The Devil is in the deployment. Discussion By
Dineo Alan Nkuna. 

Scene 1

In This Presentation I Will Cover. Close up of an olive branch on a sunset. 

Scene 2

The background of the study The main methods used throughout the study The main findings of the study

Introduction. Subsistence farmers form a large population of South Africa and play an imperative role in the economy and lately, and South Africa has faced drastic and unpredictable environmental changes This has left many farmers with increased temperatures and rainfall, which leaves the farmers with destroyed crops and farmland and hydroponic farming, which grows crops in a controlled environment, often without soil, is suggested as a solution However, hydroponic farming requires skill, knowledge, and constant monitoring. This project aims to explore the use of Artificial Intelligence in the form of Machine Learning models and the Internet of Things to automate and monitor hydroponic systems.. 

Scene 3

Subsistence farmers form a large population of South Africa and play an imperative role in the economy and lately, and South Africa has faced drastic and unpredictable environmental changes

 This has left many farmers with increased temperatures and rainfall, which leaves the farmers with destroyed crops and farmland and hydroponic farming, which grows crops in a controlled environment, often without soil, is suggested as a solution

 However, hydroponic farming requires skill, knowledge, and constant monitoring. This project aims to explore the use of Artificial Intelligence in the form of Machine Learning models and the Internet of Things to automate and monitor hydroponic systems.

The Background. This research took place in KwaZulu-Natal and the Eastern Cape provinces of South Africa. High poverty and food insecurity levels often characterize these provinces. Both provinces also house the highest concentrations of subsistence farmers in South Africa, presenting us with an opportunity to work closely with subsistence farmers while co-deploying hydroponic tents and exploring ways for addressing food insecurity . This paper reports on the deployment in KwaZulu-Natal. There is a global agreement that the increasing climate change issue poses a significant threat to humanity because it increases temperatures and cause more unpredictable rainfall patterns . In South Africa, climate change is expected to increase food insecurity and poverty levels among rural people who depend on subsistence farming . It is, therefore, important to note that subsistence farmers account for over four million of the South African population. These farmers grow crops for their own family’s consumption and to reduce food insecurity as surplus produce is sold informally. Subsistence agriculture is crucial for ensuring rural people’s livelihood and food security. Unfortunately, climate change directly impacts the poor because they cannot withstand climate-related challenges. The communities in the Eastern Cape and KwaZulu-Natal regions especially experience poverty in South Africa. This increases the number of rural subsistence farmers forced to grow crops to feed their families. Sadly, climate change has negatively affected their efforts due to extreme weather conditions, namely droughts, increased temperatures, and storms, which cause soil erosion and have them unable to cultivate the land, resulting in massive crop losses . And so, it's not surprising that different farming methods, such as hydroponics, are on the rise.. 

Scene 4

This research took place in KwaZulu-Natal and the Eastern Cape provinces of South Africa. High poverty and food insecurity levels often characterize these provinces. Both provinces also house the highest concentrations of subsistence farmers in South Africa, presenting us with an opportunity to work closely with subsistence farmers while co-deploying hydroponic tents and exploring ways for addressing food insecurity . This paper reports on the deployment in KwaZulu-Natal. There is a global agreement that the increasing climate change issue poses a significant threat to humanity because it increases temperatures and cause more unpredictable rainfall patterns . In South Africa, climate change is expected to increase food insecurity and poverty levels among rural people who depend on subsistence farming . It is, therefore, important to note that subsistence farmers account for over four million of the South African population. These farmers grow crops for their own family’s consumption and to reduce food insecurity as surplus produce is sold informally. Subsistence agriculture is crucial for ensuring rural people’s livelihood and food security. Unfortunately, climate change directly impacts the poor because they cannot withstand climate-related challenges.  The communities in the Eastern Cape and KwaZulu-Natal regions especially experience poverty in South Africa. This increases the number of rural subsistence farmers forced to grow crops to feed their families. Sadly, climate change has negatively affected their efforts due to extreme weather conditions, namely droughts, increased temperatures, and storms, which cause soil erosion and have them unable to cultivate the land, resulting in massive crop losses . And so, it's not surprising that different farming methods, such as hydroponics, are on the rise.

The Background (part 2 of 2). Green and dry land. 

