AI-powered tomato-picking robot learns to think before acting

A tomato picking robot developed by researchers at Osaka Metropolitan University evaluates the difficulty of harvesting each piece of fruit before attempting to pick it. Assistant Professor Takuya Fujinaga of the university's Graduate School of Engineering created a system that combines image recognition with statistical analysis. The robot examines visual details including the tomato, its stems, and any obstructions from leaves or other plant parts to select the best approach. This method, described as harvest ease estimation, shifts focus from mere detection to predicting the likelihood of a successful pick. In testing the system reached an 81 percent success rate. Approximately one quarter of the successful harvests occurred after the robot adjusted its angle following an unsuccessful front facing attempt. Fujinaga noted that the research establishes ease of harvesting as a quantitatively evaluable metric. The work brings agricultural robots closer to making informed decisions and acting intelligently, according to the researcher. The findings were published in the journal Smart Agricultural Technology. Fujinaga envisions a new form of agriculture where robots and humans collaborate. Robots will automatically harvest tomatoes that are easy to pick while humans will handle the more challenging fruits.