Skip to content

Update outdated openarm_bringup README - #98

Merged
kou merged 1 commit into
enactic:mainfrom
pradeepsuryad:update-bringup-readme
Oct 6, 2026
Merged

kou merged 1 commit into
enactic:mainfrom
pradeepsuryad:update-bringup-readme

Conversation

@pradeepsuryad

Copy link
Copy Markdown
Contributor

Follow-up to #97, as discussed there.

openarm_bringup/README.md still described the launch interface from before #91:

  • openarm.launch.py (single arm) no longer exists. openarm.bimanual.launch.py is the only launch file.
  • hardware_type and can_interface are no longer launch arguments. Fake vs. real hardware is use_fake_hardware, and each arm has its own right_can_interface / left_can_interface.
  • The arm_type default is now openarm_v2.0, not v10.
  • The Quick Start said "fake hardware" but passed hardware_type:=real.

This updates the Quick Start (v2.0 with fake hardware, and v1.0 with real hardware), the launch file list and the parameter list. Every argument and default in the list matches what openarm.bimanual.launch.py declares.

Test

On ROS 2 Jazzy:

  • ros2 launch openarm_bringup openarm.bimanual.launch.py: both mock hardware interfaces load, and joint_state_broadcaster, both joint trajectory controllers and both gripper controllers are configured and activated.
  • ros2 launch openarm_bringup openarm.bimanual.launch.py arm_type:=v10 use_fake_hardware:=false: OpenArmHW loads with CAN=can1, ..., can_fd=enabled. No CAN device was attached, so it stops when it opens the socket.
  • ros2 launch openarm_bringup openarm.bimanual.launch.py robot_controller:=forward_position_controller: both forward position controllers are activated.

Note

I left arm_prefix out of the parameter list because setting it currently breaks controller loading. With arm_prefix set, openarm.bimanual.launch.py switches the controllers file to openarm_bimanual_controllers_namespaced.yaml. That file was removed in #91, when openarm_v10_bimanual_controllers.yaml was renamed to openarm_bimanual_controllers.yaml, and the namespaced variant wasn't re-added. For example, ros2 launch openarm_bringup openarm.bimanual.launch.py arm_prefix:=leader gives:

[WARNING] [launch_ros.actions.node]: Parameter file path is not a file: .../openarm_bringup/config/controllers/openarm_bimanual_controllers_namespaced.yaml
[ERROR] [leader.controller_manager]: The 'type' param was not defined for 'joint_state_broadcaster'.
[FATAL] [leader.spawner_joint_state_broadcaster]: Failed loading controller joint_state_broadcaster

I can fix that in a separate PR if that would help.

The README still described the launch interface from before enactic#91:
openarm.launch.py no longer exists, and hardware_type and can_interface
are not launch arguments. Fake vs. real hardware is use_fake_hardware,
each arm has its own right_can_interface / left_can_interface, and
arm_type now defaults to openarm_v2.0.

Update the Quick Start, the launch file list and the parameter list to
match openarm.bimanual.launch.py.
@kou
kou merged commit b9d7a67 into enactic:main Oct 6, 2026
1 check passed
@kou

kou commented Oct 6, 2026

Copy link
Copy Markdown
Collaborator

Thanks!

I can fix that in a separate PR if that would help.

Please do it!

kou pushed a commit that referenced this pull request Oct 20, 2026
Follow-up to #98, as discussed there.

With `arm_prefix` set, `openarm.bimanual.launch.py` switches the
controllers file to `openarm_bimanual_controllers_namespaced.yaml`,
which was removed in #91. `ros2_control_node` then starts without any
controller parameters, and no controller loads:

```
[WARNING] [launch_ros.actions.node]: Parameter file path is not a file: .../openarm_bringup/config/controllers/openarm_bimanual_controllers_namespaced.yaml
[ERROR] [leader.controller_manager]: The 'type' param was not defined for 'joint_state_broadcaster'.
[FATAL] [leader.spawner_joint_state_broadcaster]: Failed loading controller joint_state_broadcaster
```

Rather than re-adding a namespaced copy, this prefixes every node name
in `openarm_bimanual_controllers.yaml` with `/**/`, which matches the
node in any namespace, including none, and removes the file switch. One
file then covers both cases. Before #91 the two files had already
drifted apart; the namespaced one used different gripper controller
types, for example. If you'd rather keep a separate namespaced file, I
can do it that way instead.

It also adds `arm_prefix` to the parameter list in
`openarm_bringup/README.md`, which #98 left out because of this bug.

## Test

On ROS 2 Jazzy with mock hardware, checked with `ros2 control
list_controllers`:

| Launch arguments | Before | After |
|---|---|---|
| (none) | all 5 controllers active on `/controller_manager` | same |
| `arm_prefix:=leader` | no controllers loaded | all 5 active on
`/leader/controller_manager`; `/leader/joint_states` published |
| `arm_prefix:=leader robot_controller:=forward_position_controller` |
not tested | both forward position controllers, both gripper controllers
and `joint_state_broadcaster` active |
| `arm_prefix:=/leader/` | not tested | same as `arm_prefix:=leader` |

The 5 controllers are `joint_state_broadcaster`,
`left_joint_trajectory_controller`, `right_joint_trajectory_controller`,
`left_gripper_controller` and `right_gripper_controller`.
asus-youhao added a commit to asus-youhao/openarm_ros2 that referenced this pull request Oct 30, 2026
…ere is no command

Requested behaviour: /joint_actions keeps flowing when nothing is
commanded, carrying /joint_states for every limb without a command.
The event-driven model (phu enactic#84-enactic#91) went silent instead, and the
latest phu aggregator (enactic#98) also moved to a fixed-rate timer.

Publish on a --publish_rate timer (50 Hz) from startup. Per limb and
tick: a live command (arrived within max(3 measured periods, 30 ms))
is recorded; otherwise, never commanded, stopped, rejected (FPC not
loaded, wrong `joints`, wrong length) or not subscribed, the limb is
filled from /joint_states. A joint missing from /joint_states skips
the tick instead of inventing 0.0. Switches between command and
/joint_states are logged per limb. header.stamp is the publish time.

Differences from phu enactic#98: a stopped limb is filled from /joint_states
instead of holding its last command, and a missing joint skips the
tick instead of publishing 0.0. The same-step arm pairing is gone:
a timer samples the latest command of each limb.

docs/JOINT_ACTIONS.md lists what /joint_states filling costs (action
equals observation while nothing is commanded, a held FPC target is
not the measured position on real hardware).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants