3.1.2. Actuators

	- (a) Ambient display: The PRISMA is an environmental display to support learning scenarios [46]. See Figure 3a. The PRISMA can display information with its 24 LED ring, 8 × 8 LED matrix, and a liquid crystal display. This display was made available to students so they could configure it based on their interaction needs. E.g. Group #3 configured the PRISMA to provide a range of colours between blue and yellow derived from the humidity returned by the sensor.
	- (b) Interactive touch screen display. This interactive display was installed to present real-time information from all planters making sensors data visible with visual metaphors. The main screen includes a menu where the user can select which planter to explore in detail. See Figure 2 (left).
	- (c) Mobile messaging system. Group #5 configured the IoT system to send alerts by means of Telegram instant messaging app, when specific events occur. Figure 3b shows some examples of the configured alerts: "The plant has not enough light", "Congratulations! The plant is growing under the best conditions", "Security alert! There is no Internet connectivity. Check the planter". The system also notifies when the irrigation has started and finished.

#### 3.1.3. Computer and Microcontroller

The board (or microcontroller) is the main component connecting the rest of the subsystems. Its role is to receive the captured data, processing the data, and send orders to actuators to maintain the plant in the best conditions. Likewise, the processor sends data to the IoT platform where it is stored and monitored according to consistent rules. The system is continuously active to periodically read, validate, and write the value of the sensors. The IoT systems implemented included both Up2 board (computer) and/or ESP32 (microcontroller).

• *UP*<sup>2</sup> *board*. The UP Squared board is an x86 maker board based on the Intel. The UP boards are used in IoT applications, industrial automation, or digital signage. This board is equipped with an Intel Celeron N3550 and Intel Pentium N4200 System on Chip (SoC), 40 pins, 8 GB RAM, Ethernet, HDMI, and USB connectors. This case study was carried out along the semester of the

Computer Based Systems module. Hence, students were urged to implement their IoT systems using this board.

• *ESP32 microcontroller*. ESP32 is a series of low-cost, low-power SoC microcontrollers with integrated Wi-Fi and dual-mode Bluetooth. The ESP32 employs a Tensilica Xtensa LX6 microprocessor. ESP32 includes built-in antenna switches, power amplifier, low-noise receives amplifier, filters, and power-managemen<sup>t</sup> modules. In this case study, the most advantageous groups were able to adapt the processing capacity of the system and replace the board with a microcontroller.

**Figure 3.** Feedback services configured: (**a**) Prisma, a visual feedback display; (**b**) Telegram messenger to receive alerts; (**c**) IoT cloud platform. Thingsboard.io desktop dashboard.

#### 3.1.4. IoT Cloud Platform

In the initial phase, students should agree on which IoT cloud platform would be used to persist and monitor data from planters. A brainstorming session was organized to explore and test existing IoT platforms. The following features were considered to take the decision: REST API, authentication type, protocols for data collection (i.e., MQTT, HTTP, CoAP), and analytics provided. The following IoT platforms were considered: Azure IoT, DeviceHive, Kaa IoT Platform, Mainflux, SiteWhere, Thingsboard.io, Thinger.io ThingSpeak, SSo2, and Zetta. Finally, Thingsboard.io received more votes from the students at the end of the brainstorming session.

All groups created a profile in Thingsboard.io to adapt the IoT platform to the specific requirements of each planter. These were the features used by the groups:


#### 3.1.5. Irrigation System

Agricultural Engineering students and teachers were responsible for designing and installing the automatic irrigation system. The installation involved the setup of planters, motorized valves, droppers, water pipes, water counters, water filters, and pressure switches. This subsystem is formed by a relay that controls the opening and closing of a solenoid valve, which allows or not the passage of water. The relay receives a direct order from the computer/microcontroller configured by CE students. CE and AE teachers provided on-demand support but also regularly reviewed the progress performed by each group.

#### *3.2. Mutlidisciplinary Teamwork on IoT*

This study is aimed at exploring the e ffects of multidisciplinary teamwork in learning performance (RQ3). Hence, we explored the extent to which teamwork subscales can vary based on the multidisciplinarity in the composition of the work groups. The scores obtained demonstrated adequate internal consistency for 6 out of 8 scales (see Table 1). Values for Cronbach's alpha ranged from 0.71 to 0.96 revealing su fficient score reliability for "perceived learning", "expected quality", "team cohesiveness", "workload", "satisfaction" and "collaborative behavior". Nevertheless, values for Cronbach's alpha ranged from 0.42 to 0.52 revealing insu fficient score reliability for "cooperativeness" and "task complexity", and they were consistently discarded for the rest of the analysis.

