Analysis of the impact of damage rate on the performance of orange fruit harvesting robot

Abstract

• In-depth analysis of damage rates in fruit harvesting robots, shedding light on a previously underexamined aspect of agricultural automation. • Detailed exploration of the impact of fruit orientation, occlusion, and varying illumination on damage rates, offering new insights into these critical factors during robotic harvesting. • Identification and correlation of specific damage causes with distinct bruise types, such as skin peel and bruising, providing a clearer understanding of injury mechanisms. • Strategic recommendations for optimizing robotic harvesting techniques, aimed at increasing efficiency and reducing fruit damage, thereby significantly enhancing the commercial viability of automated harvesting systems. Reducing damage rates is paramount for optimizing the efficiency of fruit harvesting robots and advancing their journey towards commercial viability. Despite the crucial role that damage rates play in determining fruit quality and marketability, there is a notable lack of comprehensive and in-depth studies analyzing this aspect, especially within the context of fruit harvesting robots. Most research tends to prioritize metrics such as success rate and accuracy of fruit picking, leaving the examination of damage rates relatively overlooked. This study fills this gap by conducting a thorough examination of the factors contributing to damage rates in fruit harvesting robots, including the causes of damage, the types and sizes of bruises incurred, and the impact of occlusion, illumination conditions, and end effector orientation. Additionally, the research investigates strategies for minimizing damage rates, offering insights into optimizing fruit harvesting techniques to reduce potential damage. Occlusion, illumination, and gripper angle were found to significantly influence fruit damage. Specifically, a 10 % increase in occlusion raised damage by 1.18 %, a 100 Lumen/m 2 increase in illumination reduced damage by 10.5 %, and deviation from the optimal 90° gripper angle increased damage by 1.8 % per 10° shift. Overall, proper fruit orientation reduced damage by 40 %, minimal occlusion by 36 %, and optimal illumination by 25 %. A multiple linear regression model explained the variance in damage rate (R2 = 0.924) and achieved a low RMSE of 1.85 %, demonstrating high predictive accuracy and validating the model’s reliability in quantifying the influence of harvesting parameters. By investigating these aspects and exploring strategies for minimizing damage, the study aims to advance fruit harvesting robotics and contribute to the successful commercialization of this technology.