Create a Task and Basic Configuration
About 2229 wordsAbout 7 min
3. Create a Task and Basic Configuration
After completing the Scene Description and Pre-configuration Checks, create the task for neatly arranged cylinder loading and unloading, and configure the camera, robot, end-effector, eye-hand calibration, and ROI. This chapter establishes the basic configuration. The templates, keypoints, visual model, and Pick Points required for the target will be covered in the next chapter.
3.1 Create a Task for Loading and Unloading Neatly Arranged General Type Target Objects
Depending on the project status, create a new project or add a task to an existing project. In either case, select Neatly Arranged Loading and Unloading > Loading and unloading of neatly arranged general type target objects.
Task Type
The task type cannot be changed after creation. Do not select “Loading and unloading of Cylinder type target objects” solely because the target is called a cylinder. This case uses the model recognition and pose estimation workflow for general type target objects.
3.1.1 Create a New Project
- Open PickWiz, go to the new project page, and select New Empty Project.
- Select Neatly Arranged Loading and Unloading > Loading and unloading of neatly arranged general type target objects.
- Set the project name and storage path. The project name cannot contain Chinese characters or spaces. You can use
cylinder01as an example name. - When setting up the project for the first time, leave additional feature options disabled. First verify the basic recognition, pose calculation, and Pick Point generation workflow.
- Click Create. After the project is created, the software provides the first task by default.
For the full procedure, see Create a New Empty Project.
3.1.2 Add a Task to an Existing Project
- Click New Task in the upper-left corner of the main interface.
- Select Neatly Arranged Loading and Unloading > Loading and unloading of neatly arranged general type target objects.
- Enter a distinctive task name, such as “Cylinder Loading and Unloading.”
- Click Create and confirm that the new task is now active.
Perform all subsequent steps in this task. For instructions on creating and switching between multiple tasks, see the Task Operation Guide.
3.2 Select the Camera and Robot Configurations
- In the task information, select a camera configuration that is connected and has been tuned.
- Trigger an image capture. Confirm that the entire cylinder infeed area is visible, the 2D image is clear, and the 3D Point Cloud retains features sufficient to distinguish position and orientation.
- Select the robot configuration that corresponds to the actual equipment, and check the communication method and connection role.
- Check the robot’s number of axes, rotation expression, and coordinate system type to ensure they match the settings on the robot side.
At this stage, confirm the acquisition and communication conditions. Do not send unverified Pick Point coordinates to the robot to execute a movement. For detailed procedures, see the Camera Connection and Tuning Guide and Robot Configuration Guide.
3.2.1 Tune for Cylinder Reflections and Orientation Features
- Create an editable configuration from a camera template suited to a similar surface, and save the initial parameters and a set of acquired data. Keep the target, lighting, and viewing angle fixed so that scene changes are not mistaken for the effects of tuning.
- Inspect the 2D Image: there should be a clear boundary between the cylinder and the pallet. The end face, holes, or other actual features used to distinguish orientation must remain visible despite bright reflections. Enable the overexposure indicator to help inspect bright areas.
- Adjust the exposure gradually, triggering an image capture each time and comparing the same cylinder. Increase exposure appropriately if the image is too dark. If reflections obscure the outline, reduce exposure and check the camera mounting angle and site lighting.
- Inspect the 3D Point Cloud at the same time: check for missing data on reflective surfaces, stray points at edges, and the stability of steps and orientation features. A suitable-looking 2D image does not necessarily mean the 3D data can be used for pose calculation.
- If one exposure setting cannot accommodate both bright and dark areas, evaluate high dynamic range exposure according to the guide for the camera in use. Confirm that feature coverage improves and compare acquisition times. Do not directly apply engine settings or exposure values from other camera models.
- Once the Point Cloud is largely complete, add denoising settings one at a time and compare the results before and after filtering. Do not filter out structures needed to distinguish the cylinder’s orientation.
- Capture repeat images with a full arrangement, missing targets, edge positions, and permitted tilt. If using automatic exposure, also check whether cylinder brightness remains stable as the exposed proportion of the pallet changes.
- Save the configuration and select it in the task. After configuring the templates and model, retest 2D recognition, pose, and Pick Points using the same samples. Do not conclude tuning solely because the image is brighter or a single recognition attempt succeeds.
It is recommended to record “changed setting—2D outline—3D features—acquisition time—recognition and pose results,” changing only one setting at a time. For interface and result examples for the camera model, see the Camera Connection and Tuning Guide.
3.3 Configure the End-effector and Eye-hand Calibration
3.3.1 Configure the End-effector
- On the End-effector page, add or select a configuration corresponding to the actual gripper.
- Check the position and orientation of the tool TCP. If using a gripper model, check the coordinate relationship between the model and the robot TCP.
- Select this end-effector configuration in the task information.
