Cannot retrieve robot state via readOnce() while using AsyncPositionControlHandler

[ChloeZ427]

Description

I am using franka::AsyncPositionControlHandler for asynchronous position control, with a joint position target callback defined as follows:

auto calculate_joint_position_target =
[&initial_position, &time, &direction](
std::chrono::milliseconds period)
-> franka::AsyncPositionControlHandler::JointPositionTarget;

During execution of AsyncPositionControlHandler, I found that I am unable to retrieve the current robot state using robot.readOnce(). The call either blocks or does not return valid/updated state information while the async control handler is active.

This makes it difficult to access and update the actual robot state (e.g., joint positions, velocities, external torques) during asynchronous control execution.

Expected Behavior

While using AsyncPositionControlHandler, I would expect to be able to:

Access the current robot state during control execution

Periodically read and update the real robot state (e.g., for monitoring, logging, or feedback purposes)

For example, being able to safely call:

franka::RobotState state = robot.readOnce();

while the async control handler is running.

Observed Behavior

Calling robot.readOnce() during AsyncPositionControlHandler execution does not work as expected

Robot state cannot be retrieved or updated while the async handler is active

Question

What is the recommended way to access and update the actual robot state while using AsyncPositionControlHandler?

Specifically:

Is it expected that readOnce() cannot be used during async control?

Is there an alternative API or callback mechanism to retrieve RobotState during asynchronous position control?

Should robot state be accessed inside the async control callback itself, or via another thread-safe interface?

Any guidance or best practices would be greatly appreciated.

Environment

Robot: Franka Research 3

Control mode: AsyncPositionControlHandler

OS: Ubuntu 20.04

libfranka version: 0.18.2

[kmohyeldine]

Thanks @ChloeZ427 will have a look

[AndreasKuhner]

Hi @ChloeZ427 ,
thank you for the feedback. It is already 'on the plan' to add additional features - our first iteration is still in the 'proof-of-concept' stage. Hence, more to come!

(if you need the robot_state right now and you can't wait for our not-yet-planned update, you can just store the robot_state which is received in the getFeedback function and store it in the class + have a getter - I hope that helps as an intermediate workaround).

Cheers,
Andreas

[ChloeZ427]

Thank you for your help! @AndreasKuhner @kmohyeldine

However, after reviewing the libfranka source code, I noticed that there isn't a function called getFeedback as you mentioned. The documentation only defines getTargetFeedback:

auto position_control_handler = async_position_control_handler_result.handler;
auto target_feedback = position_control_handler->getTargetFeedback();

Moreover, the related definition only includes:

// This structure holds feedback information about the current target state.
struct TargetFeedback {
TargetStatus status{TargetStatus::kIdle};

std::optionalstd::string error_message{};
};

This structure contains only a status field and an error message—there is no robot state information included. Where can I find the actual robot state data?

[AndreasKuhner]

Sorry, yes, I meant the getTargetFeedback function.

Where can I find the actual robot state data?

So far we didn't expose it with that new interface. Yet, if you need it, you could also just put that data field into the TargetFeedback an return it along :) (until we found a nice way to do it).

Cheers,
Andreas