This package provides a Task and Motion Planning (TAMP) server that offers ROS services to interact with the Panda stacking agent.
First, ensure this package is in a Catkin workspace (e.g. /catkin_ws/src/stacking/stacking_ros).
Now, build the Catkin workspace
catkin_build -DCPYTHON_EXECUTABLE=$(which python3)
Ensure that this Catkin workspace is being sourced:
setup /catkin_ws/devel/setup.bash
To check that the package works correctly, try running some of these commands:
rospack find stacking_ros
rosmsg list | grep stacking
First, start up a Panda agent that communicates with the planning server. For example,
python3 -m run_towers --use-action-server --blocks-file learning/domains/towers/final_block_set.pkl --num-blocks 10
The scripts/planning_server.py file contains the planning server that will feed plans to the Panda agent. To run this, you have a few options.
rosrun stacking_ros planning_server.py --blocks-file learning/domains/towers/final_block_set.pkl --num-blocks 10
To do this, you will need to ensure the stacking repo is on the Python path. You can force this, e.g.
export PYTHONPATH=$PYTHONPATH:/catkin_ws/src/stacking
First, go to the top-level folder of the stacking repository. Then,
python3 stacking_ros/scripts/planning_server.py --blocks-file learning/domains/towers/final_block_set.pkl --num-blocks 10
First, start the planning server with the block list of choice. Keep it handy because we need to use it for every separate call below.
rosrun stacking_ros planning_server.py --blocks-file learning/domains/towers/final_block_set.pkl --num-blocks 10
Then, start the Panda agent server with the same block list.
rosrun stacking_ros panda_agent_server.py --blocks-file learning/domains/towers/final_block_set.pkl --num-blocks 10
Finally, start active learning.
python3 -m learning.experiments.active_train_towers --exec-mode sim --use-panda-server --block-set-fname learning/domains/towers/final_block_set.pkl