The tool configuration must match its actual installed state. The Pick Point location on the cylinder and the approach direction will be set with reference to the target coordinate system in the next chapter. A tool model that appears correctly displayed is not, by itself, proof that the picking pose is correct.
For the procedure, see the End-effector Configuration Guide.
3.3.2 Select the Eye-hand Calibration Result
- Complete the appropriate eye-hand calibration according to the camera mounting method and the robot’s number of axes.
- Check the calibration result and confirm that it meets the site’s positioning accuracy requirements.
- In the task information, select the calibration configuration corresponding to the current camera, robot, and installed state.
Do not directly reuse calibration results from other equipment or a different installed state. For the calibration method, see the Eye-hand Calibration Procedure. For the checking method, see Calibration Result Check and Analysis.
3.4 Set the ROI for the Cylinder Infeed Area
The ROI limits the effective area for recognition and pose calculation. It must cover actual variations in incoming targets while reducing interference from the pallet, supports, and adjacent areas.
3.4.1 Add an ROI and Load Data
- On the ROI page, click Add ROI and set a distinctive name.
- Use Capture Image to acquire the current scene, or import historical images and Point Clouds from the same acquisition.
- Confirm that the cylinder arrangement in the image matches the Point Cloud before setting the region.
Historical Data
Historical data captured in ultra-fast mode may not contain a Point Cloud file. If a Point Cloud needs to be imported, first generate Point Cloud data in the historical data, then return to the ROI page to import it. Do not mix images and Point Clouds from different acquisitions.
3.4.2 Set the ROI 2D Workspace
- In the 2D image, draw a box around the full range of incoming cylinder positions, rather than just one cylinder currently being tested.
- Leave margins for row and column position deviations, target tilt, and infeed offsets. Avoid cutting off the cylinder outline or local features used to identify its orientation at the region boundary.
- If there are multiple layers or height variations, check coverage at the permitted infeed position closest to the camera.
- While covering the range above, exclude as much surrounding material and environmental area as possible if it is not part of this task.
- When using a KINGFISHER stereo camera, set the region in the left-eye image and check that the automatically synchronized right-eye ROI also covers the target.
3.4.3 Set the ROI 3D Workspace
- Inspect the Point Cloud from multiple viewpoints and use the infeed plane, such as the pallet or conveyor-belt plane, as the reference for adjustment.
- Adjust the ROI orientation so that the positive Z direction points upward. If the infeed plane is tilted, make the Z axis perpendicular to that plane and point upward, and align the X/Y directions with the incoming target arrangement as closely as possible.
- Adjust the ROI size and position to cover all permitted cylinder positions and height variations, leaving the necessary margins.
- If the pallet is not used for recognition, the bottom of the ROI can be moved above the pallet, but this must not cut off valid Point Cloud data needed for cylinder pose matching.
- Exclude non-target areas such as supports, guardrails, and adjacent workstations as much as possible. Recheck the boundaries from both top and side views.
The ROI 3D represents a range in real space, so there is no need to account for perspective size changes as with an image region. Do not address recognition failures by expanding the ROI without checking its bounds. First check whether the target has been cropped, then determine whether other visual parameters need adjustment.
3.4.4 Set a Grasp Region as Needed
If the full range for cylinder recognition is large but the process permits Pick Points only within a smaller area, you can also set an ROI 3D Grasp Region and configure the corresponding Pick Point filtering later.
The workspace retains the data needed for recognition and pose calculation; the grasp region limits where output Pick Points are permitted. Do not crop cylinder features from the workspace merely to narrow the range of Pick Points.
ROI and Motion Safety
Setting an ROI or grasp region does not replace checks of the robot motion path, reachability, and collision risks. Before the first coordinated operation, verify the Pick Point coordinates, tool approach direction, and withdrawal path.
When finished, save the ROI and select this configuration in the task information. For interface instructions, see the ROI Configuration Guide.
3.5 Check and Save the Basic Configuration
Before leaving this chapter, confirm the following:
- The current task type is “Loading and unloading of neatly arranged general type target objects”; no other scene was selected by mistake.
- The current camera configuration can reliably acquire images and Point Clouds that meet the requirements.
- The corresponding robot, end-effector, and checked eye-hand calibration configurations have been selected in the task information. If calibration is not yet complete, do not proceed to coordinated operation with the robot.
- The ROI has been saved and linked to the current task. Edge positions and permitted height variations are within the effective range.
- The basic configuration has been saved. Target configuration that has not yet been completed will be handled in the next chapter.
The task is not yet ready for full operation. The next chapter will cover creating and aligning the Point Cloud template and configuring keypoints, the Visual Model, and Pick Points of the Target Object. After completing all required configurations, verify visual operation and coordinated operation with the robot.
Return to Case Overview · Previous Chapter: Scene Description and Pre-configuration Checks · Next Chapter: Point Cloud Template and Target Configuration