With the growing world population, the need for agri-food products is rapidly increasing. The UN estimates an increase in demand of 70% in the next 30 years. At the same time, agriculture needs to reduce the impact on the environment in terms of, e.g., pollution, energy, greenhouse gasses and water usage. Furthermore, the availability of human labor to perform the physically exhausting and repetitive tasks is reducing. To put it short, agriculture needs to produce more with less and should move toward sustainable and autonomous food production. Agricultural robots are an important solution to achieve this goal. They can perform precision agriculture, providing just the right treatment to the right location at the right time, which can drastically reduce the environmental impact. Robots furthermore can perform the physically challenging and repetitive tasks, providing room for higher-skilled labour in agriculture. However, there are technical challenges to get robots to operate in the complex agri-food environments; (1) there is a lot of variation in the appearance, shape and properties of the biological products, as well as in the environmental properties, such as weather and climate conditions, (2) the environments are cluttered with many objects, causing limited free space and many occlusion, and (3) the plants and produce are delicate and vulnerable.
My research focusses on the development of methods to deal with these challenges in agri-food robotics. With methodologies from artificial intelligence and machine learning, the challenges of variation can be approached. The challenge of clutter and occlusion can be alleviated by moving away from the traditional sense-think-act paradigm in robotics towards integrated sensorimotor control, including active perception. And the interaction with delicate plants and produce requires soft robotic manipulators and end-effectors.