A Modular Dual-Arm Apple Harvesting Robot with Enhanced Field Performance
This paper presents a modular dual-arm apple harvesting robot featuring a vertically stacked design and five key technological advancements in perception, motion control, and coordination, which demonstrated an 80.0% success rate and high fruit quality in commercial field trials.
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
Imagine a busy apple orchard where the biggest problem isn't the trees or the fruit, but the lack of people to pick them. Farmers are running out of hands to harvest their crops. This paper introduces a new robot designed to solve that problem: a modular, dual-arm apple picker that works like a highly organized, two-handed chef in a kitchen, but instead of chopping vegetables, it's picking apples.
Here is a breakdown of how this robot works, using simple analogies:
1. The Setup: A Vertical Tower, Not a Flat Table
Previous robots the team built had two arms sitting side-by-side (like two people standing next to each other). This new version stacks the arms one on top of the other (like a tall bookshelf).
- Why? In the old design, the robot had to move sideways to reach different trees, which took a lot of time. With the vertical stack, the robot stops at one tree, and the top arm picks from the top branches while the bottom arm picks from the bottom branches simultaneously. It's like switching from walking back and forth across a room to standing in one spot and reaching up and down.
2. The Eyes: A "Smart Camera" That Never Gets Confused
Orchards are messy. Leaves block the view, the sun is blinding, and apples are clustered together.
- The Solution: The robot uses a special "foundation model" (a type of super-smart AI) that acts like a detective who can spot an apple even if it's half-hidden behind a leaf or the sun is glaring. It doesn't just guess; it uses a two-step process: first, it finds the general area of the apple, and then it zooms in to draw a perfect outline around it, ignoring the leaves.
- The Shield: To stop the sun from blinding the camera, the robot has a retractable "sunscreen" (like a car's sunshade) that pops up over the workspace before it starts working.
3. The Hands: The "Vacuum and Twist" Technique
The robot doesn't use fingers to grab apples (which could bruise them). Instead, it uses a soft silicone suction cup (like a gentle plunger) and a special twisting motion.
- The Approach: Instead of zooming straight at the apple (which might knock it off if the robot's aim is slightly off), the robot first stops a few inches away. Then, it performs a slow, gentle linear sweep (like a windshield wiper) toward the apple. This creates a "safety corridor" where it can gently touch the fruit without smashing it.
- The Twist: Once the suction cup is attached, the robot doesn't just pull straight back. It twists the apple like unscrewing a jar lid. This breaks the stem cleanly with less force, which is much gentler on the fruit than a hard yank.
4. The Teamwork: Sharing a Single "Vacuum Lung"
The robot has two arms, but they share one single vacuum pump (like two people sharing one pair of lungs).
- The Problem: If both arms try to suck at the same time, neither gets enough power.
- The Solution: The robot uses a strict set of rules (written in "temporal logic," which is like a very precise traffic light system) to manage the vacuum.
- Rule 1: Only one arm can be "sucking" at a time.
- Rule 2: As soon as one arm grabs an apple and starts pulling it back, the other arm is allowed to start its next job immediately. It doesn't wait for the first arm to finish everything; it just waits for the "lung" to be free.
- Result: This coordination is so efficient that the robot works almost as fast as if it had two separate vacuum pumps, saving money and weight.
5. The Safety Net: "No-Go Zones"
To make sure the robot doesn't crash into itself or move too jerkily, it uses a Control Barrier Function. Think of this as an invisible force field. If the robot's computer tries to make a move that would hit a joint limit or cause a violent shake, the safety filter gently nudges the command to make the movement smooth and safe, ensuring the robot never jerks the fruit or breaks its own parts.
6. The Results: How Well Did It Do?
The team tested this robot in real commercial orchards in Washington State during the 2025 harvest season.
- Success Rate: Out of nearly 1,740 attempts to pick an apple, the robot succeeded 80% of the time.
- Speed: It took about 7.5 seconds to pick one apple. This is a significant improvement over their previous models.
- Fruit Quality: The apples were handled very gently. 91.2% of the robot-picked apples were still in the highest possible grade ("Extra Fancy"), meaning they were perfect for the grocery store. The bruising rate was very low (between 2.4% and 4.9%).
What's Still Hard?
The robot isn't perfect yet.
- Heavy Leaves: If an apple is almost completely hidden by thick leaves, the robot sometimes can't find it or can't get a good grip.
- Branches: Sometimes branches get in the way and knock the apple loose before the robot can grab it.
- Human Help: Right now, a human still has to drive the tractor and position the robot. The goal for the future is to make the whole thing drive itself.
In short, this paper describes a robot that is faster, smarter, and gentler than previous versions, proving that machines can successfully harvest delicate fruit in the messy, unpredictable environment of a real orchard.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.