Answer from AI Robotics Australasia. .
The sensors you will actually be offered
Structured light projects a pattern and reads how it deforms. Stereo uses two cameras and the difference between them. Time of flight measures how long light takes to return. Each family has parts it likes and parts it hates. Shiny, transparent, and very dark surfaces are the usual argument.
Resolution and field of view have to match the bin or the pallet, not a lab sample. A sensor that sees a single part beautifully may not see a full stillage well enough to plan a grasp at the bottom.
What the software has to add
A cloud is not a pick. Something has to segment the part you want, choose a grasp the gripper can hold, and check that the approach does not collide. Then the robot moves, the scene changes, and the loop runs again. Cycle time lives in that loop.
Calibration between the sensor and the robot is not optional. A camera that is bumped and not rechecked will miss systematically, and the gripper will be blamed.
When to specify it
Specify 3D when the part pose is not controlled: jumbled bins, slumped stacks, or mixed layers. Do not specify it for a flat, fixtured, well-lit pick that 2D already solves.
The decision belongs in a trial with your parts, your finish, and your container. We would rather change the presentation than sell a sensor that cannot see the real bin.
Related questions
What is a point cloud in robotics?
A point cloud is the set of 3D points the sensor measured on the surfaces it could see. Grasp planning uses that shape to decide where a gripper can close without a collision.
Structured light or time of flight?
Structured light is often chosen when you need a denser cloud at shorter range, such as a bin. Time of flight is often chosen for larger scenes. The right answer is whichever returns a usable cloud on your actual part, proved in a trial.
Why can't a normal camera guide a bin-picking robot?
A single camera does not measure height. In a jumbled bin the part's height and tilt are the difference between a grasp and a collision. That requires a 3D sensor, or a mechanical presentation that removes the unknown.
Does 3D vision work on shiny metal?
Sometimes, with the right sensor, optics, and exposure. Sometimes it does not, and the part needs a different presentation. Treat shine as a trial item, not as a footnote.
Related systems
- AI Bin PickingRandom bin picking with 3D vision and spatial AI. Handles mixed SKUs, irregular shapes, and cluttered bins without fixed fixtures.
- AI Quality InspectionInline visual inspection with deep learning defect detection. Catch surface flaws, dimensional errors, and assembly defects at line speed.