T22:39:50Z -8.093868 -34.935516 16119.4
[4] J. Gutberlet. Cooperative urban mining in Brazil: Collective practices in
2019/10/14T22:58:40Z -8.076668 -34.920277 17280.1 selective household waste collection and recycling. Waste Management,
2019/10/14T23:16:08Z -8.076668 -34.920277 18620.7 45:22–31, 2015.
2019/10/15T00:09:42Z -8.112012 -34.893047 18620.7 [5] K. Ashton. That ’ Internet of Things ’ Thing. RFiD Journal, 22:97–114,
2019/10/15T01:54:03Z -8.139658 -34.949577 19852.4 2009.
[6] Y. Liu and Z. Liu. An Optimized Method for Dynamic Measurement
of Truck Loading Capacity. 2018 3rd IEEE International Conference
on Intelligent Transportation Engineering, ICITE 2018, pages 120–124,
2018.
[7] K. Mekki, E. Bajic, F. Chaxel, and F. Meyer. A comparative study
of LPWAN technologies for large-scale IoT deployment. ICT Express,
5(1):1–7, 2019.
[8] T. Pflanzner and A. Kertesz. A survey of IoT cloud providers. 2016
39th International Convention on Information and Communication Tech-
nology, Electronics and Microelectronics, MIPRO 2016 - Proceedings,
pages 730–735, 2016.
[9] S. Yang, W. Yang, and Y. Yang. Measurement of vehicle-load using ca-
pacitance and acceleration transducers. Journal of Physics: Conference
Series, 76(1):1–6, 2007.
[10] Y. D. Zhao and X. L. Liu. The study of vehicle load monitoring system.
Applied Mechanics and Materials, 505-506:384–387, 2014.

Fig. 5. Results and Weights from one small route

V. F UTURE W ORK
As we concluded the Proof of Concept with the measure-
ments on the cloud and presented the possibility of integration
with other systems and data processing pipelines, the next step
is to create new hardware and an unassisted setup process and
operation. The difficulty of installation, problems of operation,
battery charges, and others are mapped and will be improved
in the next phase.
The mapped activities on the hardware side for the pilot
phase with more trucks are the improvement of the device’s
hardware, charge circuit, and conception of any kind of a self-
healing system. At the software side, including cloud and edge,
the first action is the design of a dashboard and an API to
integrate with the ERP (Enterprise Resources Planning).

R EFERENCES
[1] A. Andrzejczak, P. Lecczycki, P. Pietrzak, B. Peckoslawski, and
A. Napieralski. Module hardware structure of wireless vehicle weight
measurement system. Proceedings of the 21st International Conference

410

Authorized licensed use limited to: University of Gothenburg. Downloaded on December 21,2020 at 16:53:30 UTC from IEEE Xplore. Restrictions apply.

You might also like