Generates 3D point clouds without CAD or image pre-registration, enabling robots to instantly recognize and grasp unknown objects.
.webp)
Process
- Automatic detection of object position, orientation, and gripping method
- Real-time, collision-free robot path generation
- Flexible gripping with Bridgestone Soft Robotics’ “TETOTE and” gripper (suction or four-finger grasp)
Challenge
- Wide variation in product size, type, and shape
- Managing a large variety of SKUs typical in e-commerce fulfillment
- Difficulty handling glossy, thin, or semi-transparent objects using conventional methods
Solution
- Masterless Picking: Eureka’s 3D processing pipeline enables object picking without prior data registration or training.
- Robust Recognition: The AI-based stereo 3D camera operates without pattern projection, allowing accurate recognition of glossy, semi-transparent, and thin objects with minimal sensitivity to reflections and lighting.
- High-Precision Calibration: Integrated 3D camera and robot control simplify calibration and facilitate flexible robot/end-effector selection.
- Optimized Motion Planning: AI-generated grasping poses combined with advanced motion planning and collision avoidance ensure robust, reliable, and high-speed picking.
Outcomes
- Demonstrated the feasibility of scalable piece-picking automation for e-commerce fulfillment.
- Enabled the robotic hand to adaptively choose suction or fingers based on item characteristics.
- Achieved high reliability and speed in picking a wide variety of SKUs, expanding the scope of automation in logistics.