[Bug]: Task Continues Execution After a dispenser_result Reports Failure

[alex-roba]

Before proceeding, is there an existing issue or discussion for this?

• I have done a search for similar issues and discussions.

OS and version

24.04

Open-RMF installation type

Source build

Other Open-RMF installation methods

No response

Open-RMF version or commit hash

main

ROS distribution

Jazzy

ROS installation type

Binaries

Other ROS installation methods

No response

Package or library, if applicable

No response

Description of the bug

In the current implementation of Open RMF task execution, when an action like a dispenser interaction reports a fail result (e.g., via dispenser_result), the overall task is correctly marked as failed.
However, the task sequence does not halt. Any remaining activities or phases scheduled after the failed dispenser action continue to be executed by the robot.

Steps to reproduce the bug

1. A task is dispatched that includes a dispenser action followed by other actions (e.g., moving to a new location).
2. The dispenser action fails and sends a dispenser_result with a failure status.
3. The task's overall state is reported as failed in the RMF dashboard/API.
4. Despite the failure, the fleet adapter proceeds to execute the next phase in the task sequence (e.g., the robot starts moving to the next waypoint).

Expected behavior

No response

Actual behavior

No response

Additional information or screenshots

No response

[mxgrey]

Unfortunately the current task system doesn't have a way to specify how a failed event should be handled. Stopping the task isn't always the right approach, e.g. a navigation failure could mean that human intervention is needed, not that the task should be cancelled.

The recommendation for current generation system integrators is to monitor the task state and trigger a cancellation when you detect a failure that warrants the task to be cancelled for your application.

In the next generation, workflows will allow users to specify the exact behaviors they want, including potentially how to recover from errors.