Mapping CRAM Actions to SOMA Tasks

[seba90]

Hi guys,

I created a table with the current CRAM actions which the robot is executes while setting up a breakfast table. It would be great if we can find official SOMA label for the corresponding CRAM actions which I can later use for the NEEM recording. The table is structured as follows:

• CRAM Action - The Name of the CRAM Action
• SOMA Task - The according SOMA Label
• CRAM Comment - @gaya- and her CRAM team can comment on e.g. the questions from the SOMA team
• SOMA Comments - SOMA team can add questions and comments which are related to the CRAM Action
• Status - The current status, e.g if own defined labels are used, currently in discussion or done.

CRAM Action	SOMA Task	CRAM Comment	SOMA Comment	STATUS
Accessing	Accessing			Seba: I defined the SOMA Label by myself
Closing	Closing	DONE
Delivering	Delivering	Seba: I defined the SOMA Label by myself
Detecting	Detecting	Seba: I defined the SOMA Label by myself
Fetching	Fetching	Seba: I defined the SOMA Label by myself
GetActionGrasps	GetActionGrasps	Seba: Is a reasoning task and I defined the SOMA Label by myself
GetActionGripperOpening	GetActionGripperOpening	Seba: Is a reasoning task and I defined the SOMA Label by myself
GetActionGrippingEffort	GetActionGrippingEffort	Seba: Is a reasoning task and I defined the SOMA Label by myself
GetActionTrajectory	GetActionTrajectory	Seba: Is a reasoning task and I defined the SOMA Label by myself
GetContainerClosingDistance	GetContainerClosingDistance	Seba: Is a reasoning task and I defined the SOMA Label by myself
GetContainerOpeningDistance	GetContainerOpeningDistance	Seba: Is a reasoning task and I defined the SOMA Label by myself
GetLocationPoses	GetLocationPoses	Seba: Is a reasoning task and I defined the SOMA Label by myself
Grasping	Grasping	DONE
Lifting	Lifting	DONE
looking	LookingAt	DONE
searching	LookingFor	DONE
putting	Lowering	DONE
MonitoringJointState	MonitoringJointState	Seba: I defined the SOMA Label by myself
Going	MovingTo	DONE
navigating	Navigating	DONE
opening	Opening	DONE
ParkingArms	ParkingArms:	Seba: I defined the SOMA Label by myself
Perceiving	Perceiving	DONE
picking-up	PickingUp	DONE
placing	Placing'	DONE
PositioningArm	PositioningArm	Seba: I defined the SOMA Label by myself
pulling	Pulling	Seba: I defined the SOMA Label by myself
pushing	Pushing	DONE
reaching	Reaching	DONE
releasing	Releasing	DONE
retracting	Retracting	DONE
sealing	Sealing	Seba: I defined the SOMA Label by myself
SettingGripper	SettingGripper	Seba: I defined the SOMA Label by myself
Transporting	Transporting	Seba: I defined the SOMA Label by myself

[mpomarlan]

There is already a Transporting task.

I see lots of GetX tasks. This suggests to me the following approach: how about we have a reasoning task called GetParameter which we define in SOMA, and any other GetX tasks as required in the table get defined somewhere else (e.g. in a CRAM-specific owl) as subclasses of GetParameter?

[mpomarlan]

Can you give me a definition/description for Accessing and Sealing? Likewise for SettingGripper?

Finally, is Detecting an action/task or an event? To me it seems detecting is what might happen while performing a search or monitoring task, rather than a task itself.

[gaya-]

Accessing and Sealing and silly names for "achieving a container opened or closed state", which includes navigating, finding the container, (one day it should also include detecting the handle) doing failure handling, and executing the actual opening and closing trajectories.

SettingGripper is setting gripper joint to the given joint value. It usually is used to open the gripper to a specific value (not completely) to grasp things in tight spaces. But one can also use it to close the gripper when grasping, but typically closing the gripper assumes completely going to 0 joint angle, and SettingGripper can be used to close the gripper to a specific value. Shortly said, it's a moving a gripper joint action.

Detecting is an action because our perception works on demand.

[mpomarlan]

Ok, that's clarifying most of my questions. About detecting: so how is it supposed to work? You make a query to the perception system of the form "can you see a cup here?"

[gaya-]

Yes.

[ayden175]

What is the state of this issue? Is it still open or do all CRAM actions now have a corresponding mapping (or are not planned to be added)?