Means and standard deviation were calculated taking into account the composition of the groups. The results illustrated in Table 2 show that by taking together all the values of the teamwork scale in a range of 1 to 7, group B obtained an average rating slightly higher than group C. On the contrary, group A obtained a rating of 0.59 points lower than group B. Looking at the subscales individually, the results concluded that group B obtained the highest scores for "collaborative behaviour", "satisfaction", "team cohesiveness", and "perceived learning". On the other hand, group C obtained the highest scores in "expected quality" and "perceived learning". On the contrary, group A obtained the lowest scores in all subscales.


**Table 2.** Teamwork means and standard deviations by group composition.

With regard to the marks obtained by the students in the final evaluation on a scale of 0 to 10 (10 being the best score), the results showed that group B obtained the upper average grade, followed by group C. The group A scored 1.60 points lower than group B, and 1.16 points lower than group C.

The results obtained in the Shapiro–Wilk test (p-value = 0.00057) confirmed the normal distribution of the overall teamwork data sample. Exploring the teamwork subscales independently, p-values lower than 0.05 and the observations of the Q-Q plots confirm that "Collaborative behaviour", "Satisfaction", "Team cohesiveness", "Expected quality", and "Workload" samples are normally distributed. However, the p-values obtained in the "perceived learning" sample deviate from normality. Hence, "perceived learning" was consistently discarded in the ANOVA test.

An ANOVA test was performed to identify significant differences between the mean values (Table 3). On the one hand, the test resulted in significant values for the grades obtained in the evaluation (See grades in Figure 4a). On the other hand, the test resulted in Pr(>F) = 0.17 (which is slightly higher to the coefficient of significance 0.1) for the overall teamwork scales, and consequently non-significant values for the overall teamwork scales. Exploring the values obtained in the subscales, the ANOVA test resulted in significant values for "expected quality" (Figure 4b).


**Table 3.** Analysis of Variance ANOVA. Significance Pr(>F) > 0.1.

**Figure 4.** Boxplots contrasting group composition.

This study aimed at exploring the significance of the relationship between the means obtained in teamwork subscales and the grades. Additionally, we investigated the relationships within teamwork subscales (Table 4). We anticipated that learning performance (grades) would be positively correlated with teamwork subscales. Contrary to our expectations, though, the results of the analysis do not depict a significant correlation between them. Additionally, we aimed at exploring potential correlations within teamwork subscales. The results from the correlation analysis show that there is a significant very strong positive correlation between "cohesiveness" and "satisfaction". Similarly, "cohesiveness" has a strong positive correlation with "expected quality", "perceived learning", and "collaboration". Likewise, "expected quality" has a strong positive correlation with "satisfaction".

#### Frequently Used Channels to Communicate

Pr(<F) Significance codes: 0 '\*\*\*' 0.001 '\*\*' 0.01 '\*' 0.05 '.' 0.1 ' ' 1.

In this case study, we aimed at exploring the most frequently used channels to communicate while working in groups (RQ4). Students were assigned an optional task in which they could report once a week (using an online form) what channels they had used to communicate among team colleagues. There were 272 reports from 47 different students along the semester (Table 5). There was a mean of 5.78 reports by student. Whatsapp, Telegram and Teams were the most frequently used channels.


**Table 4.** Pearson's correlation analysis (\* Correlation significance < 0.01).

**Table 5.** Channels used to communicate among team colleagues. Frequency of usage.


*3.3. Educational Initiatives to Promote Environmental Awareness in the Campus*

Based on the lessons learned along the semester in which the IoT planters were developed, students were encouraged to sugges<sup>t</sup> alternative educational initiatives to promote environmental awareness in the campus. These were the most relevant actions:

There were different students suggesting interactive systems to improve IoT planters. Student #612 suggested developing a mobile app inspired on the Tamagotchi metaphor to take care of the IoT planters. The Tamagotchi would simulate the real IoT planter where participants could vary feeding substances or watering frequency to explore how the real planter would behave. Likewise, she suggested that the Tamagotchi might learn from the experience enabling using artificial intelligent algorithms. Following a similar approach, student #311 proposed to show a status summary of all planters in a large ambient display at the campus using Tamagotchi metaphor. Particularly, ambient displays were pinpointed as a key channel to promote environmental awareness in the campus. Student #313 proposed to show graphs illustrating water expenses by student-day, student-semester, etc., contrasting the number of litters with the ones contained in a glass of water, in a swimming pool, or the ones used by the IoT planters. Student #307 proposed to explore the operation of the planter to contrast the current energy supply with a self-supply system based on solar panels. Student #112 put forward using Twitter as feed to post actions done on the planters so anyone could track how IoT planters take care of the plants. Additionally, he also suggested posting environmental variables collected by the sensor making the data visible illustrations that catch the attention of users. These social interactions might help students understand the real environmental conditions in the campus and consequently to take action to reduce pollution in a local area. Similarly, student #311 suggested creating a hashtag (e.g., #smartIoTplanters) to track the evolutions of the plants across social networks. Student #307 proposed that voluntary students could take responsibility to take care of the IoT planters along the semester. He encouraged them to create an internal competition in which the students who were able to configure the planters with the most suited parameters according to plant, environmental conditions, but also considered introducing machine learning algorithms would save extra money for credits to study

