How IoT and machine learning are automating agriculture

A new generation of farmers is turning to technologies such as the internet of things (IoT) and machine learning to automate agricultural production, alleviating the need to toil on the land while keeping a watchful eye on their crops.

Joining their ranks is Phoebe Xie, director and co-founder of Singapore-based agrotechnology startup AbyFarm. Teaming up with technology service provider SPTel, Xie is building a smart hydroponics farm in a greenhouse that uses a plethora of IoT sensors, including video cameras, to keep the farm humming around the clock.

“To run a self-regulating farm at optimal temperature with optimal water and nutrient supply, and to control the quality and taste of vegetables and fruits, we need IoT sensors,” she said, adding that the farm and its sensors collect and monitor thousands of data points, including humidity and temperature.

With the data and in certain environmental conditions, processes and actions are automatically triggered to protect crops from the elements. For instance, if the temperature or humidity gets too high, fans, water curtains and roof shades are activated in the greenhouse.

Xie said the smart farm, located at a rooftop carpark in Singapore, is also equipped with sensors that monitor the pH and electrical conductivity levels of water. Among the sensors, which are connected to a Lora low-power wide area network, is a dozer that automatically releases acidic or alkaline nutrients to maintain optimal pH levels. 

Whereas the farmers of old had to physically inspect their crops, AbyFarm’s agronomists and farmers can do so remotely using video cameras and image recognition technology to identify crops that might be succumbing to disease.

“If the image recognition finds that a plant is likely to be sick, the farmer will be alerted to go onsite to check how the plant is right now,” said Xie, adding that agronomists will also advise farmers on taking corrective actions to prevent disease or restore the health of their crops.

Like any machine learning system, AbyFarm’s algorithms get smarter over time. Each time a crop is affected by disease, data about the occurrence and corrective action is fed into the system, enabling its algorithms to formulate solutions for other farmers with similar crop issues – without consulting an agronomist.

With that, said Xie, even those who are new to the field can use the system to guide them on treating diseased crops, which can be isolated from other crops to prevent cross-infections.

AbyFarm’s automated farming system, which is hosted on SPTel’s private cloud, can also advise farmers on the best time to transplant their crops after germination and harvest them later for sale, said Xie.

Heng Kwee Tong, vice-president for engineering and customer solutions at SPTel, said the company has built up a software-defined platform to help companies like AbyFarm overcome the challenges of deploying applications such as IoT.

“The common struggle that companies like AbyFarm face with building an IoT solution is that it is quite intensive in terms of ICT infrastructure,” said Heng. “You’ve got to find the radio connectivity like Lora, subscribe to a service provider and connect your sensors to a sensor collector.

“We make it easy by investing in a platform and because we are a service provider with edge hub assets for deploying radio connectivity, all AbyFarm needs to do is to bring their sensors onboard.”

Going forward, Heng said SPTel is looking to support next-generation video analytics capabilities through edge cloud services that can be used to crunch workloads closer to where they reside, reducing latency.

AbyFarm was one of six companies to win contracts from the Singapore Food Agency last year to build urban farms at nine rooftop carparks in public housing estates across Singapore.

The goal is to collectively produce about 1,600 tonnes of vegetables a year in a bid by the city-state – where less than 1% of land is used for agriculture – to become more self-sufficient in meeting the nutritional needs of its people.

That could well be achievable with technology that makes farming more efficient, scalable and less labour-intensive than before. “We are willing to train and educate the next generation in agrotechnology, because now you don’t need 10 farmers to take care of one farm,” said Xie. “It’s just looking at the dashboard and, with the alerts, doing what is required.”