Website Wageningen University & Research

Are you that enthusiastic researcher that likes to work on cutting-edge robotic technologies and apply them to improve processes in agriculture and food production?
Wageningen University is coordinating a large national program, FlexCRAFT: Cognitive Robots for Flexible Agri-Food Technology. Together with four other universities in the Netherlands, we aim to develop core-robotic technologies to allow robots to operate in agri-food environments.
The research challenge
Agri-food environments are very challenging for current robotic systems due to the high level of variation in the appearance and shape of natural objects, the uncontrolled illumination, and the complexity of the environments. Consider, for instance, a robot that needs to harvest tomatoes in a greenhouse. All tomatoes vary in shape and color, the illumination differs depending on the position of the sun and clouds, there are many plants closely packed together, and the fruits might be hard to find due to occlusions by leaves or other plants and fruits. Perception and action are very challenging and current robotic solutions do not meet the requirements of the industry.
The FlexCRAFT program will improve the state-of-the-art by developing new methods for (1) active perception, (2) probabilistic world modelling, (3) learning, planning and control, and (4) gripping and manipulation. These methods will be tested and evaluated in three use-cases: (a) harvesting, (b) food-processing, and (c) food packaging.
More information can be found on:
The two PhD positions
PhD1: This position focusses on the development of novel methods for active perception. By allowing a robot to move around in the environment to observe the scene from different viewpoints, the robot can build a more complete and more accurate representation of its environment. This allows a harvesting robot, for instance, to deal with occlusions and observe the tomato that is covered by a leaf. Moreover, the robot can use its arms to push the leaf aside. Your task is to develop methods to optimally plan actions in order to gain as much relevant information about the environment as possible.
PhD2: Many current robots are still based on the sense-think-act paradigm. This paradigm ignores the fact that sensing and action are tightly coupled and that an action immediately leads to new sensations. In this position, you will develop a novel predict-act-sense-compare paradigm, which is based on the skill to predict the consequences of an action. Based on the current estimation of the world state, the robot needs to learn to predict what will happen when it performs an action, a so-called forward model. This prediction can then be compared to the actual observation after performing the action. This allows the robot to continuously learn and improve the forward model. The forward model can then be used to improve the efficiency of perception, by using the sensory predictions and to plan future actions to perform tasks.

To apply for this job please visit