Unable to run the Expert Planner

[yanrui89]

I am trying to generate my own data and tried running the following command python dagger_training.py --settings_file=config/dagger_settings.yaml as suggested in the README.

If I understand correctly, ln 102 of dagger_training.py calls the run_mppi_expert function, which is in ln 60 of this code https://github.com/uzh-rpg/agile_autonomy/blob/main/planner_learning/src/PlannerLearning/PlannerLearning.py. The run_mppi_expert function seems to be publishing to a topic called /hummingbird/start_label.

And this seems to be the place where the code gets stuck. and it seems to be just waiting for the expert to be done. However, I did a search on this repo and no where else seems to be subscribing to the /hummingbird/start_label topic. Hence, the code gets stuck. Anyone has any idea if I needed to run something else to get the expert running?

Thank you.

[antonilo]

Can you try to increase the sleep time before starting Unity?

[mw9385]

Hi, I have a similar problem. After running the following command python dagger_training.py --settings_file=config/dagger_settings.yaml, the drone stands still and does not move even though enough time has passed. It kept repeating reset environment. By following your command on #17, I made the directory data in data_generation directory. Unfortunately it still doesn't work.

When I debugged the python script dagger_training,py, I found that __self.learner.maneuver_complete always shows False__ and I have difficulty in finding functions that updates the variable. Anyone has similar problems and solved the issue?

Thank you.

[yanrui89]

Can you try to increase the sleep time before starting Unity?

Hi Antonio, thanks for your reply. However, I tried lengthening the sleep time but it is still not working.

So when I start running the dagger_training.py. I managed to get the above. So the quadcopter started running. Is this the expert running? It doesn't seem to be the expert running as it didn't go past https://github.com/uzh-rpg/agile_autonomy/blob/d95acdd8191c73aad4bb6f5bd454ff7a7ad76384/planner_learning/dagger_training.py#L102 in the code.

So my code is stuck right here at ln 102.

[yanrui89]

Hi, I have a similar problem. After running the following command python dagger_training.py --settings_file=config/dagger_settings.yaml, the drone stands still and does not move even though enough time has passed. It kept repeating reset environment. By following your command on #17, I made the directory data in data_generation directory. Unfortunately it still doesn't work.

When I debugged the python script dagger_training,py, I found that self.learner.maneuver_complete always shows False and I have difficulty in finding functions that updates the variable. Anyone has similar problems and solved the issue?

Thank you.

May I know which part of the code you are looking at? Is your code stuck at https://github.com/uzh-rpg/agile_autonomy/blob/d95acdd8191c73aad4bb6f5bd454ff7a7ad76384/planner_learning/dagger_training.py#L81

Mine seemed to be stuck at this point https://github.com/uzh-rpg/agile_autonomy/blob/d95acdd8191c73aad4bb6f5bd454ff7a7ad76384/planner_learning/dagger_training.py#L102

[mw9385]

Hi, my code get stuck at this line. https://github.com/uzh-rpg/agile_autonomy/blob/d95acdd8191c73aad4bb6f5bd454ff7a7ad76384/planner_learning/dagger_training.py#L79 When I check the code, the variable self.learner.maneuver_complete is update in callback function __callback_fly__ in PlaneBase.py. https://github.com/uzh-rpg/agile_autonomy/blob/d95acdd8191c73aad4bb6f5bd454ff7a7ad76384/planner_learning/src/PlannerLearning/PlannerBase.py#L77 The callback function is updated through the topic name /hummingbird/agile_autonomy/start_flying. So, I echoed the topic name /hummingbird/agile_autonomy/start_flying, but I didn't get any message.

So, I tried with some other topics that updates /hummingbird/agile_autonomy/start_flying. It is updated from the node /test_primitive. When I tried to echo some topics starting with /test_primitive (I cannot correctly remember this), I got some error message like: Error: Cannot load message class for [planning_ros_msgs/Trajectory]. Are your message built?

Finally, I found that the topics from __global_planner is not correctly build. The package rpg_mpl_ros__ has been ignored during the catkin build step.

