We are facing immense challenges in agriculture. There is a growing demand for nutritious and healthy food. At the same time, negative impact on the environment must be reduced. Autonomous precision agriculture enabled by artificial intelligence can make a large contribution to these challenges. However, we identified two major obstacles for widespread application of AI in agriculture and these are addressed in this proposal.
First, the dynamic environments in agriculture demand robust artificial intelligence. Each field has unique characteristics and these characteristics are highly dynamic. Plants and weeds grow and their appearances are influenced by diseases. External conditions such as daylight and moisture are changing. Although modern AI techniques are able to generalize, agriculture places extreme demands on robustness.
Second, flexible data management strategies and technologies are required to apply AI in the fields. The goal is to use all available data to make the best decisions. Scientific research has shown impressive results. However, the next step is the translation of these findings to applicable technologies.
The SensAI project is the joint effort of a large industrial robotics partner, three SMEs, and two knowledge institutions from Canada and the Netherlands. The goal is to develop AI for environmentally friendly agriculture. To develop more robust AI, several approaches are researched and developed. One approach is lifelong learning where trained networks are continuously updated for maximum performance. Another approach is data management for multimodal sensor fusion.
SensAI delivers value by applying the best machine learning approaches to actual robots. This allows the robots to make optimal decisions. Moreover, the project explores commercializing and sharing robust AI and quality datasets using new online markets and platforms. This makes the developed data and algorithms available for companies from all over the world.