Scene 5

The communities in the Eastern Cape and KwaZulu-Natal regions especially experience poverty in South Africa. This increases the number of rural subsistence farmers forced to grow crops to feed their families. Sadly, climate change has negatively affected their efforts due to extreme weather conditions, namely droughts, increased temperatures, and storms, which cause soil erosion and have them unable to cultivate the land, resulting in massive crop losses . And so, it's not surprising that different farming methods, such as hydroponics, are on the rise.

The method (part 1 0f 5 ). Sweetwaters - KwaZulu-Natal Province: Sweetwaters is a low-income rural area located 97 km outside Durban, South Africa in the uMgungundlovu district. This area has an average household The Devil is in the deployment: Lessons learned while deploying an AI and IoT-enabled hydroponics grow tent with rural subsistence farmers in South Africa COMPASS ’24, July 08–11, 2024. To recruit participants, we partnered with the Human Sciences Research Council in KwaZulu-Natal , who have been actively involved and working closely with farmers in the province.. 

Scene 6

Sweetwaters - KwaZulu-Natal Province: Sweetwaters is a low-income rural area located 97 km outside Durban, South Africa in the uMgungundlovu district. This area has an average household The Devil is in the deployment: Lessons learned while deploying an AI and IoT-enabled hydroponics grow tent with rural subsistence farmers in South Africa COMPASS ’24, July 08–11, 2024. To recruit participants, we partnered with the Human Sciences Research Council  in KwaZulu-Natal , who have been actively involved and working closely with farmers in the province.

The Methods (part 2 of 5)
Participant Demographic Information, Land Size and Farming experience. 

Scene 7

The Methods (part 2 of 5)
Participant Demographic Information, Land Size and Farming experience

Gender Age Farming land size Number of years farming F 67 22m x 80m 5 F 69 38m x 21m  21 F 57 1 Hectare 17 F 48 7m x 6m 3 M 44 50m x 50m 5

The community liaison recruited five farmers, two of which were community growers who provides food to community members free of cost. The information of our participants is detailed in this Table

The method (part 3 0f 5). Electronics protoboard. 

Scene 8

We decided to use technology probes. Technology probes are simple, flexible, adaptable technologies that help researchers understand the needs and desires of their participants, allow researchers to field test technologies in a real-world setting, and enable users to think about new technologies . We built and deployed a fully functional IoT and AI-enabled hydroponics tent and accompanying mobile application to allow our participants time to experience the technologies to elicit easier and more meaningful conversations about the community’s design requirements.
  We were fully aware that the system, in its initial form, did not consider the design requirements and needs of the community but would instead serve as a probe to start conversations and hopefully enable participants and researchers to think about and express the system’s design requirements, and share their experiences To mitigate the erratic weather patterns currently present in South Africa  nsors transmitted real-tim

The method (part 4 of 5). We implemented a fully enclosed hydroponic system, which included temperature, light, and humidity control. We, therefore, used a dark box hydroponic grow Tent with 80cm x 80cm x 160cm dimensions, which contains a 72-hole step-down tiered hydroponic food grade PVC grow stand . The humidity and temperature in the tent are controlled with an oscillating multi fan and a mixed-flow inline extractor fan . For the necessary lighting and heat we made use of a full spectrum LED grow light . The necessary nutrients are supplied via an eight-liter water reservoir fitted with a submersible water pump 11 capable of pumping water meters to reach the top tier of the growing stand. The system is gravity-fed from here with an outlet pipe draining into the water reservoir to complete the water circuit. IoT network and mobile application . To enable the farmers to visualize and help monitor and maintain the controlled environment, we installed a DHT22 temperature and humidity sensor 12, an ambient light sensor as well as pH and Electrical Conductivity sensors. These sensors transmitted real-time. 

Scene 9

We implemented a fully enclosed hydroponic system, which included temperature, light, and humidity control.
 We, therefore, used a dark box hydroponic grow Tent with 80cm x 80cm x 160cm dimensions, which contains a 72-hole step-down tiered hydroponic food grade PVC grow stand .  The humidity and temperature in the tent are controlled with an oscillating multi fan and a mixed-flow inline extractor fan . For the necessary lighting and heat we made use of a full spectrum LED grow light . The necessary nutrients are supplied via an eight-liter water reservoir fitted with a submersible water pump 11 capable of pumping water meters to reach the top tier of the growing stand. The system is gravity-fed from here with an outlet pipe draining into the water reservoir to complete the water circuit.
 IoT network and mobile application . To enable the farmers to visualize and help monitor and maintain the controlled environment, we installed a DHT22 temperature and humidity sensor 12, an ambient light sensor as well as pH and Electrical Conductivity sensors. These sensors transmitted real-time

