For autonomous robots to function in the real world, artificial intelligence is not sufficient; they also need reliable sensors to see in changing conditions, and robust control of a physical robot platform. To achieve such physical intelligence, new research is needed in the cross section between sensor design, machine learning and robot control. Agricultural robotics is a particularly demanding application, with its large seasonal variations, and high diversity in crops and environment. Today, there are a lot of work-intensive agricultural operations that must be done manually by seasonal workers, for instance pruning, weeding and picking. These tasks are very difficult to automate and requires technology that is beyond state-of-the-art in robotics, computer vision and sensors. The goal of this project is to develop new knowledge and methods that will enable robotic solutions for such agricultural tasks. To achieve this, we need new 3D sensors suited for a demanding outdoor environment, more explainable and reliable deep learning methods for 3D data analysis, a new multi-arm robot system and new methods to ensure safety and reliability of the autonomous operation. The project will particularly focus on how to handle uncertainty from sensors, learning-based methods and control in a unified manner. The methods developed in this project will be developed and tested on real crops and agricultural environments from the start, and will also be relevant for other applications of robotics and autonomy in varied and demanding environments.
To push the boundaries of human-like physical intelligence in robots by developing methods for safe and reliable sensing, learning and control of an autonomous multi-arm agri-robot platform – a step towards human-like physical intelligence.
This project will move the research front within field robotics by developing new methods and knowledge in the cross section between 3D sensing, deep learning, control and safety assurance, which will enable robust and efficient interaction with agricultural crops. By working on real use-cases and data from the start, we ensure relevance for the industry and end-users, and are forced to focus on real-world challenges using sensor data and incorporating reliability measures.
The project will do this through developing for new knowledge within the whole cycle from sensing to interaction; 1) Accurate 3D sensing in highly varying outdoor settings for moving platforms. 2) Learning-based 3D analysis for robotic interaction 3) Design and control of a compact multi-arm robotic system on a moving platform, and 4) Safety and reliability of the autonomous operation. In particular there is a need for new knowledge on how to combine uncertainty of sensors, learning-based components and control, to ensure the overall reliability of the system as a whole in an unstructured and changing environment.
We will use an iterative approach with theoretical studies, simulations and proof-of-concept demonstrations and evaluations, and work on real crops and use-cases from the start to ensure relevant data collection and test-cases. Using agri-robot interaction as a proofing ground to develop methods that can be transferred to other applications and domains.