sjwang2015 / aemcarl Goto Github PK

Dear owners,

As far as I can tell it doesnt seem like the code for your proposed AEMCARL solution is included in this repo. Would it be possible to add this?

Many thanks,

Sahil

Could you share me your husky_ws that had been amended? Thanks!

After examining the predict function(line 310) in crowd_nav/policy/multi_human_rl.py it seems that even during testing the values for each of the actions are determined by propogating the current state and deriving the value estimate as follows value = reward + pow(self.gamma, self.time_step * state.self_state.v_pref) * next_state_value. Surely at test time you cannot do this and you must use the model only to give you value estimates since you wouldn't have access to a reward and estimates of the next_state_value? Am I missing something?

大佬你好,请问AEMCARL算法是"actenvcarl.py"么,他的具体实现在./common目录下的那些么,谢谢大佬

Hi.
I really appreciate for sharing your work.

I'm very interested in your work that conducted in Gazebo.
Could you also share that codes?

Thank you.

Hello, do you train faster with CPU or with GPU? How long did it take you to train your policy to converge? Thanks!