ECO4RUPA: 5G-IoT Inclusive and Intelligent Routing Ecosystem with Low-Cost Air Quality Monitoring

Round 1

Reviewer 1 Report

The authors proposed an inclusive and intelligent routing ecosystem with the objective of calculating healthy routes according to the profile and particular needs of each citizen. It is very well explained and performance evaluation is well done.

However I have some comments regarding data collection methodology:

-          WSN, i.e Wireless Sensor Networks assume mesh communication topology (ZigBee for example), and cooperation between nodes of the network. Is it your network architecture like this, or your nodes directly communicate with the server without multi-hop communications? In that case this is not a Wireless Sensor Networks, but rather IoT Sensor network.

-          What is the measurements resolution in seconds/minutes?

-          Why do you use 5G for this measurements? It is rather small and sporadic amount of data and for that you don't need 5G, but rather LTE-M, NB-IoT, LoRa etc.  Except if in observed  5G network is implemented massive MTC 5G?

-          So you have data from public monitoring stations and from your devices. How many devices is deployed?

-          If devices are used by cyclist how are reliable collected data? I guess that device has a GPS module so you can have a position were data are collected, but if the cyclist are driving for example 20-30 km/h and measurements are conducted every 1 minute then you will have a quite large distance between two measurements.

-          Recommendation for mounting of AQ low-cost devices is on the height 1.5-4m above the ground. Where doy ou mount devices on the bicycle, and how do you consider reliability of measurements?

-          Please add one paragrapfh which explains shortcomings and limitations of the proposed ecosystem,  in the sense of needed number of devices, costs, reliability, calibration , maintenance etc...

Minor editing of English language required.

Author Response

Dear editor,

Thanks for giving us the chance to address the comments provided by the anonymous reviewers. This is a reviewed version of a previously submitted manuscript (information-2522909) to be considered for publication in Information. We have done our best to address all the comments raised by the reviewers properly. We attach a response letter indicating how all their comments have been addressed. The submitted manuscript uses red to indicate the changes applied.

We hope this new version satisfies the high-quality standards of your journal.

Best regards,

Jaume Segura-Garcia

Reviewer 1:

Comment:

The authors proposed an inclusive and intelligent routing ecosystem with the objective of calculating healthy routes according to the profile and particular needs of each citizen. It is very well explained and performance evaluation is well done.

However I have some comments regarding data collection methodology:

-      WSN, i.e Wireless Sensor Networks assume mesh communication topology (ZigBee for example), and cooperation between nodes of the network. Is it your network architecture like this, or your nodes directly communicate with the server without multi-hop communications? In that case this is not a Wireless Sensor Networks, but rather IoT Sensor network.

Answer:

Thank you for your comment. In our architecture for our application, we combine different Air Quality (AQ) monitoring nodes, both official (static) stations and the proposed low-cost nodes, called Eco4rupa AQ nodes. As the reviewer says, the latter, the Eco4rupa nodes, connect directly to the server. We do not use multihop communications. Nevertheless, if it were necessary, we could consider this option, but in this scenario, it is out of the scope. 

The goal of these nodes is to improve the coverage in poor AQ official monitoring areas and they have been designed to connect through different types of wireless communications, as shown in Figure 6 of the paper, but in a direct way. The technology used is given by the available services in the area.

We have clarified this issue of multihop networks in the new version of this paper in order to avoid this confusion.

Comment:

-          What is the measurements resolution in seconds/minutes?

Answer:

The minimum and recommended time frequency used for AQ monitoring is 10 minutes, according to the ISO 11771:2010, ISO 37122:2019 and European Regulation Directive 2008/50/EC. It is worth mentioning that the monitoring of meteorological conditions also follows a similar sampling process, based on ten-minute periods.

In practice, this time interval is enough to provide accurate information for the route planner and this time scale is sufficient for this purpose.

We have introduced this detail in the new version of the manuscript.

Comment:

-          Why do you use 5G for this measurements? It is rather small and sporadic amount of data and for that you don't need 5G, but rather LTE-M, NB-IoT, LoRa etc.  Except if in observed  5G network is implemented massive MTC 5G?

