Rover gets stuck in infinite reset loop in inspect world, cannot move at all

[alexleel]

When I run this command:
python3 sb3_SAC.py --load False --mode train --world inspect --vision False
The robot does not move at all.

1. ros2 launch roverrobotics_gazebo 4wd_rover_gazebo.launch.py
2. $python3 ign_ros2_Nav2_topics.py inspect rover_zero4wd
3. python3 sb3_SAC.py --mode train --load False --world inspect
   The output is: $python3 sb3_SAC.py --mode train --load False --world inspect
   Using standard lidar model
   world is inspect
   /home/wruser/.local/lib/python3.10/site-packages/gymnasium/spaces/box.py:235: UserWarning: WARN: Box low's precision lowered by casting to float32, current low.dtype=float64
   gym.logger.warn(
   /home/wruser/.local/lib/python3.10/site-packages/gymnasium/spaces/box.py:305: UserWarning: WARN: Box high's precision lowered by casting to float32, current high.dtype=float64
   gym.logger.warn(
   [WARN] [1774530610.705180763] [turtlebot_controller]: No actual robot detected. Running in simulation mode.
   Creating new model
   Using cpu device
   Starting training mode
   ################inspect Environment Reset

Generated heading: 18.712741018598567°

New target x,y: -20.35, -23.47
1774530624.4372308
Logging to ./tboard_logs/SAC_inspect_20260326_2109/SAC_1

Please help to check

[jackvice]

can you check if cmd_vel is being published?

[alexleel]

OK , I can check it later, seems some python script changed and I can not run it as README.md ,can you please help to check if the command i run is OK?

[alexleel]

Hi Jackvice,

I have checked the below is the ros2 topic list when I run ros2 launch roverrobotics_gazebo 4wd_rover_gazebo.launch.py:
$ros2 topic list
/camera/image_raw
/camera_info
/clock
/cmd_vel
/imu/data
/joint_states
/odometry/wheels
/parameter_events
/rosout
/scan
/tf
/world/default/dynamic_pose/info

Howerver , no actual /cmd_vel published when I run python3 sb3_SAC.py --mode train --load False --world inspect:

Generated heading: 146.87016123046877°
New target x,y: -22.92, -21.53
1774749862.5863094
| rollout/ | |
| ep_len_mean | 1 |
| ep_rew_mean | -10 |
| time/ | |
| episodes | 20 |
| fps | 0 |
| time_elapsed | 275 |
| total_timesteps | 20 |
too far, x, y is 0.0 0.0 , episode done. ************** reward -10

Generated heading: -148.63371837403741°

New target x,y: -23.17, -22.40
1774749875.7032788
too far, x, y is 0.0 0.0 , episode done. ************** reward -10
inspect Environment Reset

Generated heading: -43.433753366237084°

New target x,y: -26.34, -19.15
1774749888.8283353
too far, x, y is 0.0 0.0 , episode done. ************** reward -10

Generated heading: -107.6070401332865°

New target x,y: -23.07, -22.22
1774749901.9403417
too far, x, y is 0.0 0.0 , episode done. ************** reward -10
..
$ros2 topic echo /cmd_vel

linear:
x: 0.0
y: 0.0
z: 0.0
angular:
x: 0.0
y: 0.0
z: 0.0

linear:
x: 0.0
y: 0.0
z: 0.0
angular:
x: 0.0
y: 0.0
z: 0.0

linear:
x: 0.0
y: 0.0
z: 0.0
angular:
x: 0.0
y: 0.0
z: 0.0

There are some issues shown in the Gazebo simulation—the robot still won't move. Could you please help take a look? Could it be that because the code was modified, the way to run the Python scripts has also changed accordingly?

Thanks,
alex

[jackvice]

Looks like SAC is not sending drive commands. And yes, the code has been modify a number of times for my ongoing dissertation work. and so it might be best to pull an older commit from like 3 or 4 months ago. Unfortunately, I am on work travel and so I wont be able to try reproducing your results until Sunday April 5th.

[jackvice]

see Active_Vision_README.md for latest methods.