search

Ericonaldo / visual_wholebody

Train a loco-manipulation dog with RL
https://wholebody-b1.github.io/
Other
65 stars 5 forks source link

About the Reward Function_rewards Alive #7

Open hanzhi0410 opened 1 week ago

hanzhi0410 commented 1 week ago

Hello, thank you for your work. I have read your code and found that there is an interesting item in the reward function called _rewards _alive. I am not sure what the specific purpose of doing this is. Can you explain it to me? Looking forward to your reply

hatimwen commented 1 week ago

In the paper, it's mentioned that the alive reward encourages the robot not to fail. In the code, this is implemented by awarding 1 point for each step the robot takes, which motivates the agent to run for as many steps as possible, essentially encouraging it to stay alive and avoid failure. This is my understanding.