So, I think you should check whether your rpg_mpl_ros packages are correctly built. In my setting environment, when I tried to build the package isolated through the command catkin build rpg_mpl_ros, I got error message that the given package is not in the src directory.

[yanrui89]

@mw9385 I am assuming you managed to resolve your issue and could run the dagger_training.py? I checked and there were no such error messages. Thanks for the advice.

Did you managed to get your code past this line? https://github.com/uzh-rpg/agile_autonomy/blob/d95acdd8191c73aad4bb6f5bd454ff7a7ad76384/planner_learning/dagger_training.py#L102

Seems like this calls the function here https://github.com/uzh-rpg/agile_autonomy/blob/d95acdd8191c73aad4bb6f5bd454ff7a7ad76384/planner_learning/src/PlannerLearning/PlannerLearning.py#L63 and then it is supposed to publish to a topic /hummingbird/start_label. Do you know where is the subscriber to this topic?

[mw9385]

@yanrui89 Not yet. I am still struggling. My code get stuck here as the variable self.learner.maneuver_complete is not updated to true. https://github.com/uzh-rpg/agile_autonomy/blob/d95acdd8191c73aad4bb6f5bd454ff7a7ad76384/planner_learning/dagger_training.py#L79

I will check the subscriber and publisher of /hummingbird/start_label. What about checking nodes and topic through __rqt_graph__? In the rqt_graph, /hummingbird/start_label --> /hummingbird/generate_label_#(1~7) --> /hummingbird/labeling_completedKeep in touch :)

[yanrui89]

So above you mentioned that the 'So, I tried with some other topics that updates /hummingbird/agile_autonomy/start_flying. It is updated from the node /test_primitive.'

So the callback_fly is the callback function from the subscriber to the topic /hummingbird/agile_autonomy/start_flying and the publisher to the topic is this? https://github.com/uzh-rpg/agile_autonomy/blob/c9ded9a6ecf0aded8319850c81df23a65bf97960/data_generation/agile_autonomy/launch/simulation.launch#L59

Which seems like it is launching this node. https://github.com/uzh-rpg/agile_autonomy/blob/c9ded9a6ecf0aded8319850c81df23a65bf97960/data_generation/agile_autonomy/src/agile_autonomy.cpp#L58

Is this similar to what you are thinking? I am not sure why the publisher to the /hummingbird/agile_autonomy/start_flying didn't get updated in your case...

[mw9385]

@yanrui89 Hi, the topic /hummingbird/agile_autonomy/start_flying is initiated when the global planner finished making trajectories; however it seems that my global planner is not working. When I tried to reach the package through roscd rpg_mpl_ros, I cannot access to it although other packages such as agile_autonomy, flightrender are fine. Could you check whether you could reach the rpg_mpl_ros package through roscd command?

[yanrui89]

@mw9385 I can't seem to access rpg_mpl_ros through the roscd command. But I was pretty sure there weren't any errors during the building of rpg_mpl_ros

[mw9385]

@yanrui89 Thanks. I don't know what is the cause of problem then. I rebuild the given packages couple of times but the drone still doesn't move. Can docker environment cause the issue? I am build the packages in docker environment.

So to sum up, you succeed in building without any errors. But you get stuck at some point and the drone cannot move. As your global planner can generate global trajectories (as we could see from your figures above), we need to check on generate_label topic. Did you check the topics related to global planner (/test_primitive/~) are publishing messages?

[antonilo]

Sorry but we do not provide support for docker

[BAYMAXer]

I have the same problem. Did you solve it? thank you very much!

Can you try to increase the sleep time before starting Unity?

Hi Antonio, thanks for your reply. However, I tried lengthening the sleep time but it is still not working.

So when I start running the dagger_training.py. I managed to get the above. So the quadcopter started running. Is this the expert running? It doesn't seem to be the expert running as it didn't go past

https://github.com/uzh-rpg/agile_autonomy/blob/d95acdd8191c73aad4bb6f5bd454ff7a7ad76384/planner_learning/dagger_training.py#L102

in the code. So my code is stuck right here at ln 102.

[Zedzou]

hello, have you solve it? I have the same problem, but dont know how to solve.