How to use Openai with ROS (Python3)
Hi guys, I really hope someone will answer because I pretty stuck. I've been following the Construct tutorial on how to use ROS with Python3 but I still have trouble launching the training.
I am working on my machine (ROS kinetic + Ubuntu 16.04) and so far I have 3 workspaces: catkin_ws, python3_ws and simulation_ws.
All the workspaces have been compiled with:
catkin_make -DPYTHON_EXECUTABLE:FILEPATH=/home/user/catkin_ws/py3/bin/python
This is the structure (I put only their src/ folders):
/home/p16325mr/catkin_ws/src/ <-- here's the stuff that import the algorythms
- ├── my_sawyer_openai_example
- ├── openai_ros
- └── spawn_robot_tools
/home/p16325mr/python3_ws/src <-- here's the stuff written in python3
- ├── geometry
- ├── geometry2
- └── stable-baselines
/home/p16325mr/simulation_ws/src
- ├── intera_common
- ├── intera_sdk
- ├── sawyer_robot
- └── sawyer_simulator
I launch the simulation in one terminal (which works) and the training in the other terminal (after activating py3venv) The training (roslaunch my_sawyer_openai_example train_a2c.launch) raises error:
Traceback (most recent call last): File "/opt/ros/kinetic/lib/gazebo_ros/spawn_model", line 32, in <module> import tf.transformations as tft [...]
ImportError: dynamic module does not define init function (init_tf2)
It's obviously related to the python3 virtual environment... Is there a workaround for the problem?
