[Computer-go] Cross thoughts from space trajectory design

"Ingo Althöfer" 3-Hirn-Verlag at gmx.de
Thu May 4 13:17:49 PDT 2017


Hello,

in the last few weeks I was much engaged in a 
"Global Trajectory OPtimization Competition",
which is organized every 20 months or so.
The task (to be solved within 4 weeks) was to design
space trajectories who deorbited 123 pieces of space
debris from earth orbits. Typically a mission would 
consist of several sub-missions where each sub-mission
deleted several debris pieces.

The final results are here (Team Jena on rank 10)
https://kelvins.esa.int/gtoc9-kessler-run/leaderboard/
There are also links to problem description and so on.

My partner Dietmar Wolz (well known from the Eternity Puzzle
scene) wrote:
> Because of the high search depth (> 100) even despite our 
> attempts to improve diversification, at higher depths
> long parts of the branches share the same trajectory. 
> Earlier missions have no chance to be replaced, even 
> if they block better missions in the end...

The search depth comes from the debris. Each piece adds +1
to the depth. The problem of "early dominators" is well known
from computer go (on MCTS base).
Has there recently been progress to achieve better chances for
"late comers"?

Ingo.




More information about the Computer-go mailing list