Try to learn ObstaclePathing

weidler / RLaSpa

Reinforcement Learning in Latent Space

MIT License

5 stars 1 forks source link

Try to learn ObstaclePathing #4

Closed weidler closed 5 years ago

weidler commented 5 years ago

[x] Learn obstacle pathing using state of the agent
[x] Learn obstacle pathing using visual input

HansBambel commented 5 years ago

There is an issue in obstaclepathing where it is possible to have an obstacle over the target.

HansBambel commented 5 years ago

Implemented with DQN

HansBambel commented 5 years ago

Reopened, now working on visual input

HansBambel commented 5 years ago

I think we need to do #9 first for efficient visual representation. Otherwise we have to flatten a list every time we need to get the state. List flattening: flat_list = [item for sublist in l for item in sublist]

HansBambel commented 5 years ago

Maybe it is also helpful to have the visual representation be given back as state. Then we dont need to construct the visual representation every time. We only update the representation.

HansBambel commented 5 years ago

Since there are multiple representation learners that are able to learn using visual input we can close this issue.