| """Scripted policy against a running SUPERBEAM server (run right; right+jump near gaps/walkers). |
| |
| python play_scripted.py [--url http://localhost:8000] [--level classic] |
| """ |
| import argparse |
| import os |
| import re |
| import sys |
|
|
| sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) |
| from superbeam_env import SuperBeamAction, SuperBeamEnv |
|
|
|
|
| def policy(obs): |
| ahead = [int(v) for v in re.findall(r"\((-?\d+) px ahead\)", obs.facts)] |
| if "standing on ground" in obs.facts and any(0 <= d < 8 for d in ahead): |
| return "right+jump" |
| return "right" |
|
|
|
|
| def main(): |
| ap = argparse.ArgumentParser() |
| ap.add_argument("--url", default="http://localhost:8000") |
| ap.add_argument("--level", default=None) |
| a = ap.parse_args() |
| kw = {"level": a.level} if a.level else {} |
| with SuperBeamEnv(base_url=a.url).sync() as env: |
| r = env.reset(**kw) |
| print(r.observation.screen); print(r.observation.facts) |
| n, total = 0, 0.0 |
| while not r.done: |
| r = env.step(SuperBeamAction(action=policy(r.observation))) |
| n += 1 |
| total += r.reward or 0 |
| o = r.observation |
| print(o.screen); print(o.facts) |
| print("won=%s dead=%s steps=%d frames=%d progress=%.2f reward=%.1f level=%s state=%s" |
| % (o.won, o.dead, n, o.frames, o.progress, total, o.level, env.state())) |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
