While implementing algorithm in new data set -thor-201706291201-Linux64 (discrete mode)the move commands are not working, the robot perception is remains unchanged while inputting move commands. But Rotation commands,Look UP and Down is working at the same time .
Please help me out
Posted by: telephus @ July 11, 2017, 8:06 p.m.Here is a snippet of code that will move the agent around randomly in the discrete scenes:
env = robosims.controller.ChallengeController(
unity_path="thor-201706291201-OSXIntel64.app/Contents/MacOS/thor-201706291201-OSXIntel64"
)
env.start()
output = []
with open("targets-test.json") as f:
t = json.loads(f.read())
for target in t:
env.initialize_target(target)
for i in range(100):
env.step(action=dict(action=random.choice(['MoveLeft', 'MoveRight', 'MoveAhead', 'MoveBack'])))
Can you try running this or compare it to what you are running to ensure you are working with the correct targets/scenes?
Posted by: AI2_admin @ July 14, 2017, 6:58 p.m.I managed to solve the problem of data set,
But unfortunately while I run in new data set , the target_found() function not returning True value while the robot reaches the target.
I used thor-201706291201 with thor-challenge-targets-test file in Mac OS
I have uploaded the result using thor-201705011400-OSXIntel64
with thor-challenge-targets
Could you please help me out here
Posted by: telephus @ July 15, 2017, 8:35 p.m.I managed to solve the problem of data set,
But unfortunately while I run in new data set , the target_found() function not returning True value while the robot reaches the target.
I used thor-201706291201 with thor-challenge-targets-test file in Mac OS
I have uploaded the result using thor-201705011400-OSXIntel64
with thor-challenge-targets
Could you please help me out here
Posted by: telephus @ July 15, 2017, 8:36 p.m.To make the test set blind, we do not return True when the target is found.
Posted by: AI2_admin @ July 16, 2017, 3:06 p.m.