Gamenot
commented
1 year ago Hello @Edward11235, to recount, I believe your are saying the following is true:
- "car-E0_0_0-E0_0_max--59446--11620-0-0.3" is created at 22.4s
- "car-E0_0_0-E0_0_max--59446--11620-0-0.3" is taken over by the ego at 22.4s (after 224 steps from 1 call to
reset()until an ego agent enters into the simulation) - Ego has control of "car-E0_0_0-E0_0_max--59446--11620-0-0.3" for 100s (1000 steps, 1000 calls to
step()) - Total elapsed simulation time is 122.4s
- Agent's elapsed simulation time is 100s
If this is the behaviour you are seeing, it is intended. We have to simulate the time leading up to the ego's entry into the simulation. You can reduce this if you wish by checking against env.smarts.step_count which will give you the steps taken.
Q
In terms of usability, is there some observation that you think should be providing the total simulation time/steps or be clarified in name/documentation?
High Level Description
For each iteation, I have 1000 iterations, which means that I call step() method 1000 times and my episode should be 100s given that my environment runs at 10 frames per second. On my scenario.py, when I define Mission for ego_mission, I set entry_tactic=IdEntryTactic(start_time=0, actor_id="car-E0_0_0-E0_0_max--59446--11620-0-0.3". The traffic vehicle with actor_id="car-E0_0_0-E0_0_max--59446--11620-0-0.3" is created at 22.4 seconds. After I 'hijack' this car, my episode runs for 122.4 (= 100 + 22.4) seconds, which means that time is reset to zero when I take control of this car. This is not supposed to happen.
Version
I am using master branch.
Steps to reproduce the bug
No response
System info
Ubuntu 20.04, Python 3.8
Error logs and screenshots
No response
Impact (If known)
No response