RoboCrop: Teaching robots how to pick tomatoes (phys.org)

Osaka Metropolitan University researcher Takuya Fujinaga has developed an innovative approach to training robots for tomato harvesting, addressing critical labor shortages in agriculture. Unlike traditional models that focus solely on fruit recognition, Fujinaga’s method employs a “harvest-ease estimation” framework, allowing robots to evaluate the feasibility of picking each tomato based on various factors, such as fruit clustering and occlusion. This shift enables robotic systems to not only recognize ripe tomatoes but also to determine the most effective approach for harvesting them. The results of this research, published in Smart Agricultural Technology, demonstrate an impressive 81% success rate in robotic harvesting attempts. This model significantly improves the potential for automated agricultural solutions by allowing robots to adapt their strategies—shifting their approach direction to successfully harvest fruits that previous attempts missed. Fujinaga’s work paves the way for smarter agricultural robots that could redefine farming practices, with robots and humans working together: robots will efficiently harvest easily accessible crops while humans manage more complex tasks, potentially revolutionizing efficiency and productivity in the agricultural sector.