Para-meter Sample Optimal-Range EC 0-4.5 1.8. - 2.5
 Humidity 60-80 6.5-7.5 PH 4.0-7.0 5.8-6.3 Temp 15-28 18.3-25.0

Sample and Optimal values in synthetic dataset

The method (part 5 of 5). We used the datasets to train the needed RFC models with the default parameters of the sci-kit learn Random Forest Classifier library. X and y datasets representing independent and dependent variables required by the RFC were created from the synthetic datasets. These were further split into training and testing sets in the ratio 80:20, enabling the models to be evaluated using separate data from that used for training. The trained models were evaluated using accuracy scores and confusion matrices. We finally deployed the ML models to the Arduino Micro-controller by porting the Python model to C++ using the micromlgen Python package. The models predicted the class labels based on sensor data, and the AI took action accordingly, as specified in the second table. The reliability and validity of the program were achieved through unit testing and boundary value testing . The parameter values at the class boundaries were used to check if the program classified correctly on the transition between low to optimum status and between high and optimum and if the AI took the correct action. The hydroponic stand drains slowly using only gravity to drain the water. This process takes time and enables the plants to survive up to 48 hours until all the nutrient-rich water is drained from the hydroponic stand.. 

Scene 10

We used the datasets to train the needed RFC models with the default parameters of the sci-kit learn Random Forest Classifier library. X and y datasets representing independent and dependent variables required by the RFC were created from the synthetic datasets.  These were further split into training and testing sets in the ratio 80:20, enabling the models to be evaluated using separate data from that used for training. The trained models were evaluated using accuracy scores and confusion matrices. We finally deployed the ML models to the Arduino Micro-controller by porting the Python model to C++ using the micromlgen Python package. The models predicted the class labels based on sensor data, and the AI took action accordingly, as specified in the second table. The reliability and validity of the program were achieved through unit testing and boundary value testing . The parameter values at the class boundaries were used to check if the program classified correctly on the transition between low to optimum status and between high and optimum and if the AI took the correct action. The hydroponic stand drains slowly using only gravity to drain the water. This process takes time and enables the plants to survive up to 48 hours until all the nutrient-rich water is drained from the hydroponic stand.

Main Findings. In the study, it became clear that technology alone, including grow tents and AI, could not successfully produce crops without the involvement of rural subsistence farmers. These farmers played a crucial role in monitoring plant health, manually checking for pests, root rot, nutrient burn, leaf spot, and humidity-related issues. They also pruned unhealthy leaves to prevent stunted growth. Human intervention was essential; for instance, unconnected power cords led to potential crop damage, highlighting the need for regular checks. Farmers began weekly inspections, documenting plant progress and communicating issues to researchers. This collaboration fostered a bond, transforming farmers into co-growers alongside technology. Despite facing crop failures, farmers expressed optimism about hydroponic farming's potential. They noted faster growth rates compared to traditional soil methods, indicating a need to adjust expectations for harvest timelines. Older farmers appreciated the reduced physical demands of hydroponics, allowing them to farm more easily. This shift in mindset reflected a broader acceptance of new practices and the benefits of technological integration in agriculture. The researchers also supported the farmers by providing seedlings whenever replanting occurred, ensuring they could continue to cultivate despite setbacks.. 

Scene 11

In the study, it became clear that technology alone, including grow tents and AI, could not successfully produce crops without the involvement of rural subsistence farmers. These farmers played a crucial role in monitoring plant health, manually checking for pests, root rot, nutrient burn, leaf spot, and humidity-related issues. They also pruned unhealthy leaves to prevent stunted growth. Human intervention was essential; for instance, unconnected power cords led to potential crop damage, highlighting the need for regular checks. Farmers began weekly inspections, documenting plant progress and communicating issues to researchers. This collaboration fostered a bond, transforming farmers into co-growers alongside technology.

 Despite facing crop failures, farmers expressed optimism about hydroponic farming's potential. They noted faster growth rates compared to traditional soil methods, indicating a need to adjust expectations for harvest timelines. Older farmers appreciated the reduced physical demands of hydroponics, allowing them to farm more easily. This shift in mindset reflected a broader acceptance of new practices and the benefits of technological integration in agriculture. The researchers also supported the farmers by providing seedlings whenever replanting occurred, ensuring they could continue to cultivate despite setbacks.