search

duckietown / gym-duckietown

Self-driving car simulator for the Duckietown universe
http://duckietown.org
Other
52 stars 19 forks source link

The simulator's position is too small when testing with RL algorithm. #204

Closed zhaoworking closed 4 years ago

zhaoworking commented 4 years ago

timestep: 487 | reward: 2.9815825234448745 INFO:gym-duckietown:Pos: [0.16776178 0. 0.29571088] angle 4.266275815950335 INFO:gym-duckietown:Pos: [0.18177636 0. 0.27306946] angle 4.049985441742995 timestep: 488 | reward: 3.515600246424384 INFO:gym-duckietown:Pos: [0.18177636 0. 0.27306946] angle 4.049985441742995 INFO:gym-duckietown:Pos: [0.20296474 0. 0.24407919] angle 4.112490319974118 timestep: 489 | reward: 3.9223733389333497 INFO:gym-duckietown:Pos: [0.20296474 0. 0.24407919] angle 4.112490319974118 INFO:gym-duckietown:Pos: [0.2160342 0. 0.22504895] angle 4.10870028266937 timestep: 490 | reward: 3.8976755001671903 INFO:gym-duckietown:Pos: [0.2160342 0. 0.22504895] angle 4.10870028266937 INFO:gym-duckietown:Pos: [0.20268746 0. 0.24788006] angle 4.258060757087562 timestep: 491 | reward: 3.624382276087298 INFO:gym-duckietown:Pos: [0.20268746 0. 0.24788006] angle 4.258060757087562 INFO:gym-duckietown:Pos: [0.18788337 0. 0.27996628] angle 4.3021555331039965 timestep: 492 | reward: 3.378131035184803 INFO:gym-duckietown:Pos: [0.18788337 0. 0.27996628] angle 4.3021555331039965 INFO:gym-duckietown:Pos: [0.17724695 0. 0.30111433] angle 4.190612199495656 timestep: 493 | reward: 3.0978186555579086 INFO:gym-duckietown:Pos: [0.17724695 0. 0.30111433] angle 4.190612199495656 INFO:gym-duckietown:Pos: [0.18037936 0. 0.29412895] angle 4.391082117996021 timestep: 494 | reward: 2.886240731778871 INFO:gym-duckietown:Pos: [0.18037936 0. 0.29412895] angle 4.391082117996021 INFO:gym-duckietown:Pos: [0.18032057 0. 0.29688446] angle 4.9910317739651795 timestep: 495 | reward: 2.896586904763131 INFO:gym-duckietown:Pos: [0.18032057 0. 0.29688446] angle 4.9910317739651795 INFO:gym-duckietown:Pos: [0.1788154 0. 0.29126345] angle 4.957022812380346 timestep: 496 | reward: 2.8599158783976595 INFO:gym-duckietown:Pos: [0.1788154 0. 0.29126345] angle 4.957022812380346 INFO:gym-duckietown:Pos: [0.17792745 0. 0.27865244] angle 4.608344250004112 timestep: 497 | reward: 3.0078651414030046 INFO:gym-duckietown:Pos: [0.17792745 0. 0.27865244] angle 4.608344250004112 INFO:gym-duckietown:Pos: [0.17331981 0. 0.29421553] angle 4.240753406541636 timestep: 498 | reward: 3.0340761428067067 INFO:gym-duckietown:Pos: [0.17331981 0. 0.29421553] angle 4.240753406541636 INFO:gym-duckietown:Pos: [0.17084038 0. 0.30560628] angle 4.755370910399716 timestep: 499 | reward: 2.976090021658111 Why my postion is so small until the end of the tset ? I tested with this file.

zhaoworking commented 4 years ago

@AndreaCensi

zhaoworking commented 4 years ago

why my speed is 0 ? it is spinning aroud all the time.

AndreaCensi commented 4 years ago

@CourchesneA ?

bhairavmehta95 commented 4 years ago

This might just be an RL thing, where the agent hasn't trained enough yet.

zhaoworking commented 4 years ago

@CourchesneA ?

INFO:gym-duckietown:Pos: [0.10907588 0. 0.08145748] angle -1.4140188871210604 INFO:gym-duckietown:Pos: [0.1077757 0. 0.06491242] angle -1.5707276305686322 timestep: 1229 | reward: 0.5541583695478534 INFO:gym-duckietown:Pos: [0.1077757 0. 0.06491242] angle -1.5707276305686322 INFO:gym-duckietown:Pos: [0.1072284 0. 0.06348076] angle -0.8405789910143403 DEBUG:gym-duckietown:No tile found at [ 0.14709685 0. -0.00442732] (0, -1) DEBUG:gym-duckietown:Invalid pose. Collision free: True On drivable area: False DEBUG:gym-duckietown:safety_factor: 1.0 DEBUG:gym-duckietown:pos: [0.09121964 0. 0.04560005] DEBUG:gym-duckietown:l_pos: [ 0.14709685 0. -0.00442732] DEBUG:gym-duckietown:r_pos: [0.03534243 0. 0.09562742] DEBUG:gym-duckietown:f_pos: [0.15125248 0. 0.1126527 ] INFO:gym-duckietown:Stopping the simulator because we are at an invalid pose. timestep: 1230 | reward: -10 INFO:gym-duckietown:Pos: [0.1072284 0. 0.06348076] angle -0.8405789910143403 Total T: 1230 Episode Num: 8 Episode T: 82 Reward: 388.986034 DEBUG:gym-duckietown:No tile found at [0.2196841 0. 0.59513809] (0, 1) DEBUG:gym-duckietown:Invalid pose. Collision free: True On drivable area: False

And I always encounter this interruption , leaving the simulator stopped. Then i have to wait for a long time for the next training procedure.

zhaoworking commented 4 years ago

This might just be an RL thing, where the agent hasn't trained enough yet.

oh , i see. And could u take a look at the interruption problem?

bhairavmehta95 commented 4 years ago

It's not an interruption, it's the training of the algorithm that takes a long time, especially if you're working on a laptop or CPU based hardware. Not much we can do about that.

zhaoworking commented 4 years ago

It's not an interruption, it's the training of the algorithm that takes a long time, especially if you're working on a laptop or CPU based hardware. Not much we can do about that.

Does it mean that i need a GPU on my laptop , and then it will process faster?