Application · Pick and place
Follow a part from conveyor capture to accepted placement in a fixture. This example explains the working plane, movement tracking, gripper offset and controller outcomes, then defines a capacity calculation that counts each time loss once.

Consider parts arriving on a conveyor and being placed into a fixture with a vacuum gripper. This is an illustrative implementation scenario. The camera must provide enough information for the grip, and the robot must deliver each accepted part within the fixture tolerance.
Record the permitted part position, height and tilt, conveyor movement, accessible gripping surface and destination tolerance. The broader 2D versus 3D comparison covers the technology choice; the implementation here follows one part through the cell.
If the gripping surface stays at a known height and permitted tilt is sufficiently constrained, a 2D image may provide horizontal position and in-plane rotation. Check the smallest locating feature and the allowed movement after capture. The part must remain distinguishable from the belt across normal surface and lighting variation.
If height or tilt changes the contact point or approach direction, test the additional spatial information required. A 3D measurement must be valid on the actual surface and throughout the permitted range. If parts overlap or interlock, locating one part is insufficient: grip accessibility and withdrawal also require evaluation.
For a stopped conveyor, define how the controller confirms the stop and prevents movement during the relevant operation. For a moving conveyor, associate each image with the machine's validated position reference and account for processing and robot delay. A timestamp alone does not establish the distance travelled when speed varies.
Test minimum and maximum speed, acceleration where relevant and the closest permitted part spacing. Verify that the measured part is still the one the robot approaches. Specify what happens when a result arrives after the tracking window or when two detections are ambiguous.
Document camera, conveyor or working-plane, robot-base and tool coordinate frames. State units, axis directions and rotation conventions. The detected reference point must be converted to the actual gripper contact point using the defined part offset and tool calibration.
Validate at independent points near the centre and edges of the working area, including height extremes. Confirm both pickup and placement against their respective tolerances. A pickup that succeeds with a compliant gripper can still place the part outside the fixture requirement.
Return a cycle identifier, valid-result status and the coordinates required by the robot. Define separate outcomes for no part, uncertain detection, unreachable grip, failed pickup and unavailable destination. The machine may retry, hold, divert or request assistance according to the agreed process.
Test a stale result, missing result, disconnected interface and restart. Confirm pickup using the selected gripper's supported method, and test whether a dropped part or double pickup can be detected where those failures matter. Validate approach, clearance and withdrawal with the actual tooling.
Choose a timing boundary and state it in the report. For a representative production window, measure total elapsed seconds T and the number of accepted placements N. Include retries, recovery and other stops that belong within that window.
Time per accepted placement = T ÷ N. Observed hourly output = 3,600 × N ÷ T.
For illustrative arithmetic, 1,800 elapsed seconds and 300 accepted placements give 6 seconds per accepted part and an equivalent 600 accepted parts per hour over that window. These are example inputs, not reported installation results. Do not multiply this observed rate by an availability factor for stops already included in T.
Alternatively, a planning model can use an explicitly defined operating rate multiplied by a separately measured availability fraction. State which losses belong to each term and count each once. Do not mix operating-only cycle time with an all-inclusive denominator. If there are no accepted placements, report the failed run rather than dividing by zero.
Agree the part mix, presentation extremes, placement tolerance, allowed intervention frequency and accepted-output requirement. Keep counts of distinct parts, attempts, accepted placements and failure types. Test product changeover and restoration after replacing or moving a camera.
Compare measured manual loading hours, ergonomic demands and recovery work with equipment, gripper, mounting, integration and validation hours multiplied by their rates. Include maintenance and configuration changes. Value additional output only where demand and downstream capacity make it usable; avoid counting the same recovered output again as a separate downtime saving.
Bring representative parts, conveyor motion data, fixture drawings, gripper information and required output to a feasibility discussion. The first trial should resolve the uncertain grip and tracking conditions before final equipment selection.
Feasibility
Share representative parts, extreme presentations, required cycle time and placement tolerance. A trial shows whether 2D is sufficient or which 3D information is required.