Answer:

Since our goal is to cover different scenarios in urbanized areas, the Eco4rupa low cost AQ monitoring nodes have been designed to be flexible and can integrate all these technologies as shown in Figure 4 of the paper. In particular, we use one technology or another based on the available wireless services at the sampling point in the deployment. In fact, currently we are using NB-IoT for the communication, which is one of the 5G-enabled technologies, to send the data from these nodes to the server every 10 minutes as mentioned above.

We have improved the wording of the paper to clarify this issue in the new version of the manuscript.

Comment:

          So you have data from public monitoring stations and from your devices. How many devices is deployed?

Answer:

Yes, for the healthy route planner application, we obtain pollution measurements from nearby AQ monitoring nodes. The number of monitoring nodes depends on the area under test. 

According to Directive 2008/50/EC of 21 May 2008 on ambient air quality and cleaner air for Europe, the number of sampling points in each zone or agglomeration should be at least one sampling point per 2 million inhabitants or one sampling point per 50 000 km2 , where the latter criterion results in a higher number of sampling points, but not less than one sampling point per zone or agglomeration. 

Figure 1.- Detail of the AQ monitoring stations used in the area of Burjassot (Valencia)

In particular, in the area under test where we have carried out our experiments shown in Section 5 of the paper, Burjassot city, we have 6 official AQ monitoring stations nearby, within a distance lower than 5 Km. Figure 1 (included in the revision as Figure 9 in the paper) depicts the location of these 6 stations. Notice that this is a square area of 10 km x 10 km,  with 1.3 inhabitants. In this case, it was not necessary to install any extra AQ monitoring node. The Eco4rupa AQ monitoring nodes are used in areas with low coverage of these official AQ stations.

We have included these details in the new version of the manuscript.

Comment:

-          If devices are used by cyclist how are reliable collected data? I guess that device has a GPS module so you can have a position were data are collected, but if the cyclist are driving for example 20-30 km/h and measurements are conducted every 1 minute then you will have a quite large distance between two measurements.

Answer:

Thank you for your comment. We realized that our wording was a bit misleading. 

Actually, we do not use any mobile AQ nodes, all of them are static. As mentioned before, the Eco4rupa AQ nodes can be placed wherever they are needed, they are easy to install, but they are not mobile (or installed in any vehicle) but static. The aim of these nodes is to enhance the coverage provided by the official AQ monitoring stations, which are permanently placed. Nevertheless, as shown in the related work, we can find different initiatives based on mobile AQ monitoring nodes, but this approach is different from ours.

We have enhanced the wording in the paper to clarify this issue and avoid this confusion.

Comment:

-          Recommendation for mounting of AQ low-cost devices is on the height 1.5-4m above the ground. Where doy ou mount devices on the bicycle, and how do you consider reliability of measurements?

Answer:

As mentioned above, we do not use mobile AQ monitoring nodes. We have improved the wording accordingly to avoid this confusion.

Comment:

-          Please add one paragrapfh which explains shortcomings and limitations of the proposed ecosystem,  in the sense of needed number of devices, costs, reliability, calibration , maintenance etc…

Answer:

Thank you for your suggestion. In this paper, we describe the different activities carried out within the Eco4rupa project, and in particular we focus on the performance and functionality of the healthy route planner application. The information used to feed this application, comes from both official static AQ stations (mainly) and the low-cost Eco4roup AQ nodes when it is necessary. These nodes have been designed for scenarios with poor coverage by AQ official monitoring areas and as it is mentioned in the manuscript in Section 2 and 3, their performance and accuracy is obviously lower compared to the official stations. 

On one hand, these low-cost nodes are more flexible in terms of deployment and lower cost, but on the other hand, they are less accurate (as it is detailed in Section 2 of the paper) and their raw measurements must be further processed (including calibration) in order to provide accurate AQ monitoring data. This process is out of the scope of this paper. Besides, the life time of the sensors equipped on these low cost nodes is shorter (between 6 and 12 months, and due to their cost we replace them when needed) compared with the official stations; however, the official stations require a weekly maintenance.

