A dual-arm robotic platform built by NeuraPick Robotics that sees, reaches, and picks ripe strawberries at commercial speed — fully autonomous, row by row, all day long.
A differential-drive mobile base, dual precision manipulators, and an NVIDIA-first perception stack — engineered to navigate greenhouse rows and harvest at the pace growers need.
Two independent robotic arms work in coordinated tandem, each equipped with a Da Vinci-inspired cable-driven end effector. A precision jaw grasps the fruit while an integrated blade makes a clean calyx cut — gentle enough to preserve shelf life, fast enough to hit commercial throughput.
YOLO-based real-time detection identifies ripe berries by color, size, and occlusion state. OAK 4D stereo depth cameras deliver millimeter-accurate 3D localization for each target — enabling precise reach planning even in dense canopy conditions.
ROS 2-native navigation stack with differential-drive kinematics. EVE-01 travels at 0.1 m/s during active harvesting, autonomously transitioning between rows and returning to collection points. RadioMaster ELRS RC provides field-ready manual override.
Every inference cycle runs on-board. No cloud dependency, no latency penalty, no connectivity risk. Designed for the reality of greenhouse operations.
NVIDIA Jetson-powered compute delivers real-time inference for detection, depth mapping, and motion planning — all at the edge with zero cloud dependency.
OAK 4D stereo cameras combine RGB and depth in a single pipeline, feeding fruit coordinates directly to the arm motion planner at high frame rates.
Dual servo DC drive wheels with 1:50 traction gearing deliver the torque and control needed for smooth greenhouse-floor navigation at harvesting speeds.
Inspired by the Da Vinci surgical system. Cable-driven jaw provides compliant fruit grasping while an integrated blade achieves clean, consistent calyx cuts.
A dedicated 400W servo motor with 1:70 gear ratio enables tight-radius maneuvering through narrow greenhouse aisles with encoder-precise positioning.
RadioMaster TX16S Mark II with ELRS provides long-range manual override directly publishing to ROS 2 cmd_vel — ready for any field scenario.
A tightly integrated pipeline where every stage runs locally on EVE-01 — no round trips, no cloud queues, no single points of failure.
OAK 4D cameras capture RGB + depth at high frame rate, fused into a unified point cloud
YOLO model classifies ripeness, estimates berry pose, and ranks pick priority in real time
Motion planner generates collision-free arm trajectories coordinated with base movement
Cable-driven gripper grasps and blade cuts in a single fluid motion, depositing into tray
We're partnering with forward-thinking growers for field trials. Let's talk about what autonomous harvesting can do for your operation.