1. Introduction
As the urban population is increasing, there are several challenges faced by large metropolitan cities. Some of these challenges include vehicle route guidance to avoid traffic jams [
1,
2], the effective utilization of the health care system [
3,
4], and efficient waste management [
5,
6]. The advancement in technology has provided several techniques and tools that can assist in solving these challenges, thus making way for smarter and cleaner cities [
7,
8].
Waste management is an important component of future smart cities [
9,
10]. The improper management of waste materials can be detrimental to the environment in many ways. This can significantly enhance land pollution and damage the soil, thus hurting human health and the ecosystem. Similarly, toxic materials can damage soil fertility leading to lower agricultural output. Waste mismanagement is also dangerous for marine animals and disrupts the supply of clean water to humans.
Waste materials are also a source of air pollution [
11]. In particular, through burning waste materials, the respiratory health of humans can be badly affected [
12]. Similarly, waste dum** can cause the release of methane-based greenhouse gas emissions [
13,
14]. This is one of the leading causes of climate change and global warming [
15,
16]. Other harmful effects of improper waste management include difficulty in extracting raw materials and damage to the animal habitat. Lastly, if the waste materials are not properly disposed of, it can lead to infectious diseases and other health issues.
Owing to the above issues, it is thus critical to design mechanisms for the efficient collection and disposal of waste materials [
17,
18,
19]. Sustainable, eco-friendly, and technology-assisted strategies are needed to develop waste dum** systems. With the increase in population as well as the amount of waste generated from households and industries, it is also pertinent to design methods for time-efficient and effective waste collection [
20,
21,
22,
23,
24].
The collection of waste relies on effectively using waste pickup trucks. Since the number of such waste pickup trucks is limited in number and the amount of waste is increasing day by day, managing the routes of these waste trucks and their pickup capacity are two critical issues. It is important to optimize the routes of waste pickup trucks so that cost-effectiveness and sustainability are achieved. For this purpose, technologies like the Internet of Things (IoT), Global Positioning System (GPS), and routing algorithms can be combined to formulate the best routes for waste pickup. The factors to be considered for waste vehicle routing include vehicle densities on the road and waste-related data. The goal is to select routes that reduce fuel consumption and maximize the disposal of waste.
The second issue is the management of the loading capacity of trucks that are used for picking up the waste. For better load management, the waste material type, waste priority, and truck capacity must be considered. Moreover, sensors can be installed, and regular waste fill-up data can be monitored to optimize the collection of waste. The efficient filling of waste pickup trucks is necessary to avoid extra trips and reduce fuel consumption. Moreover, it is vital to pick up waste of a critical nature and dump it in a time-efficient manner. Thus, the problem of waste pickup is essentially a resource management and allocation problem.
This paper focuses on the efficient allocation of waste material to waste pickup trucks. In this regard, a novel technique is proposed that uses the 0/1 knapsack algorithm to fill the waste pickup truck up to its loading capacity. The proposed algorithm takes into account factors such as waste volume, waste toxicity, and truck loading capacity, which are monitored using IoT sensors, and allocates waste to the trucks that maximize the waste utility while taking into account the truck’s loading capacity. The proposed technique was implemented in MATLAB software (version 2021) and compared against two other recent techniques from the literature. Simulation results show that the proposed technique improves the highly toxic waste collection by 47%.
The paper’s organization is as follows.
Section 2 provides a review of related techniques and mechanisms.
Section 3 describes the system model. The proposed technique is presented in
Section 4. The implementation of the proposed work and evaluation of results is provided in
Section 5. The conclusion is presented in
Section 6.
3. System Model
In this work, we considered multiple waste points located in different metropolitan locations, where different types of waste are collected. There are three waste bins in each of the waste locations carrying three different toxicity levels of waste material: high, medium, and normal toxicity levels. The bins will be filled by users based on their toxicity category; for example, if the waste is hazardous, it will be placed in the most toxic category bin. Similarly, general waste can be placed in the medium-toxic category bin. Lastly, the recyclable material can be placed in the least toxic category bin. It should be noted that classifying toxic category was not part of this study, and we rely on users to place the trash in the different bins. For classification purposes, IoT-based gas sensors can be used to measure the toxicity level of waste [
30] to measure the hazard level of the waste for human health [
31,
32].
The waste bins located throughout the metropolitan area were of the same size and differentiated with different colors. Bins in each location were equipped with IoT sensors to monitor the filled capacity of the bin and the time since the bin had been emptied or replaced. IoT-based ultrasonic sensors were placed to measure the volume of waste in each bin [
33]. All of the waste from different locations was collected by a dumper and dumped in a main garbage area. The dumper had a limited bin-carrying capacity and space to place a specific number of complete bins in it. It was supposed that the bins will be replaced with empty bins. All bins were supposed to be of the same size. A system model of the garbage collection is shown in
Figure 2.
