facebookresearch / spot-sim2real

Spot Sim2Real Infrastructure
MIT License
83 stars 7 forks source link

Phantom pick at the end of ASC demo when robot is supposed to go to dock #21

Open KavitShah1998 opened 1 year ago

KavitShah1998 commented 1 year ago

We saw this several times during RL-R demo and also during tests inside apartment space in Fremont.

Robot on finishing all tasks from waypoints.yaml tries to navigate to the base. While doing so if it sees any detections even from a non-registered receptacle, it does and picks the object (from object_targets inside waypoints.yaml) and takes it to its desired location (again from object_targets inside waypoints.yaml).

On reaching this desired location, it thinks that it reached the dock & sits down there itself.

Messages from log & attached image are evidence for this. The chair on which robot detects a dinosour as a "penguin" is not mapped as a receptable at all, but spotty does pick it nevertheless.

policy_inference: 0.0408 run_actions: 0.8101 get_observations: 0.0008 total: 0.8517
[base_env.py]: Saying: Finished object rearrangement. Heading to dock.
open gripper called in SpotMobileManipulationBaseEnv

...

[bounding_box]: Detected: ball, penguin
policy_inference: 0.0361 run_actions: 0.5098 get_observations: 0.0008 total: 0.5467
gater: gaze corrective: 0.047, -0.025
raw_base_ac: [0.00, -0.00]  raw_arm_ac: [-0.94, 0.90, 0.99, 0.86]
base_action: None   arm_action: [-0.25, -2.56, 2.02, 0.00, 1.57, 0.00]
DETECTIONS STR
penguin,0.09405618906021118,197,120,249,177
[bounding_box]: Detected: penguin
Locked on to target 1 time(s)...
policy_inference: 0.0043 run_actions: 0.5114 get_observations: 0.0012 total: 0.5169
gater: gaze corrective: 0.062, 0.091
raw_base_ac: [0.00, -0.00]  raw_arm_ac: [-0.34, 0.97, 0.51, -0.98]
base_action: None   arm_action: [-0.28, -2.48, 2.07, 0.00, 1.49, 0.00]
DETECTIONS STR
penguin,0.19008836150169373,175,68,226,133
[bounding_box]: Detected: penguin
Locked on to target 2 time(s)...
policy_inference: 0.0759 run_actions: 0.5098 get_observations: 0.0010 total: 0.5867
gater: gaze corrective: -0.015, -0.021
raw_base_ac: [0.00, -0.00]  raw_arm_ac: [0.33, 0.98, -0.58, -1.00]
base_action: None   arm_action: [-0.25, -2.40, 2.02, 0.00, 1.40, 0.00]
DETECTIONS STR
penguin,0.33202460408210754,144,51,210,122
[bounding_box]: Detected: penguin
Locked on to target 3 time(s)...
policy_inference: 0.0192 run_actions: 0.5111 get_observations: 0.0009 total: 0.5312
gater: gaze corrective: -0.009, 0.099
raw_base_ac: [0.00, -0.00]  raw_arm_ac: [0.77, 0.98, -0.76, -1.00]
DETECTIONS STR
penguin,0.36048218607902527,148,37,203,113
[bounding_box]: Detected: penguin
Saved grasp image as /home/kavitshah/fair/spot-sim2real/spot_rl_experiments/spot_rl/grasp_visualizations/1685142245.4967258.png
Lost lock-on!
GRASP CALLED: Aiming at (x, y): (250, 106)!
[base_env.py]: Saying: Grasping penguin
Current grasp_point_in_image state:  MANIP_STATE_SEARCHING_FOR_GRASP
Current grasp_point_in_image state:  MANIP_STATE_SEARCHING_FOR_GRASP
Current grasp_point_in_image state:  MANIP_STATE_SEARCHING_FOR_GRASP
Current grasp_point_in_image state:  MANIP_STATE_SEARCHING_FOR_GRASP
Current grasp_point_in_image state:  MANIP_STATE_MOVING_TO_GRASP
Current grasp_point_in_image state:  MANIP_STATE_MOVING_TO_GRASP
Current grasp_point_in_image state:  MANIP_STATE_MOVING_TO_GRASP
Current grasp_point_in_image state:  MANIP_STATE_MOVING_TO_GRASP
Current grasp_point_in_image state:  MANIP_STATE_MOVING_TO_GRASP
Current grasp_point_in_image state:  MANIP_STATE_MOVING_TO_GRASP
Current grasp_point_in_image state:  MANIP_STATE_MOVING_TO_GRASP
Current grasp_point_in_image state:  MANIP_STATE_MOVING_TO_GRASP
Current grasp_point_in_image state:  MANIP_STATE_MOVING_TO_GRASP
Current grasp_point_in_image state:  MANIP_STATE_GRASP_SUCCEEDED
base_action: None   arm_action: [0.07, -2.76, 2.02, 0.00, 1.22, 0.00]
[base_env.py]: Saying: Navigating to kitchen_sink

...

base_action: [-0.05, 0.00, -0.51]   arm_action: None
policy_inference: 0.0095 run_actions: 0.5094 get_observations: 0.0140 total: 0.5328
gater: nav  corrective: 0.007, 0.011, -0.015, 0.029
raw_base_ac: [0.11, -0.94]  raw_arm_ac: [0.00, 0.00, -0.00, -0.00]
base_action: [0.06, 0.00, -0.49]    arm_action: None
[base_env.py]: Saying: Executing automatic docking
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
Dock not found... trying again
/home/kavitshah/miniconda3/envs/spot_ros/lib/python3.8/site-packages/rospy/core.py:146: DeprecationWarning: The 'warn' method is deprecated, use 'warning' instead
  _rospy_logger.warn(msg, *args, **kwargs)
/home/kavitshah/miniconda3/envs/spot_ros/lib/python3.8/site-packages/rospy/core.py:598: DeprecationWarning: isAlive() is deprecated, use is_alive() instead
  if t.isAlive():

/home/kavitshah/fair/spot-sim2real/spot_rl_experiments/spot_rl/grasp_visualizations/1685142245.4967258.png

KavitShah1998 commented 1 year ago

1685142245 4967258

@SergioArnaud & @jimmytyyang know about this issue.

@akshararai have you seen this issue before?