Skip to main content

Manipulation

This section focuses on practical robot-arm workflows, showing how perception, pose estimation, motion planning, and control are connected to let a robot interact with objects.

While Robotics explains how a robot senses, perceives, plans, and acts, Manipulation applies those concepts to tasks performed by a manipulator, such as reaching, picking, placing, and stacking objects.

Sensing, perception, planning, and action in a manipulation workflow

Manipulation Workflow​

A typical vision-based manipulation workflow contains the following stages:

  • Object Perception
    Detect the target object from RGB or RGB-D camera data.

  • Pose Estimation
    Estimate the object's 3D position and orientation in the camera frame.
    Transform the detected pose into the robot's planning frame, typically the robot base frame.

  • Motion Planning
    Generate a valid robot-arm trajectory to the target using a planning framework such as MoveIt.

  • Arm Action
    Send the planned trajectory to the robot controller, monitor execution, and coordinate the gripper when required.

What You Will Learn​

  • How camera perception is connected to robot-arm applications
  • How RGB, depth, and camera calibration are used for 3D position estimation
  • Why camera-to-robot calibration and TF are required in a complete system
  • How object poses become motion-planning targets
  • How URDF, joint states, TF, MoveIt, and robot controllers work together

Hands-on Samples​

Manipulation topics are paired with teaching-oriented examples and recorded ROS bags, allowing you to:

  • Observe perception results and robot-arm motion without connecting a physical camera or manipulator
  • Understand the relationship between object detection, pose estimation, motion planning, and robot action
  • Study repeatable workflows before adapting them to your own camera, robot model, and controller

Manipulation bridges the gap between robot perception and physical interaction with the environment.