Suppose that each of the waste collecting bins is categorized as
,
, or
for highly toxic waste, medium-toxic waste, and normal-toxic level waste material, respectively. The volume of each of the waste collection bins is
V. If waste collected in the
high-toxic bin is
,
medium-toxic bin is
, and
normal-toxic bin is
, and there are
N high-toxic bins placed at the different locations of the metropolitan area, then the total volume of the high-toxic waste (
), the total volume of the medium-toxic waste (
), and the total volume of the normal-toxic waste materials (
) are calculated as follows:
The total waste material available (
) from all the waste bins after a specific amount of time is the collective sum of all this waste:
The dumper replaces these bins after a certain time interval of the day. Suppose a dumper has a bin-carrying capacity for placing
C bins. If it picks up bins placed at the different locations of the areas up to their maximum capacity with
X high-toxic bins,
Y medium-toxic bins, and
Z normal-toxic bins, then the total amount of waste collected by a dumper (
) in its route is calculated as follows:
The toxicity of waste depends on the waste material as well as the duration since it was placed there. As more time lapses, the toxicity of the waste also increases. The toxicity values of a unit amount of the highly toxic, medium-toxic, and normal-toxic bins after a certain time are represented as
,
, and
, respectively, and the waste collected after a certain time
t is measured as
x amount of highly toxic,
y amount of medium-toxic, and
z amount of normal-toxic waste. The toxicity values of waste collected from high-toxic bins (
), medium-toxic bins (
), and normal-toxic bins (
) after time
t are calculated as follows:
The toxicity level of all the collected waste bins (
) from the different toxic level bins after a certain time
t is calculated as follows:
5. Results
In this section, we will present some of the results that were attained while simulating the waste collection based on our proposed mechanism. We labeled the high-, medium-, and normal-toxicity bins with numbers 10, 5, and 2, respectively. We chose the time since the last time the bin was collected from a set through a uniform distribution. Similarly, we chose the weight/volume from the set . For the simulations, we set the capacity of the vehicle to be equal to 10 bins. If the number of critical-waste bins was greater than the capacity of the vehicle, then another vehicle was sent. The number of critical-waste bins was the number of bins that had a priority value greater than the threshold value, which in our case, was equal to 100. The number of bins was set to be 40, i.e., 40 high-toxicity bins, 40 medium-toxicity bins, and 40 low-toxicity bins.
Figure 3 illustrates the total toxicity value of the waste collected, showing the impact of deploying varying numbers of waste bins. Our proposed method was compared with three strategies: first bin first (FBF), which prioritizes bin collection based on location; largest bin first (LBF), which plans routes focusing on the largest bins nearing their capacity, assigning them a higher collection priority; and longest delay (LD), which grants higher priority to bins with longer collection delays. The capacity of the vehicle for this figure was set to be equal to 10 bins. The number of bins was increased from 20 bins to 40 bins in a step size of 5. From
Figure 3, we can see that the scheme yielded the highest value of the overall toxicity for all values of the number of bins. Secondly, we can see that the value of toxicity increases from 1350 to 2700 for the proposed scheme when the number of bins was increased from 20 bins to 40 bins. The longest-delay scheme performs better as well. This is because the toxicity value depends on the time delay as well, and a larger delay value contributes to a higher toxicity value, as can shown in Equations (
6)–(
8). The FBF and LBF schemes performed the worst, as both location and volume do not have any involvement in the toxicity calculations.
Figure 4 shows the distribution of the waste collected based on high-, medium-, and normal-toxicity waste bins. The parameters for this figure are the same as in
Figure 4. In this figure, we can see that the proposed scheme performed best for the high-toxicity waste bins, while the medium-toxicity waste bins were also comparable to those of the LD scheme. The normal-toxicity waste bins were collected in small proportions due to their priority value being in the low region. The normal-toxicity level bins would be collected with a longer delay as the material that is present in the bin does not become toxic rapidly with time.
Figure 5 shows the results of the collected waste toxicity versus the truck capacity. The number of bins of each type was set to be equal to 40 bins. In this figure, we can see that the toxicity value of the collected waste is proportional to that of the LD scheme. The LD scheme is directly based on the parameter that defines the toxicity value of the waste, which is why it achieves a better performance in this case. It is important to mention here that as in the case of
Figure 3,
Figure 5 considers the overall toxicity value of all the waste bins combined. We can see that the FBF and the LBF schemes achieve a lower toxicity value of the collected waste as they do not directly consider the toxicity of the waste.
Figure 6 is the breakdown of the results from
Figure 5 into high-, medium-, and normal-toxicity bins collected, and its overall toxicity in the subset. We can see that although the LD scheme was performing better in
Figure 5, the amount of high-priority waste being collected by the proposed scheme is way better than in any of the schemes, including the LD scheme. For medium-toxicity bins, the proposed scheme achieves comparable results, and for normal-toxicity bins, the LD scheme achieves the highest value of the overall toxicity waste being collected.
From
Figure 7,
Figure 8,
Figure 9 and
Figure 10, we show the results for the amount of waste that was collected for various scenarios. For this set of figures, we kept the same values of the parameters as discussed in the discussion for
Figure 3,
Figure 4,
Figure 5 and
Figure 6.
