🚧 Work in progress: this documentation is still being written and reviewed. Check details against the source code.

Vision task action server

Expose generic vision-task execution and TF reset, and warm the spincoater inference model in a daemon thread. motion_kind is inferred from the detected-pose result rather than the executed target type.

beambot.action_servers.vision_task_server · Source

Perception and mapping · Guide · Symbol index

How it is used

The orchestrator sends generic detector-and-motion requests to beambot_vision_task. This server constructs VisionTaskStages, delegates each goal, and copies detection and execution results into VisionTaskAction.Result.

create_stages() connects beamline camera settings to the coordinator. Startup also warms the spincoater model in a background thread; the TF-reset service updates the engine after tool-model changes.

Start here

Callable	Purpose
main	Run VisionTaskActionServer through the shared multithreaded ROS server runner.
VisionTaskActionServer.create_stages	Load beamline camera settings and construct the generic vision coordinator.
VisionTaskActionServer._execute	Run vision stages, map error strings to failed results, and copy unchecked vacuum_ok and optional detected approach. motion_kind is NONE when last_detected_pose is truthy, otherwise CARTESIAN_POSE, including joint-target executions.

Functions

main()

Run VisionTaskActionServer through the shared multithreaded ROS server runner.

Entry point · Source

main(args=None)

Parameter	Meaning
args	Optional ROS initialization arguments; None uses normal rclpy argument handling.

Returns: None when the server exits.

Errors: KeyboardInterrupt during spin is handled by run_server; initialization/construction and other executor errors propagate.

Side effects: Initializes ROS, constructs the action node, spins until shutdown and destroys the node/context.

VisionTaskActionServer

Concrete BaseActionServer around the generic vision coordinator. Its execution override supplies vision-specific result fields; camera, detection, and motion sequencing live in VisionTaskStages.

Inherits from: BaseActionServer.

VisionTaskActionServer.create_stages()

Load beamline camera settings and construct the generic vision coordinator.

Extension hook · Source

create_stages(self)

Returns: VisionTaskStages instance.

Errors: Configuration or VisionTaskStages initialization errors propagate.

Side effects: Reads configuration, initializes the vision engine and logs camera configuration.

VisionTaskActionServer._execute()

Run vision stages, map error strings to failed results, and copy unchecked vacuum_ok and optional detected approach. motion_kind is NONE when last_detected_pose is truthy, otherwise CARTESIAN_POSE, including joint-target executions.

Extension hook · Source

_execute(self, goal_handle)

Parameter	Meaning
goal_handle	Accepted VisionTaskAction handle.

Returns: VisionTaskAction.Result with optional metre XYZ and XYZW quaternion.

Errors: Stage exceptions propagate to the base wrapper.

Side effects: Runs perception/motion; fills result fields. Goal finalization belongs to BaseActionServer.

VisionTaskActionServer.__init__()

Register beambot_vision_task and its TF reset service, then start best-effort spincoater model warmup.

Lifecycle · Source

__init__(self)

Errors: Vision/configuration/ROS initialization errors propagate; warmup errors are handled in its worker.

Side effects: Creates stage/camera/ROS resources and starts a daemon inference thread.

Internal callbacks and helpers

VisionTaskActionServer._warmup_spincoater_model()

Start a daemon thread that loads the shared sample detector and performs one dummy inference to initialize its resources.

Internal helper · Source

_warmup_spincoater_model(self)

Returns: None immediately after starting the thread.

Errors: Thread-start errors propagate; worker exceptions are logged and suppressed.

Side effects: Starts a background thread; may load model weights and initialize inference resources.

Referenced in source (1)

• VisionTaskActionServer.__init__ — beambot.action_servers.vision_task_server

VisionTaskActionServer._warmup_spincoater_model._warmup()

Load the spincoater sample model and infer on a black 640×640 uint8 RGB image with confidence 0.5.

Internal helper · Source

_warmup()

Returns: None.

Errors: Catches import/load/inference exceptions and logs a skipped-warmup warning.

Side effects: Loads/caches model weights, allocates a dummy image and initializes model inference resources; logs readiness.

Referenced in source (1)

• VisionTaskActionServer._warmup_spincoater_model — beambot.action_servers.vision_task_server

VisionTaskActionServer._reset_tf_callback()

Delegate stage TF reset and populate a successful Trigger response.

Callback · Source

_reset_tf_callback(self, request, response)

Parameter	Meaning
request	Trigger request; unused.
response	Response object to populate.

Returns: Same response with success=True and reset message.

Errors: Reset exceptions propagate.

Side effects: Replaces engine TF resources and mutates response.

Referenced in source (1)

• VisionTaskActionServer.__init__ — beambot.action_servers.vision_task_server