related modules at the university. Thanks to advances in technology and machine learning, chatbots have become more popular than ever in recent years. Student #912 proposed using a chatbot (text or voice chatbot) to automate remote actions regarding the planters using commands. He named it *ChatbIoT* and these would be a sample command: (1)–User: "Hi Planter 3!, can you please water the lettuces at 18:00".–ChatbIoT: "For how long?". User: "Five minutes please". ChatbIoT: "Got it!. I will water the lettuces for 5 minutes at 18:00. Is that correct?". User: "Yes, thank you Planter 3!". ChatbIoT: "Okay, I will do so. Hasta la vista, baby.". The ChatbIoT might also help to remember recurrent actions performed on the planter (e.g., when was the last time that the hydroponic garden was fed).

Many students suggested gamification activities to raise awareness on environmental issues in the campus. Hence, di fferent actions and rewards were considered to implement these games:


Di fferent students suggested organizing interactive workshops in the course of the semester. For example, Student #107 suggested creating a physical open space to show the potential benefits of using renewable energy in the campus. Student #212 proposed to create a mailbox where students could contribute with ideas to make the campus more sustainable. Student #313 argued on the importance of healthy habits, foods and plants, and considered that regular workshops should be organised to make students aware of its benefits in long term. Similarly, student #607 urged to practically explore the composition, substances, and pesticides in cultivation soils using sensors, and to promote the understanding on bio food. Student #411 recommended organizing a hackathon to develop IoT software/hardware to reward the most eco-e fficient developments (i.e., computation, energy consumption).

There were di fferent students suggesting the creation of associations to promote achievable actions to support the sustainable development goals. Student #107 suggested that creating an association would have power enough to push a "Campus without plastics "action. Student #111 believed that an association should be able to promote the replacement of toilets to save water, i.e., "a presence sensor might count the number of litters wasted every day. A monthly summary could be presented visual displays at the campus". Student #312 recommended promoting situational awareness organizing excursions to recycling plants, or places a ffected by pollution. Additionally, she proposed to create a compost area in the campus garden. Last but not least, student #307 suggested the use of alternative media channels such as Radio Campus Sur to disseminate good practices in the context of environmental awareness.

#### **4. Discussion and Conclusions**

The rapid spread of IoT technologies has triggered the educational challenge of training future engineers to be able to design complex sensor-based ecosystems. In Industry 4.0, it is essential to educate students about the need to implement technological solutions that meet the objectives of the 2030 agenda, and to ensure that the engineers of the future consider ecological issues in their implementations.

This work presents the results of a case study in which students of Agricultural Engineer and Computer Engineer were assigned the task to create IoT systems to promote environmental awareness in the context of a university campus.

Multidisciplinary groups followed the PBL methodology to investigate IoT solutions that varied in the sensors, actuators, microcontrollers, plants, soils and irrigation system they used (RQ1). Section 3.1 elaborates on the three types of planters implemented, namely, *hydroponic system*, *vertical garden*, and *rectangular planters*. Sensors were configured considering the singularities (e.g., irrigation system) of each planter, but also reflecting on the particular care required by each plant. The holistic architecture of the components represented in Figure 2, shows that the implementation of the IoT planters covered up to 9 sensors (thereof 5 were analog, and 4 were digital). The working groups had the ability to engineer functionally di fferent systems using the same sensors (Section 3.1.1). Di fferent feedback systems were assembled to foster understanding on environmental issues in the campus i.e., LCD lights, ambient displays [46], mobile applications, and desktop-oriented dashboards (Section 3.1.2).

The overall architecture implemented in the case study shows that all groups were consistent featuring a three-layered architecture (Figure 5):


**Figure 5.** Frequently implemented three-layered architecture in IoT ecosystems.

The architecture described in this section is frequently implemented in di fferent engineering areas [47,48]. Therefore, IoT cloud platforms are providing improved services to facilitate a seamless

integration of IoT ecosystems (Section 3.1.4). Due to the scalable nature of the proposed architecture, the ecosystem is easy to extend, and to adapt towards further implementations in educational contexts.