Figure 7 shows the amount of waste collected with a varying numbers of bins, which increases from 20, in a step size of 5, to 40. In
Figure 7, we can see that the amount of waste collected increases with the increased number of bins. We can see that our proposed mechanism collects waste in an amount that is comparable to the largest-waste-bin-first scheme, while our proposed performs better than the FBF and the LD schemes. The downside of the increased value for the LBF scheme is that it achieves a lower value of toxicity, as can be seen from the trends shown in
Figure 3,
Figure 4,
Figure 5 and
Figure 6. The LD scheme was performing well for toxicity values, as is shown in
Figure 3,
Figure 4,
Figure 5 and
Figure 6, but in terms of the amount of waste collected, it performed poorly, as shown in
Figure 7. Meanwhile, the FBF scheme performed worst in all the cases because the waste collection was not conducted on any specific criteria, and so the scheme acts like random bin collection, as it will be collecting bins that come in its way.
A breakdown of the values in
Figure 7 in terms of the high-, medium-, and normal-toxicity bins is shown in
Figure 8, where we can see that for high-toxicity bins, the proposed method performs better than any of the other schemes. The performance in the medium-toxicity waste bin collection is also better than those for the LBF and LD schemes, while the LBF scheme achieved a slightly better value for the medium-toxicity waste bins and achieved the best value for the low/normal-toxicity waste bins. The results in
Figure 4 and
Figure 8 confirm that our proposed method achieves a higher priority for the high-toxicity waste bins while the medium- and the normal-toxicity waste will grow in priority if it is delayed by a large value or the bin is about to be filled.
Figure 9 shows the amount of waste collected versus the carrying capacity of the vehicle which was increased from 10 to 20 in a step size of 20. The numbers of bins for
Figure 9 and
Figure 10 were set to be equal to 40. From
Figure 9, we can see that the amount of waste collected increases with the increased capacity of the truck. We can see that the value increases from 50 to about 97 for the proposed scheme when the capacity of the truck is increased from 10 to 20. The LBF scheme performance is neck-in-neck with that of the scheme in terms of the amount of waste collected for the lower capacity of the truck and achieves slightly higher when the capacity of the truck is greater than 15. The FBF and the LD schemes perform worst as they do not rely on the amount of waste for calculating their bin collection priority. LBF collects bins based on their weight/volume so that is why it achieves a higher value for the amount of waste collected, but it performs poorly when we consider the toxicity measure of the collected waste, as can be seen in
Figure 5 and
Figure 6.
When we break down the results in
Figure 9 to show the amount of waste collected from the high-, medium-, and low/normal-toxicity waste, as shown in
Figure 10, we can see that the proposed method performs significantly better than the LBF scheme while it performs multi-fold higher than the FBF and the LD scheme for the reasons that were discussed in the explanation of
Figure 9. The proposed scheme performs better than the LBF scheme for the medium-toxicity waste bins as well and only performs poorly for the low/normal-toxicity waste bins as only a small set of the normal-toxicity waste bins made it into the high-priority zone due to a longer wait time or due to the amount of waste that might become full in a short time.
Cost Analysis of the Proposed Method with Legacy System
Our proposed method optimizes waste collection and reduces the number of required trips by deploying an IoT system. This section presents a cost analysis of the proposed system, demonstrating that over the long term, the initial investment in IoT deployment will prove beneficial compared to the legacy method. For this analysis, we assumed that the vehicles are already available, so the capital cost of the legacy system is not considered. We considered that there are N bin locations, each with three bins. The distance between two bin locations was set at 100 m; thus, if there are 100 bin locations, our vehicles would need to travel approximately 10 kilometers for each trip. For the legacy systems, we estimated a per-kilometer cost of USD 20, 30 and 40, encompassing fuel, maintenance, salaries, and other associated expenses in waste collection.
Regarding the proposed system, we evaluated the improvements in terms of the reduced number of trips required, as illustrated in
Figure 11. The analysis indicates a clear reduction of at least 50% in trips required for the optimized system compared to those of the legacy systems.
Turning to the additional costs for the proposed system, we considered that each bin is equipped with an ultrasonic sensor, priced at around USD 50. These sensors are WiFi-enabled and can utilize the WiFi network available at the end-user premises. Additionally, we anticipated a maintenance cost of approximately 1% of the total sensor costs per week. Furthermore, there will be an operational cost, estimated at 2% of the total cost of all the sensors per week. In assuming there will be five trips per week, the cost analysis is depicted in
Figure 12.
Referring back to the proposed algorithm, as indicated in
Figure 11, we observed that the number of trips will be halved compared to the legacy system. Consequently, the cost per week of operation will also be reduced by half. Although the proposed model entails higher initial costs due to sensor installation, the weekly operational expenses will decrease. It is evident that the accumulated costs will break even at approximately 30 weeks for the case where the operational cost is 20 USD per km, while it will break even at 20 and 17, when the cost per km is USD 30 and 40, respectively.