perception module clutering testing

[IsMoNgai]

perception

This module includes techniques of filtering pointclouds, clustering samples, and localization of arm in physical environment.

The command below should create two gui one for snap shot and one for filtering setting:

ros2 launch interbotix_xsarm_perception xsarm_perception.launch.py robot_model:=rx150 use_pointcloud_tuner_gui:=true use_armtag_tuner_gui:=true

If there is issues with the orientation showing on rViz it could be the aprilTag not being setup facing to the camera in the right way.

One can disable the power of the arm to orient the arm in the right place for snap pose:

ros2 service call /rx150/torque_enable interbotix_xs_msgs/srv/TorqueEnable "{cmd_type: 'group', name: 'arm', enable: false}"

ros2 service call /rx150/torque_enable interbotix_xs_msgs/srv/TorqueEnable "{cmd_type: 'group', name: 'arm', enable: true}"

However, there is no clusters being created yet at this moment. Future Tasks will be finding the right tune of parameters to create clusters on rViz2 and run pick and place script to see if that works. It will be a great reference to learn how localization works.

Pick and place command:

ros2 run interbotix_xsarm_perception pick_place.py 

[IsMoNgai]

This module is working and it does generate some clusters.
demo of the module

Future implication:
We can train NN to do clustering and generate the best grasp point based on factors like center of mass, type of objects...etc.
To do that we can use this perception script as our expert for our NN to learn from (base model). Before that, we need to have a better understanding of how this clustering module works and how to improve this perception module better by tunning the parameters. Currently, this script will make false clustering and result in grasping air.

TO DO:

• Give a detail explanation and documentation of what this script does and how it generates the points
• Identity potential improvements on the script
• Identity what each of the parameters does and how it affects the clustering result

[bjl0440]

documentation doc
https://docs.google.com/document/d/13SFaRaPYzUIo0-h7GQvkMhN4XuCS_wSxV1yleTGuq64/edit?usp=sharing