This study achieved the objective of fostering environmental awareness on the campus by implementing systems that addressed the following important issues:


Moreover, students were able to sugges<sup>t</sup> alternative educational initiatives to raise awareness about environment issues in the campus (RQ2). The reported results show that the experience with the smart IoT planters facilitated students to envision multiple ingenious strategies to improve IoT planters towards educational purposes (Section 3.3). Different proposals suggested the inclusion of interactive visual and acoustic displays to increase the impact on students and employees of campus. The proposals included the use of social networks, and radio media to share the data collected by the sensors, and the creation of learning objects (in videos, text) to disseminate good practices in the use of water and energy or alert about the pollution in the campus. Furthermore, students suggested solutions based on gamification strategies, designing ingenious reward policies for active users raising environmental awareness. In this context, students were able to identify different achievable activities that could be promoted within any campus. These initiatives imply an important base of knowledge towards implementing further actions to promote environmental awareness in educational contexts.

Multidisciplinary education is key to tackle complex projects covering different areas of knowledge. The implementation of smart IoT planters in educational contexts demanded technical expertise to assemble hardware components and programing software interfaces. Nonetheless, it also required specific expertise on agronomics to select convenient plants, soils, planters, and, to install suitable irrigation systems. In this case study, we wanted to investigate how multidisciplinarity working in groups might impact learning performance and the expected quality of the outcomes (RQ3). The results presented in Section 3.2 show that multidisciplinary work among students from different areas of knowledge is ostensibly beneficial for learning. The groups with the highest degree of multidisciplinarity obtained higher *grades* (Figure 4a). Likewise, a similar effect was observed for the perception of *expected quality* of the work (Figure 4b). Exploring the subscales comprising teamwork, the results of the analysis show that the perception of group cohesion (*team cohesiveness*) and *satisfaction* to work in groups are strongly correlated (Table 4). These findings are consistent with previous research concluding that *expected quality* predicts satisfaction with teamwork [36]. These results rekindle the need to promote work in multidisciplinary groups from different areas of expertise to achieve a deeper knowledge, and to create functionally efficient IoT systems in Higher Education contexts.

Work-in-groups implied both face-to-face and remote to collaboration among group members. Hence, students could freely use different channels of communication to coordinate their activities. On the one hand, they had a default institutional platform (based on Microsoft Teams), which featured

synchronous and asynchronous messaging system. Alternatively, they might use any personal messaging applications or social networks. The results collected during the course show that students were more active using personal messaging application from their own mobile devices (Table 5): i.e., WhatsApp, Telegram. However, students were reluctant to use social networks for academic purposes: i.e., Facebook, Instagram (RQ4). These results must be interpreted together with recent conclusions on mobile learning showing that the new generation of teachers would be willing to use personal devices to guide students in educational contexts [49].

The complexity of this study lied in the implementation of a PBL activity across a full semester synchronizing students from two di fferent engineering degrees (Computer Engineering and Agricultural Engineering), campuses, and modules (Computer based systems, and Irrigation and Drainage Technology Systems). This approach is especially di fficult to implement since engineering degree syllabi are usually tightly restricted to specific areas of expertise. The results reported in this work represent an important technical knowledge base towards the implementation of IoT ecosystems in CE and AE educational contexts. Likewise, the conclusions of this evaluation provide evidence of the need to encourage work in multidisciplinary teams to train engineers towards Industry 4.0.

Further research should explore alternative associations between multidisciplinary groups to define suitable IoT architectures towards suggesting supplementary learning/teaching paths for industrial, civil, naval, aerospace, or forestry engineering studies.

**Supplementary Materials:** The following are available online at https://vimeo.com/392389154, Video eUrbanGarden: Campus Sur. UPM.

**Author Contributions:** Conceptualization, B.T., V.G.-A., C.G.-C., and S.B.-A.; methodology, B.T.; software, B.T. and S.B.-A.; validation, V.G.-A. and C.G.-C.; formal analysis, B.T.; investigation, B.T.; resources, V.G.-A. and C.G.-C.; data curation, B.T.; writing—original draft preparation, B.T.; writing—review and editing, B.T., V.G.-A., C.G.-C., and S.B.-A.; visualization, B.T.; supervision, B.T., V.G.-A., C.G.-C., and S.B.-A.; project administration, V.G.-A. and C.G.-C.; funding acquisition, B.T., V.G.-A. and C.G.-C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work received partial support from the Colegio Oficial de Ingenieros Técnicos Agrícolas de Centro, and the Escuela Técnica Superior de Ingeniería de Sistemas Informáticos. Likewise, this work has been co-funded by the Madrid Regional Government, through the project e-Madrid-CM (S2018/TCS-4307).

**Acknowledgments:** The authors would like to thank Pedro Martínez Jorde for his invaluable support and advice assembling the components. Likewise, the authors would like to thank Trinidad González González for her expert assistance checking English language and style.

**Conflicts of Interest:** The authors declare no conflict of interest.
