Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Minecraft's respawn grace period can skew results #88

Closed
DaveyBiggers opened this issue Jun 21, 2016 · 1 comment · Fixed by #366
Closed

Minecraft's respawn grace period can skew results #88

DaveyBiggers opened this issue Jun 21, 2016 · 1 comment · Fixed by #366
Labels
Milestone

Comments

@DaveyBiggers
Copy link
Member

Minecraft appears to have a (not unreasonable) grace period after respawning, during which the agent is invulnerable. This can break experiments if the agent tends to end each mission by dying.

For example, remove lava from the <AgentQuitFromTouchingBlockType> section of cliff_walking_1.xmland then run thetabular_q_learning.pyexample. The agent will quickly die in lava, then be respawned for the next mission. If, in the new mission, it enters lava again before the grace period has expired, it will be able to run around happily for a while, thus breaking the q-table.

@DaveyBiggers DaveyBiggers self-assigned this Jun 21, 2016
@DaveyBiggers DaveyBiggers removed their assignment Aug 15, 2016
@DaveyBiggers
Copy link
Member Author

This is due to EntityPlayerMP.respawnInvulnerabilityTicks - set to 60 at construction. (Equates to 3 seconds at normal game speed.)

Annoyingly private, but easy enough to get around with reflection.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging a pull request may close this issue.

1 participant