We have included the following paragraph in the conclusion section to highlight these pros and cons:

“Notice that this healthy route planner application uses data from both official static AQ stations (mainly) and the low-cost Eco4rupa AQ nodes when it is necessary. These Eco4rupa nodes are simpler and easier to be deployed, but on the other hand they are less accurate and their raw measurements must be further processed (including calibration) in order to provide accurate AQ monitoring data. Besides, the life time of these sensors equipped on these low cost nodes is shorter, between 6 and 12 months. However due to their cost, we can replace them when needed, compared with the official stations that require weekly maintenance.”

Author Response File: Author Response.pdf

Reviewer 2 Report

1. The system model is not clear. What is the specific requirement for the use case?

2. The routing planner is too simple. How is the routing actually designed?

3. The validation part is not explicit. The KPI for for data collecting or routing is not provided.

4. The presentation should be thoroughly improved.

5. More similar existing works can be compared as follows:

[1] Maximizing spatial–temporal coverage in mobile crowd-sensing based on public transports with predictable trajectory. International Journal of Distributed Sensor Networks. 2018;14(8). 

[2]  "Collaborative Learning of Communication Routes in Edge-Enabled Multi-Access Vehicular Environment," in IEEE Transactions on Cognitive Communications and Networking

Carefully revision is needed.

Author Response

Dear editor,

Thanks for giving us the chance to address the comments provided by the anonymous reviewers. This is a reviewed version of a previously submitted manuscript (information-2522909) to be considered for publication in Information. We have done our best to address all the comments raised by the reviewers properly. We attach a response letter indicating how all their comments have been addressed. The submitted manuscript uses red to indicate the changes applied.

We hope this new version satisfies the high-quality standards of your journal.

Best regards,

Jaume Segura-Garcia

Reviewer 2:

Comment:

1. The system model is not clear. What is the specific requirement for the use case?

Answer:

In this paper we describe a healthy route planner application within the activities carried out within the Eco4rupa project. The goal of this application is to calculate healthy walking and/or cycling routes according to the particular citizen's profile and needs, in particular when they require specific care in case of respiratory diseases.

This application has been designed to reduce the exposition to specific air pollutants based on these profiles. We show several profiles as examples, such as citizens with asthma and pregnant women, where each profile defines a set of weights for the different air pollutants taking into account scientific studies (as depicted in Section 4.1) that determine the hazardousness of their exposure.

In order to determine and analyze the different routes from a pollution exposure point of view, we estimate the distribution of the pollutants over the city map grid. This process is done using interpolation techniques, in particular Ordinary Kriging, and mapping the quantity of each pollutant over this grid of the city map. Then, weighting the different pollutants we define a complex metric that is used as a cost function in order to run the shortest path algorithm.

Thus, the requirement for the use case is to provide a specific profile (or choose a default one trying to minimize a global pollution exposure), and these profiles are given by a set of weights for the different air pollutants.

We have included this detail in the new version of the manuscript.

Comment:

2. The routing planner is too simple. How is the routing actually designed?

Answer:

Of course, we run a well-known shortest path first algorithm. However, our contribution and novelty is given by this complex metric used with the goal to minimize the pollution exposure for the citizens based on their profile, as well as the whole process carried out to do this. 

As mentioned before, once we know the estimated quantity of each pollutant over this grid of the city map, we use this complex metric to run the shortest path first algorithm. As a result, it provides the route with less pollution exposure according to the citizen’s profile.

Thus, once the citizen asks for a specific route from one source to a destination, selecting the route mode (walking and/or cycling) and based on the latest Air Quality (AQ) information retrieved from the monitoring nodes (less than 10 minutes that is the sampling frequency for AQ monitoring), we estimate the concentration of these air pollutants over the city map, calculate the best healthy route (less pollution exposure) and show it to the citizen. 

We have included this explanation in the new revised version of the paper to improve its reading. 

