CLASP: A Cluster-Level Autonomous Selective Picking Robot with a Soft Rolling-Band Gripper for Fresh-Market Blueberry Harvesting
This paper introduces CLASP, a cost-effective autonomous robot featuring a soft rolling-band gripper and a global-to-local perception system that enables efficient, selective harvesting of fresh-market blueberries at the cluster level by gently detaching only mature fruit while preserving immature berries.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the quiet rows of a blueberry farm, a paradox plays out every summer. The fruit destined for the fresh market—plump, unblemished berries meant for the grocery store shelf—must be picked by hand. Machines exist that can shake bushes and catch the falling fruit, but they are too rough; they bruise the delicate skin and strip away berries that are not yet ripe, leaving only a mushy mix suitable for jam or processing. The problem is that blueberries do not ripen all at once. On a single small cluster, a deep blue, ready-to-eat berry often sits right next to a hard, green one. To harvest the ripe fruit without damaging the rest, a picker must gently pull the mature berry off its stem while leaving the immature one behind. This requires a level of touch and judgment that has, until now, been impossible to automate, forcing farmers to rely on a shrinking workforce of seasonal laborers.
A team of researchers has built a robot designed to solve this specific puzzle. They call their creation CLASP, a machine that does not try to grab individual berries one by one, but instead treats the entire cluster as a single unit. The robot's "hand" is a soft, rolling band made of silicone, inspired by the way human pickers use their thumbs to rub berries against the stem. When the robot finds a cluster, it wraps these soft bands around the fruit and begins to roll them. As the bands turn, they create a gentle friction that pulls the ripe berries off their stems one after another, while the unripe ones, which are still firmly attached, stay put. The robot does not need a force sensor to know how hard it is pulling; instead, it listens to the electrical current in its motors. When the motor works harder because a berry is resisting, the robot knows it is applying force, and it adjusts its grip to stay within a safe window—strong enough to detach the ripe fruit, but weak enough to leave the green ones alone.
The system relies on a two-step vision process to find its targets. First, a camera mounted on the robot's body scans the field to spot clusters of fruit from a distance. Once a potential target is found, the robot moves its arm closer, bringing a second, handheld camera right up to the fruit to get a detailed look. This close-up view helps the robot understand the shape and orientation of the cluster, allowing it to position its soft bands correctly. The robot then executes a precise sequence: it moves its arm to the cluster, closes the bands, and begins the rolling motion. It continues this process, periodically checking to see if any ripe berries remain, until the cluster is empty of mature fruit.
In field tests conducted across several blueberry farms, the robot demonstrated that this approach works. Out of twenty-five clusters presented to the system, the robot successfully grasped and harvested twenty-three, achieving a ninety-two percent success rate. The failures were not due to the picking mechanism itself, but rather because the robot could not physically reach the cluster due to the dense leaves and branches blocking its path. When the robot did reach the fruit, it was gentle. The berries it harvested were only slightly softer than those picked by human hands, a difference the researchers attribute partly to the time delay between picking and measuring, rather than damage from the robot. The cost of building one of these units is approximately three thousand three hundred twenty-six dollars, a price point that suggests the technology could be scaled for real-world use.
The researchers also measured the force required to pull a berry off its stem and found a clear difference between ripe and unripe fruit. Ripe berries detached at a much lower force than green ones, creating a "window" where the robot could apply enough pressure to harvest the ripe fruit without disturbing the rest. By carefully controlling the motor current to stay within this window, the robot proved it could selectively harvest at the cluster level, a task that previous machines had failed to do. While the robot is not yet as fast as a human picker, it moves at a pace that is promising for a machine, handling about thirty-two berries per minute compared to a human's fifty-five. The study suggests that the bottleneck is no longer the ability to pick the fruit gently, but rather the speed of the robot's movement and its ability to navigate through the thick foliage of the farm. This work marks a significant step toward automating the most labor-intensive part of fresh-market blueberry production, offering a way to bring the delicate touch of a human hand to the machine.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.