The field performance test verifies that the average harvesting cycle of the robot with both arms reached 9 s/bunch, and the success rate of bunch identification and harvesting success rate reached 88 and 83%, respectively, which were significantly better than those of existing harvesting robots worldwide. {A model already working in the market}
A NVidia Jetson Nano developer kit is used as the center of the decision system. Its small size and powerful computing power meet the needs of running programs in harvesting (GPU: 128-core NVIDIA Maxwall, CPU: Quad-CoreARM Cortex-A57 MPCore). RealSense D435i depth camera is used as the main sensor to obtain environmental information (RGB images 1,920 × 1,080 in resolution, depth images 1,280 × 720 in resolution, with a FOV of 69° × 42°). It is manufactured by Intel, United States. It is able to cover a wider area and reduce more blind spots. The robotic arms use Techsoft TB6-R5 (Techsoft, Shenzhen, China). TB6-R5 has a payload of 5 kg and repeatable position accuracy of up to ±0.05 mm. Each robotic arm has its own controllers, and controllers receive their respective harvesting tasks and control independently. Each arm is equipped with a cut-clip end effector. They hold grape clusters while cutting grape stems. The chassis is manufactured by Sangpu Agricultural Machinery Co., Changzhou, China. This robot uses a 2D radar (LMS-111, Sick, Germany) with TOF distance detection and enables accurate measurements in a complex field environment.
| Depth | Depth technology: Stereoscopic Minimum depth distance (Min‑Z) at max resolution: ~28 cm Depth Accuracy: <2% at 2 m1 | **Depth Field of View (FOV):87° × 58°Depth output resolution:Up to 1280 × 720Depth frame rate:**Up to 90 fps |
|---|
| RGB | RGB frame resolution: 1920 × 1080RGB frame rate: 30 fps RGB sensor technology: Rolling Shutter | **RGB sensor FOV (H × V):**69° × 42° RGB sensor resolution: 2 MP |
|---|
Camera that is to be used for rover movement:
Raspberry Pi Camera Module V2-8 Megapixel,1080p (RPI-CAM-V2)