resibots/kaushik_2019_aprol

Adaptive prior selection for repertoire based adaptation in robotics

kaushik_2019_aprol's Issues

Minor changes required to get the kuka repertoires to train and to visualize the kuka experiments

Hi,
first of all, thanks for making public your work implementing map-elites with python.

I tried out your repository and found some things that needed to be changed in order to get the kuka repertoires to train and to visualize the kuka experiments:

1. Change absolute paths for relative paths (to work from the base directory) in lines:
   

   kaushik_2019_aprol/kuka_pushing_exps/map_elites_kuka_pushing.py

   Line 70 in c69f930

   boturdf="/nfs/hal01/rkaushik/projects/multi_dex_python/environments/URDF/kuka_iiwa/kuka_pusher.urdf",

   
   

   kaushik_2019_aprol/kuka_pushing_exps/map_elites_kuka_pushing.py

   Line 71 in c69f930

   floorurdf="/nfs/hal01/rkaushik/projects/multi_dex_python/environments/URDF/plane.urdf",

   
   

   kaushik_2019_aprol/kuka_pushing_exps/map_elites_kuka_pushing.py

   Line 72 in c69f930

   tableurdf = "/nfs/hal01/rkaushik/projects/multi_dex_python/environments/URDF/table_square/table_square.urdf")

2. env_params.txt is missing in each of the data/kuka_pusher_prior/kuka_pusher_map_toy_n_friction_1/ archives folders. I believe this is caused by *.txt being in the .gitignore file

---

Recommended changes:

1. Currently, steps are printed at every run in
   

   kaushik_2019_aprol/environments/kuka_pusher_env.py

   Line 212 in c69f930

   print(steps)

   
   I would suggest passing an argument to decide whether steps should be printed or not.

---

I have already made the changes mentioned above in my forked repository. If you are ok with it, I could create a pull request to yours.

some questions about experiment

Hi,
first of all, thanks for making public your work implementing map-elites with python. And there is no doubt this's an amazing experiment!
I tired your experiment and found some problems bothers me.

First,env_params.txt is missing in each of the data/hexapod_prior/ archives folders.If you can, maybe you can upload these.
so that i create env_params.txt by myself. It seems work. But If there is no leg damage，i generating the policy repertoires and run the experiment，what should i fill in blocked_legs?

I look forward to your reply and thank you very much！！！