In the context of increasingly high automation demands in industrial manufacturing, the integration of advanced technologies such as 3D cameras and industrial robots has become critically important. The solution of combining 3D cameras with industrial robots for product picking helps optimize the production process, enhance accuracy, and improve operational efficiency. These technologies not only increase productivity but also minimize the risk of errors in manufacturing.

1. Solution Architecture

 

 

 

 

 

 

 

 

 

The system consists of an Intel RealSense 3D camera, a pick-and-place robot, and control software. The robot performs the picking operation, while the software processes and manages the entire workflow.

2. Camera 3D Intel RealSense

Camera 3D Intel Realsense D415

 

The camera provides 3D point cloud data and color images, enabling the robot to accurately identify objects

3. Processing computer

  • Processing and Analyzing 3D Camera Data: The Asus NUC rugged industrial mini PC receives data from sensors such as the 3D camera. It then processes and analyzes this data to help the robot understand its surroundings, determine the position and shape of the object to be picked, and identify other crucial factors such as distance, depth, and the object’s condition.
  • Motion Planning and Path Calculation: The computer uses control algorithms to calculate the robot’s movements, from arm control to selecting the optimal path to pick up the object without collision. These algorithms ensure that the robot can move precisely and flexibly in 3D space, avoiding obstacles and completing tasks efficiently.
  • Managing Machine Learning Algorithms: The computer integrates machine learning algorithms to enhance object recognition and optimize pick-up processes. For example, the robot can learn from previous tasks to improve its object selection, identify the best grasping strategy, or handle previously unseen scenarios.
  • Performance and Time Optimization: The computer helps the robotic picking system optimize task execution time and overall process efficiency. The control and planning algorithms allow the robot to work quickly and accurately, minimizing the time required to perform each task.
  • Interaction and Communication with External Systems: The computer also manages communication between the robot and external systems, such as warehouse management systems, ERP (Enterprise Resource Planning) systems, or other infrastructure. Data exchange and information reception enable the robot to adjust its actions according to external system requirements.

Asus® NUC Rugged Chassis Element

 4. JAKA Zu 7 Robot System

With the robotic arm and 3D camera, we can perform the following tasks:

  • Replace manual labor in stages where the products fed into the system are not yet sorted.
  • Help reduce time and intermediate steps for automatic feeding into the system.
  • Accurately identify products that are mixed up in trays.
  • Integrated camera programs, just install and use, help minimize programming time and processing steps.

Cycle of Object Grabbing Robot Using 3D Camera

  1. Image Acquisition
    The Intel RealSense 3D camera is used to capture the three-dimensional images of the working area.
    The data includes:

    • Color image (RGB)

    • Depth data

    • Point Cloud

    Preprocessing & Feature Extraction

    • Clean up the point cloud data (denoise, downsample)

    • Segment the background and the object to be grasped

    • Compute geometric features (size, centroid, orientation…)

    Object Detection & Pose Estimation

    • Use object detection techniques and algorithms.
      Results:

    • Identify the type of object

    • 3D coordinates and rotation angles (6 DOF pose)

    • Detection time: 1-2 seconds

    • Graspable object size: 4cm x 2cm

    Grasp Planning & Motion Planning

    • Determine the optimal grasping position (grasp point)

    • Use motion planning algorithms (MoveIt, OMPL, RRT, TrajOpt…)

    Execution

    • Send control commands to the robot

    • The robot moves to the grasping position → grabs the object → moves to the target location

    Feedback & Error Handling

    • Force sensors, camera, or tracking system to confirm successful grasp

    • If the grasp fails, repeat the detection and planning steps

    • Cycle time for object grabbing robot: 3-5 seconds

 

Real-World Applications

Solution Deployment Timeframe: 2–3 months

Contact Information

Address: No. 76, Alley 120 Yen Lang, Dong Da District, Hanoi

Hotline: +84 888627888

Website: www.brandpc.vn

Youtube: https://www.youtube.com/@brandpcVN

Fanpage: https://www.facebook.com/BrandPCVietnam