Comment:

3. The validation part is not explicit. The KPI for for data collecting or routing is not provided.

Answer:

Thank you for your comment. In this paper, we focus on the whole system for AQ monitoring and the added value given by the offered service to provide healthy routes, taking into account the citizen’s profile. We have analyzed its integration and functionally, as a proof of concept.

Since our goal is to reduce the exposure to the pollution that could be critical for some citizens, we have carried out the performance evaluation of this route planner under this point of view. Thus, we have analyzed the performance of the route planner comparing the results with different scenarios. 

We have defined a set of KPI for this purpose. These KPI used to analyze the results are also depicted in the paper, in Section 5.1. In particular as shown in Table 3, we estimate the average pollution reduction percentages for each test by:

• % of pollution reduction (PR) 

• % of increased distance (ID)

Specifically, our target for the router planner is given by a minimum threshold of 15% of PR with an increase in distance traveled of less than or equal to 10%.

Thus, in order to calculate these parameters, in Table 2 we have measured the following attributes and values:

-the total cost with asthma (C. Asthma) and total distance with asthma (D. Asthma)

-the total cost for pregnancy (C. Preg.) and total distance for pregnancy (D. Preg.) 

-the cost with Shortest Path First (SPF) assuming asthma (C. SPF Asthma)

-the cost with shortest path first assuming pregnancy (C. SPF Preg.) 

-the distance with SPF (D. SPF)

In summary, our approach can lead to an approximately average reduction in pollution exposure of 17.82%, while experiencing an approximately average increase in distance traveled of 9.8 %. In detail, we confirm that our algorithm succeeds in reducing exposure to polluted air by 18.72% in cases of asthma, which implies an increase of 9.27% in distance traveled. For pregnant women, our algorithm succeeds in reducing exposure to air pollution by 16.91%, increasing the distance traveled by 10.32%. 

Therefore, it is shown that the route planning system achieves the goal of reducing exposure to high air pollution. These results support the effectiveness of the method in providing healthier navigation options among users, without imposing a significant increase in distance. Also, there is evidence that the optimal route for health is consistent for both the pregnancy and asthma trials. 

Comment:

4. The presentation should be thoroughly improved.

Answer:

Thank you for your suggestion. During this revision process, we have improved the wording and presentation of the paper. 

Notice that for clarity we summarized our results in Tables 2 and 3, where we show the details for the different tests carried out for a set of routes given by different sources and destinations, under different scenarios (different time tables, congestion road traffic and routes) and different user's profiles, with special interest on citizens with asthma and pregnant women.

For clarity, only a set of selected routes are shown in Figures 10, 11 and 12 (in the revision) comparing the results for the different scenarios, different profiles, given by citizens with asthma, pregnant women and the default result based from the shortest path first using distance as a metric.

Comment:

5. More similar existing works can be compared as follows:

[1] Maximizing spatial–temporal coverage in mobile crowd-sensing based on public transports with predictable trajectory. International Journal of Distributed Sensor Networks. 2018;14(8). 

[2]  "Collaborative Learning of Communication Routes in Edge-Enabled Multi-Access Vehicular Environment," in IEEE Transactions on Cognitive Communications and Networking

Answer:

Thank you for your comment and suggested references. These interesting references do not fit exactly in the scope of our approach. However, we have extracted similarities with regard to route discovery with a different goal. 

This new approach has been mentioned in the related work as a different alternative, citing them.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors addressed all my comments and I would like to propose the paper for publishing.

Just a few comments to the authors:

- NB-IoT is not part of the 5G but 4G.

- Some recommendation about deployment of AQ monitoring station could be found in:

https://fortress.wa.gov/ecy/publications/documents/1602021.pdf

https://www.london.gov.uk/sites/default/files/air_quality_monitoring_guidance_january_2018.pdf

Minor editing of English language required.

Author Response

Dear reviewer,

Thank for your comment. 

NB-IoT and LTE-M were introduced in 3GPP LTE release 13, but they were continued and improved in 3GPP releases 14 and 15 (which are the ones defining 5G).

Best regards,

Jaume

Reviewer 2 Report

The authors have addressed my concerns.

The presentation is OK.