Ball Catching with Kinova Gen 3 using RL

Reinforcement learning project aimed at having a Kinova Gen3 manipulator catch a ball in the air

Setup

1. Make sure you have the OmniIsaacGymEnvs repository cloned to your device. Follow their installation instructions if not.
2. Create a symbolic link from the directory OmniIsaacGymEnvs/omniisaacgymenvs/cgf/task/ to kinova_ball_catching_RL/config/KinovaTask.yaml
3. Create a symbolic link from the directory OmniIsaacGymEnvs/omniisaacgymenvs/cgf/train/ to kinova_ball_catching_RL/config/KinovaTaskPPO.yaml
4. Navigate to OmniIsaacGymEnvs/omniisaacgymenvs/utils/task_util.py.
5. Inside the import_tasks() function, add from kinova_task import KinovaTask.
6. Inside the task_map dictionary, add an entry "KinovaTask": KinovaTask.

Running

1. Add the isaac_scripts folder in this repo to your PYTHONPATH environment variable manually. Example: export PYTHONPATH=$PYTHONPATH:/path/to/isaac_scripts.
2. Navigate to the OmniIsaacGymEnvs/omniisaacgymenvs folder.
3. Run /path/to/your/isaac-sim/python.sh scripts/rl_train.py task=KinovaTask.
4. Additional arguments you can pass to that script include:
   • headless=True
   • num_envs=<how many robots you want to spawn>
   • test=True if you want to examine a policy
   • checkpoint=/path/to/a/checkpoint (Note that you always have to do this to examine a policy. Just setting test to True will not load a trained policy)
   • max_iterations=<how many epochs to run> The default is 100 and is pretty quick (about 400,000 timesteps with 256 robots). Setting to 1000 is pretty long but gives very good results.