About Crawler for Continuous Control
ZeratuuLL opened this issue · comments
I am trying to solve the Crawler environment in Continuous control task. I have read the Unity webpage and realized that there were two environments. One with static target and one with dynamic target. Which one is provided through the links? Thank you!
I too wanted to know the answer please.
After some test it should be the fixed goal....
I think so too. I was training my humble agent and it’s gaiining score and the visual shows that its always facing the same direction. I guess it is a fixed goal. Thanks man!
sam
… On 13 Apr 2019, at 2:42 AM, Lifeng Wei ***@***.***> wrote:
After some test it should be the fixed goal....
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub <#16 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AMvt5Sz9ja7TCTA79SP0BcSpfUGoH_cIks5vgNOpgaJpZM4bJKm->.
Np! What kind of results are you getting? The dimension info is different from Unity page so I am not sure if the benchmarks are still reliable.... I can get an average around 1800 but cannot get further....
Lifeng Wei.
From my phone and apology for any typos
…
On Apr 12, 2019 at 11:44 AM, <Samuel Pun ***@***.***)> wrote:
I think so too. I was training my humble agent and it’s gaiining score and the visual shows that its always facing the same direction. I guess it is a fixed goal. Thanks man!
sam
> On 13 Apr 2019, at 2:42 AM, Lifeng Wei ***@***.***> wrote:
>
> After some test it should be the fixed goal....
>
> —
> You are receiving this because you commented.
> Reply to this email directly, view it on GitHub <#16 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AMvt5Sz9ja7TCTA79SP0BcSpfUGoH_cIks5vgNOpgaJpZM4bJKm->.
>
—
You are receiving this because you modified the open/close state.
Reply to this email directly, view it on GitHub (#16 (comment)), or mute the thread (https://github.com/notifications/unsubscribe-auth/AVj1AKuNAhHfjsJ0xpvm-oJ4soxzgyvQks5vgNQmgaJpZM4bJKm-).
envy you! I have just started and I am trying to use PPO to solve the problem. I am only able to get to around 100 scores (average of rewards of pass 100 episodes across agents once any of the agent reached ‘done’). May I ask how many layers of network did you use?
… On 13 Apr 2019, at 5:26 AM, Lifeng Wei ***@***.***> wrote:
Np! What kind of results are you getting? The dimension info is different from Unity page so I am not sure if the benchmarks are still reliable.... I can get an average around 1800 but cannot get further....
Lifeng Wei.
From my phone and apology for any typos
>
> On Apr 12, 2019 at 11:44 AM, <Samuel Pun ***@***.***)> wrote:
>
>
> I think so too. I was training my humble agent and it’s gaiining score and the visual shows that its always facing the same direction. I guess it is a fixed goal. Thanks man!
>
> sam
> > On 13 Apr 2019, at 2:42 AM, Lifeng Wei ***@***.***> wrote:
> >
> > After some test it should be the fixed goal....
> >
> > —
> > You are receiving this because you commented.
> > Reply to this email directly, view it on GitHub <#16 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AMvt5Sz9ja7TCTA79SP0BcSpfUGoH_cIks5vgNOpgaJpZM4bJKm->.
> >
>
>
>
> —
> You are receiving this because you modified the open/close state.
> Reply to this email directly, view it on GitHub (#16 (comment)), or mute the thread (https://github.com/notifications/unsubscribe-auth/AVj1AKuNAhHfjsJ0xpvm-oJ4soxzgyvQks5vgNQmgaJpZM4bJKm-).
>
>
>
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub <#16 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AMvt5UiNFbU-q7GvI0PgoYfGZsrKodrwks5vgPoggaJpZM4bJKm->.
You can check my repo here. https://github.com/ZeratuuLL/Reinforcement-Learning/tree/master/Continuous%20Control/Crawler
It's not easy to say the structure directly....