Write4U
Valued Senior Member
This is an interesting exercise.
Teach your own AI to walk
Action Space
BipedalWalker has 2 legs. Each leg has 2 joints. You have to teach the Bipedal-walker to walk by applying the torque on these joints. Therefore the size of our action space is 4 which is torque applied on 4 joints. You can apply the torque in the range of (-1, 1)
Reward
Setting
In the beginning
DDPG Network Architecture
I have chosen the following hyperparameters for my network.
Program is provided.
https://towardsdatascience.com/teach-your-ai-how-to-walk-5ad55fce8bca#
Teach your own AI to walk
Action Space
BipedalWalker has 2 legs. Each leg has 2 joints. You have to teach the Bipedal-walker to walk by applying the torque on these joints. Therefore the size of our action space is 4 which is torque applied on 4 joints. You can apply the torque in the range of (-1, 1)
Reward
- The agent gets a positive reward proportional to the distance walked on the terrain. It can get a total of 300+ reward all the way up to the end.
- If agent tumbles, it gets a reward of -100.
- There is some negative reward proportional to the torque applied on the joint. So that agent learns to walk smoothly with minimal torque.
- Slightly uneven terrain.
- Hardcore terrain with ladders, stumps and pitfalls.
Setting
In the beginning
DDPG Network Architecture
I have chosen the following hyperparameters for my network.
Program is provided.
https://towardsdatascience.com/teach-your-ai-how-to-walk-5ad55fce8bca#