The FlexCRAFT program develops, among other things, new robot technology for the automatic harvesting of tomatoes, the processing and packaging of chicken products and the neat packaging of bags of chips and packets of biscuits in boxes of different sizes. “We are developing generic skills for robots with which they can perform actions on products in the agri-food sector that are variable in shape, size and hardness.
The objective of my research is to enable robots to accurately perceive their surroundings in complex agro-food environments, by overcoming the challenges of variation in object and environment properties and incomplete information due to occlusions. These challenges will be addressed by adopting an active perception paradigm, which suggests that by taking actions, robots can gain access to new information which is otherwise unavailable. By gathering more information using systematically planned actions, the uncertainty in perception can be minimised.
Under active perception, next-best-view planning methods will be explored, which aim to maximize the amount of novel information gained from the scene by moving the camera to new viewpoints. Under interactive perception, in-hand manipulation of objects or pushing of occlusions will be investigated to maximize information gain and minimize perception uncertainty.
Finally, the developed methods will be applied to the usecases of greenhouse harvesting and poultry processing to determine their effectiveness in real-world applications. We believe that this research will significantly improve the quality of perception in agro-food robotics and will pave way for improved automation and agro-food production.