FASTPICK: Novel active vision and picking head to robotically harvest soft fruit

季节性移民劳动力的不确定性使英国新鲜农产品和软水果行业面临前所未有的压力。英国退欧和COVID-19的直接影响已经并正在限制每年前往英国收获超过10亿英镑新鲜水果和蔬菜的69，000名季节性移民工人的供应。机器人和自动化技术提供了一个永久性的解决方案，可以将该行业与其劳动力依赖性断开，同时也为英国机器人行业创造了高技能就业机会和增长。然而，关键的挑战仍然是开发机器人技术，可以采摘密集，闭塞和生物可变集群内的水果。领先的机器人水果采摘技术可以在4秒内采摘80%的草莓。FASTPICK将开发集成到新型机器人采摘头和私有5G网络的主动视觉系统，旨在以最快速度采摘95%的水果。2秒一个浆果，和人类收割机的性能一样。这一性能消除了农业机器人系统在软水果行业大规模采用的最后技术障碍。FASTPICK将开发一种最先进的主动和动态视觉系统，该系统使用多个摄像头，包括采摘头中的视觉伺服，以创建复杂集群的3D场景，并识别抓取器的关键采摘点。FASTPICK将在Gazebo数字孪生环境中实施，可用于优化拣选控制，并作为未来机器人开发的关键工具。高速图像处理将通过私有5G网络将系统集成到云和移动的边缘计算中进行优化。一个优化的采摘头和主动视觉系统将被集成到由佐贺与林肯大学合作开发的完全自主的Thorvald机器人平台上。它将在草莓半商业作物上进行测试和示范。该合作由佐贺机器人有限公司与林肯大学（UoL）和领先的机器人系统开发商剑桥顾问公司合作领导。采摘解决方案将与Berry Garden Growers共同创建，其合作社成员生产超过45%的英国软水果。共同创造实现了有效和负责任的创新，同时也为佐贺的重要、快速和规模化的市场路线奠定了基础。该技术将面向英国和全球新鲜农产品行业销售，但二级市场存在于多个机器人应用领域。

Uncertainty over access to seasonal migrant labour is placing the otherwise vibrant UK fresh produce and soft fruit sector under unprecedented pressure. The immediate impacts of Brexit and COVID-19 have and are restricting availability to the 69,000 seasonal migrant workers who travel to the UK each year to harvest over £1bn of fresh fruit and vegetables. Robotics and automation technology offers a permanent solution that can disconnect the sector from it's labour dependency, whilst also creating high skilled jobs and growth for the UK robotics sector. However, critical challenges remain to develop robotic technology that can pick fruit within dense, occluded and biologically variable clusters. Leading robotic fruit picking technology can pick 80% of strawberries at 4 seconds per berry. FASTPICK will develop active vision systems integrated to a novel robotic picking head and private 5G network that aims to pick 95% of fruit at c. 2 seconds per berry, the same performance of human harvesters. This performance removes the final technical barrier to large scale adoption of agri-robotic systems for the soft fruit sector. FASTPICK will develop a state of the art active and dynamic vision system that uses multiple cameras, including visual servoing in the picking head to create a 3D scene of complex clusters and identify critical picking points for the gripper. FASTPICK will be implemented in a Gazebo digital twin environment that can be used to optimise picking control and as a key tool for future robotic development. High speed image processing will be optimised by integrating the system into cloud and mobile edge compute via a private 5G network. An optimised picking head and active vision system will be integrated onto the fully autonomous Thorvald robotic platform developed by Saga in collaboration with the University of Lincoln. It will be tested and demonstrated on semi-commercial crops of strawberries. The collaboration is led by Saga Robotics Ltd in collaboration with the University of LIncoln (UoL) and leading robotic system developers Cambridge Consultants. The picking solution will be co-created with Berry Garden Growers whose cooperative members produce over 45% of the UK's soft fruit. Co-creation enables effective and responsible innovation whilst also underpinning significant, rapid and scaled routes to market for Saga. The technology will be marketed to the UK and global fresh produce sectors but secondary markets exist across multiple robotic application domains.