Skip to content

Latest commit

 

History

History
83 lines (76 loc) · 3.8 KB

intro.md

File metadata and controls

83 lines (76 loc) · 3.8 KB
Raw

DarwinSport

Build image

  • To build new image:
    • Start docker Daemon
    • Build clojure standalone: lein uberjar
    • Confirm that snapshot jar name matches that in Dockerfile
    • Build: docker build -t <imagename> .
    • Start: docker run -i -t <imagename>:latest \bin\bash -e "INSTANCE=<nodeID>"
  • Note: to minimize image size/instance size check in stdout and use .dockerignore to remove unnecessary build inclusions:
    • Sending build context to Docker daemon <size>MB

Resources

  • Player config files can be found in resources/players
  • Decision code files can be found in resources/decisionfiles
  • Graphical resources can be found in resources/images

Graphical mode

  • Darwin has its own builtin graphical abilities
  • DarwinSport has a real-time top-down soccer game window. This should typically be used outside of a run
    • Note: graphical mode slows computation to realtime (throttled by framerate)

Usecase

  • The image built using this repo is meant to run on a cluster node
  • The node opens a server socket that accepts connections
  • It executes the tests for the individual received by the connection, and returns the transformed individual
  • It then accepts the next connection
  • This is meant to be run in a docker swarm with multiple containers per host
  • Testing is not parallel within a node

Player Attributes

  {:location (400 350)
   :facing-angle 0
   :possessing-ball? false
   :open? true
   :user-controlled? true                         ;signifies if a player can be user controlled   
   :autonomous-when-idle? true                    ;if player can be user controlled, this will access defined decisions in the absence of user input
   :assigned-image "images/players/team1player4.png"
   :defined-decisions "testfiles/decisionfiles/offensive_decisions.txt"}

Decision Code

Syntax example:

and ball-settle-radius? !self-ball-possessed? : action-settle-ball
and self-ball-possessed? self-space? shooting-range? : action-shoot
and self-ball-possessed? !self-space? team-mate-open? : action-leading-pass
and self-ball-possessed? self-space? : action-dribble-forward
and !self-ball-possessed? team-possessing-ball? : action-forward-run
and self-closest-to-ball? !opponent-possessing-ball? : action-recover-ball
and !opponent-possessing-ball? ball-forward? : action-forward-run
  • IMPORTANT NOTE: later actions have the potential to overwrite actions at the beginning unless :first-decision-only? true is specified in config!
    • i.e. if an instruction causes the player to move in one direction, a future action in a different direction will overwrite that state
    • Therefore, actions should be designed to be clear cut and care should be exercised when writing actions that could potentially conflict
  • All predicates can be inverted by adding ! on the front.
  • All predicate decision pairs are structured and|or p p ... p : a a ... a
  • The following are valid predicates (not all listed here):
    • self-ball-posessed?
    • self-space?
    • team-mate-open?
    • self-defensive-third?
    • self-offensive-third?
    • team-possessing-ball?
    • opponent-possessing-ball?
    • tackle-range?
    • close-to-goal?
      • (And inverses)
  • The following are valid actions:
    • action-longest-pass-forward
    • action-self-dribble-forward
    • action-short-pass-forward
    • action-clear
    • action-shoot
    • action-tackle
    • action-follow-ball
    • action-defensive-drop

Config

  • Image currently setup for raspberry pi java (for cluster)
  • TCP port 5555 used to communicate across local swarm network
  • Graphical mode should not be enabled in swarm mode as it slows computation to real time and introduces inefficiencies.