We've been trying to make the model able to respond like any human drivers when driving on the tracks. However, even if we analysed the baseline solution, there were totally nothing in the action function that makes the model respond to environment observation changes, only the TTC can access the environment observation function's fields without errors. This action function made the model act in the exact behavior like how the action function did in the original policy script, which results in a blind car steering in a rude manner randomly on the road without being able to drive at an appropriate speed and respond to social vehicles and sharp turns properly as soon as our trainings begin. Even if we try to define new functions to make the model respond against distances before social cars and sharp turns, all of our attempts ended up in constant errors which block the training process from actually starting up. Any modification changes we made in the action function resulted in errors in the same manner. @HuaweiUK We've e-mailed you the zip file containing my errors already. We are needing you to resolve our issues and we are looking forward to your reply fairly soon.
Posted by: Team115 @ Dec. 12, 2019, 11:39 a.m.@HuaweiUK we've e-mailed you our error files due to the restriction on posting attachments on the forum. Please reply in accordance to it. Thank you.
Posted by: Team115 @ Dec. 12, 2019, 1:32 p.m.