huangfu/InteractionImitation
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

adding options for regularization and relative state masking via interaction graphs during data processing and experiment running. found 0.002 regularization on actions gives up to 3m of deviation with no collisions. added shell script to run ray experiments overnight

Browse Source
This commit is contained in:
Arec
2021-08-02 14:38:06 -07:00
parent fb91ee1a62
commit 6e524cf4b5
4 changed files with 66 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
src/main.py
View File

@@ -59,14 +59,14 @@ def main(config, method='bc', train=False, test=False, loc=0, datadir='./expert_
# simulate policy
track = 4
simulate_policy(policy, loc=loc, track=track, filestr=filestr, nframes=kwargs['nframes'])
simulate_policy(policy, loc=loc, track=track, filestr=filestr, nframes=kwargs['nframes'], graph=kwargs['graph'])
# run test metrics
test_dataset = InteractionDatasetSingleAgent(output_dir=datadir, loc=loc, tracks=[track])
metrics(filestr, test_dataset, policy)
def simulate_policy(policy, loc=0, track=0, filestr='', nframes=float('inf')):
def simulate_policy(policy, loc=0, track=0, filestr='', nframes=float('inf'), graph=None):
"""
Simulate a trained policy
Args:
@@ -79,8 +79,12 @@ def simulate_policy(policy, loc=0, track=0, filestr='', nframes=float('inf')):
basepath = os.path.abspath('./InteractionSimulator')
svt, svt_path = get_svt(base=basepath, loc=loc, track=track)
osm = get_map_path(base=basepath, loc=loc)
env = gym.make('intersim:intersim-v0', svt=svt, map_path=osm,
min_acc=-np.inf, max_acc=np.inf)
if graph:
env = gym.make('intersim:intersim-v0', svt=svt, map_path=osm,
min_acc=-np.inf, max_acc=np.inf, graph=graph, mask_relstate=True)
else:
env = gym.make('intersim:intersim-v0', svt=svt, map_path=osm,
min_acc=-np.inf, max_acc=np.inf)
# env = gym.make('intersim:intersim-v0', loc=loc, track=track,
# min_acc=-np.inf, max_acc=np.inf)