search

gnns4hri / SocNavGym

GNU General Public License v3.0
15 stars 2 forks source link

Missing dependencies and test default config files #1

Closed luigiberducci closed 4 months ago

luigiberducci commented 6 months ago

Hi there, thanks for making this environment available.

While using it, I noticed a couple of missing dependencies and some issue in running the environment with config file "exp2_*yaml". It seems the human_list is empty and it raises an index error when trying to accessing it socnavgym/envs/socnavenv_v1.py:3055.

To reproduce, please find the test in the tests directory, which loops over all config files and tries to create an environment for each of them.

Hope it helps! Luigi

sushant1212 commented 5 months ago

Hi @luigiberducci Sorry for the late reply. Thanks for pointing out the bug. I've catered to it in the latest changes. I've also added a test inspired from your PR for testing the configs. Do let me know if the issue persists. Also, I couldn't find the need for pandas anywhere. Did you need it while installing?

Do share your feedback on your experience using SocNavGym!

luigiberducci commented 4 months ago

Thanks, I will close this pr for now.

About the environment, I would like to see it ported to gymnasium apo rather than gym api. This will give longer term support and compatibility with most recent RL frameworks. The changes are minimal.

About the env dynamics, I found it a bit weird to have very large integration timestep (1.0) and very restrictive bounds in the control input (max velocity). I would usually expect to control it at higher frequency.

Anyway, it is a nice env. Hope it helps to improve it! :)

sushant1212 commented 4 months ago

Thanks for your prompt feedback @luigiberducci We are planning to migrate to gymnasium, so do keep an eye out for that! Regarding the timestep and max velocities, you can configure them using the config files. The keys time_step and max_advance_robot would help you change as per your requirements.