AI-Enabled Drones: Agriculture Done Smartly

When most people think of drone videography, odds are they immediately think of videos over mountains, lookouts for unmanned aerial and underwater vehicles, or remotely operated vehicles. Today, more and more drone data is collected across the globe.

Combined with AI technologies, this data can be exploited. Useful information can be extracted for the benefit of society. Smartphones and AI-enabled drones can be transformed into tools that allow one to maintain orchard farms from production until the end product reaches a consumer’s table. The combination of internet-of-people (IoP)-enabled blockchain technologies ensures that the balance between supply and demand is efficient for each cycle of the season.

Furthermore, when digitizing the entire fruit-growing process, AI-enabled drone technology allows you to save time connecting supply and demand. From the local producers, a global supply can be estimated, and demand can be estimated from the local markets and online platforms.

Agriculture and breeding represent an integral part of the local economies of Burkina Faso, Mali, and Senegal, with nearly 60 million people altogether. In 2020, agriculture contributed around 18.4 percent to the GDP of Burkina Faso, while 32.59 percent came from industry and 40.83 percent from the services sector. Almost 90 percent of the rural population continues to rely on it directly for a living.

According to the International Trade Administration, agricultural, livestock, poultry, and transformation activities represent over 33 percent of Mali’s GDP and employ over 70 percent of Malians as of 2018. And according to the Food and Agriculture Organization, the agricultural sector is one of the pillars of Senegal’s economy, with an estimated contribution of 15 percent to the GDP in 2018 and a significant portion of the population in rural areas (60%).

However, many agricultural sub-sectors, such as shea butter, mangoes, peanuts, cashews, and biofuels, are largely untapped, and AI can help improve production while lowering costs.

Guinaga is an online platform focused on reducing the gap between demand and supply in underserved areas. It aims to follow up items step by step from production sources to consumptions using IoP, data-driven operations, mean-field-type game theory, and collective intelligence. The “manguechain” project of Guinaga uses a generative adversarial network based on fruit counting by taking aerial pictures of the mango trees in Siby, Sikasso (Mali), Bobo-Dioulasso (Burkina Faso), and Casamance.

The AI-based video processing of the mango trees allowed a more reliable estimation of the production two months before their term, saving 5 % of the losses and wastes. Mango production plays a major role in terms of the realization of food security and the improvement of the livelihoods of rural populations. The method was also used to count shea trees from AI-enabled drone images and videos.

The second project is conducted by the Grabal platform and is in the area of livestock and poultry. Grabal connects breeders and buyers. It uses an incentive loop design for breeders and buyers to solve the mismatch between supply and demand. In order to estimate the supply, producers who have food and transportation issues aim to minimize multiple pain points: patrolling pastures to locate cattle, monitoring health indicators on the road, and finding animal food.

AI-enabled drone videos have been added to the database to enrich recognition and counting. The MoutonChain project of Grabal has used AI-enabled drone technologies for monitoring animals and finding animal food in Burkina Faso, Senegal, and Mali.

Grabal used cameras and drones to capture pictures and videos from different angles. These videos are then transmitted to a Grabal server. A recently developed algorithm based on a distributionally robust generative adversarial network is implemented and used to identify areas with grass and animal food on the path of the herd, track footprints, and collect information about animals.

During our experiments, the breeders were able to monitor and track their livestock. This assists greatly in preventing theft and securing food on their way. Data from AI-enabled drones collected at local cattle markets, such as:

These markets were analyzed and provided an estimation of the weekly supply. The buyers’ weekly demand was estimated using various sources, including market AI-enabled drone data, a feedback survey from sellers, physical buyers, and Grabal’s online demand data.

The methodology takes into account the insecurity aspect as well as the climate change associated with the zone change. MoutonChain is a sub-blockchain of Grabalton that aims to link sheep from production to sale, specially designed for traceability in Tabaski. The AI-powered framework followed Banamba and Nara breeders, who conduct crossbreeding and interbreeding for animals such as:

The genetic diversity of these local names as well as the intergenerational genetic diversity needs to be studied. The AI-enabled drone methodology was especially useful with the MoutonChain follow-up, which peaks during the Eid al-Adha/Tabaski celebrations in the area from 2019 to 2022. The supply-demand mismatch creates losses and fluctuations for breeders and plenty of families.

The Grabal team retraced part of the journey of the shepherds of Nioro, Nara, Banamba, Douentza, Koro, Fatoma, San, and Bla towards the Malian capital markets and other destinations such as Abidjan, Dakar, and Conakry. Some breeders had to walk with the animals from the villages to the urban markets.

This year, there were a lot more animals that walked from the breeding area to the market than those that were transported by trucks. The observation is important since when there is less food over a long year like this one, these rams have less energy to travel such a long distance. The AI-enabled drone data collected over the roads over the three preceding months of the event is consistent with this observation.

During the “MoutonChain” peak this year, over 10 million animals were reported as being sold, including cows, sheep, and goats, across 105 sites. The reported market cap exceeded one trillion FCFA (the local currency of the area). In comparison to the previous three years, there has been a decrease in supply and an increase in price.

Minimum ram and sheep prices were less than 50 000 FCFA depending on weight and appearance, and they were frequently far from capital cities. The highest prices exceeded 50 million FCFA. Some recent ram stars, such as the Chadian Lake-originated rams and the Bali-Bali, attracted our attention by exploiting the data. An AI-assisted genetic diversity sequencing of these rams needs to be examined better to understand the market’s needs in the future.

From August 2019 to September 2022, AI-enabled drone data was carefully collected by our teams from Timadie, L&G Lab, MFTG, Guinaga, Grabal, WETE, 1m+, and SK1 Sogoloton. The data was combined with camera data. These include videos as well as pictures.

The tool improved understanding of supply-demand mismatches in shea butter production, mango production, and potato transformation, among other things. AI-enabled drones can identify resources and reduce livestock losses at the orchard level as well as on the global market, representing a significant part of the local economy. As such, we can conclude that the agriculture sector will definitely see more applications of AI in the next couple of years.