Applied AI in Agriculture at the Vancouver Linux Users Group
- Summary
This content was originally shared at the Vancouver Linux Users Group (VANLUG) event on July 19, held at the Burnaby Public Library, Metrotown Branch.
I'm an engineer with experience in both factory automation and agricultural automation. For about 10 years, I worked as a software engineer in factory automation and robotics. Later, I joined a project to develop and implement an AI-based agricultural system on farms. This experience gave me a unique perspective—what principles can be transferred from factories to farms, and what cannot. Please note that these insights come from my own project experience and might not apply to every case.
- Practical Challenges
One of the biggest challenges in agricultural automation can be described using the 5M model—a fundamental concept in quality control within factory automation. The 5Ms are: Methods, Measurement, Material, Machine, and Man.
While leading projects to implement AI systems on farms, I realized the limitations of our approach. Our AI system monitored plant environments—temperature, humidity, solar radiation, soil moisture, and CO₂ levels. It also used crop sensors to assess growth conditions. The collected data were uploaded to the cloud, where AI analyzed them to control irrigation, fertilization, and greenhouse conditions.
This worked well for Methods and Measurement:
- Methods: Irrigation and fertilization schedules were precisely maintained.
- Measurement: Sensor data were standardized, improving consistency.
However, the other three Ms—Material, Machine, and Man—remained outside the system's scope.
- Material (Seedlings): The quality of seeds and seedlings directly affects yield. Even the best AI system cannot overcome the early disadvantage in the growth.
- Machine: The cost-performance balance of machines is crucial. Some crops or environments limit which machines can be used, and AI can't overcome mechanical constraints.
- Man: Full automation often requires high external costs, so human flexibility remains essential.
- Technical Challenges
Agriculture faces technical challenges that do not exist in a factory.
- High disturbance condition: Factories maintain environments—constant temperature and lighting ensure predictable production. Farms, on the other hand, depend on natural sunlight and external weather conditions. Even in greenhouses, full control is unusual.
- Dynamic Growth: Factory machines operate periodically and predictably. Crops, however, grow uniquely—no two plants are identical. Growth speed fluctuates with light and temperature. While AI holds great potential to model crop growth, there is still no perfect simulator for predicting real-world conditions. Some uncontrollable factors always remain.
- Key to Success
To address these challenges, I immersed myself in agricultural work and carefully studied farmers' workflows. By working directly in the field, I found opportunities to ask questions and observe their daily operations. Through this hands-on experience, I received continuous feedback, identified operational bottlenecks, and discovered how our system could effectively automate their processes.
Collaboration with farmers is the key success factor for agricultural automation.
