Model-based assessment of robotic fruit harvesting cycle times

Abstract

<abstract> <bold>Abstract. </bold>In this paper fruit harvesting by robotic manipulators on a platform was simulated in order to predict fruit picking efficiency (percent of fruits reachable) and cycle times (fruits per second). Two different manipulators (Puma 560, ABB S4 2.8 robots) mounted on a mobile platform were used. The manipulators had to place the collected fruit in a bin. The kinematic models of these robots were used to assess the fruit picking efficiency. The dynamics of the manipulators were introduced into the simulation along with the kinematics to assess the fruit picking cycle times. The simulations used georeferenced fruit location data for twenty-eight high-density pear trees obtained last summer from commercial orchards. The manipulators used in this simulation achieved both 100% picking efficiencies; however their cycle times were different. The Puma 560 is faster because it has greater maximum joint accelerations than the ABB S4 2.8 robot. From the results it was also clear that 40-65% of the time was consumed in transporting the fruit to the bin.