Motion-planned arm movement will not avoid walls in an interactive scene. Do walls have a body ID like floors that should be appended to the MotionPlanningWrapper's obstacles list?

Hello, https://github.com/StanfordVL/iGibson/blob/5f8d253694b23b41c53959774203ba5787578b74/igibson/render/mesh_renderer/mesh_renderer_cpu.py#L1390 The function get_lidar_all is not working. The camera does not turn during the 4 iterations. So the result of the readings is the same chair scene rotated 90 degrees, 4 times and patched together. I am trying to reconstruct a 360 degree scene by transforming 3d streams to the global coordianate system and patching them together but nothing is working. Please help.

Hi， there is something wrong with me，when I run roslaunch gibson2-ros turtlebot_rgbd.launch
it shows: Exception:floors.txt cannot be found in model: area1 I have downloaded the entire gibson_v2 dataset, and the area1 subset does not contain the file floors.txt. How can I get floors.txt?

Hi team, thank you for maintaining this project.

My iGibson installation went fine, but I am facing an issue that seems common among many iGibson beginners.

(igib) ➜  ~ python -m igibson.examples.environments.env_nonint_example

 _   _____  _  _
(_) / ____|(_)| |
 _ | |  __  _ | |__   ___   ___   _ __
| || | |_ || || '_ \ / __| / _ \ | '_ \
| || |__| || || |_) |\__ \| (_) || | | |
|_| \_____||_||_.__/ |___/ \___/ |_| |_|

********************************************************************************
Description:
    Creates an iGibson environment from a config file with a turtlebot in Rs (not interactive).
    It steps the environment 100 times with random actions sampled from the action space,
    using the Gym interface, resetting it 10 times.
    ********************************************************************************
INFO:igibson.render.mesh_renderer.get_available_devices:Command '['/home/mukul/iGibson/igibson/render/mesh_renderer/build/test_device', '0']' returned non-zero exit status 1.
INFO:igibson.render.mesh_renderer.get_available_devices:Device 0 is not available for rendering
WARNING:igibson.render.mesh_renderer.mesh_renderer_cpu:Device index is larger than number of devices, falling back to use 0
WARNING:igibson.render.mesh_renderer.mesh_renderer_cpu:If you have trouble using EGL, please visit our trouble shooting guideat http://svl.stanford.edu/igibson/docs/issues.html
libEGL warning: DRI2: failed to create dri screen
libEGL warning: DRI2: failed to create dri screen
ERROR: Unable to initialize EGL

I went through all the closed issues related to this, but nothing helped. I also went through the troubleshooting guide and things seemed fine to me. Here are the outputs of some commands I ran to check the EGL installation:

• (igib) ➜  ~ ldconfig -p | grep EGL
  	libEGL_nvidia.so.0 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libEGL_nvidia.so.0
  	libEGL_mesa.so.0 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libEGL_mesa.so.0
  	libEGL.so.1 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libEGL.so.1
  	libEGL.so (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libEGL.so
  

• (igib) ➜  ~ nvidia-smi
  Thu Mar 31 15:28:55 2022       
  +-----------------------------------------------------------------------------+
  | NVIDIA-SMI 510.47.03    Driver Version: 510.47.03    CUDA Version: 11.6     |
  |-------------------------------+----------------------+----------------------+
  | GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
  | Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
  |                               |                      |               MIG M. |
  |===============================+======================+======================|
  |   0  NVIDIA GeForce ...  On   | 00000000:01:00.0  On |                  N/A |
  | 41%   39C    P8    20W / 215W |    331MiB /  8192MiB |      2%      Default |
  |                               |                      |                  N/A |
  +-------------------------------+----------------------+----------------------+
  

• Reinstalling after USE_GLAD set to FALSE didn't work either.

• (base) ➜  ~ ./iGibson/igibson/render/mesh_renderer/build/query_devices
  2
  
  (base) ➜  ~ ./iGibson/igibson/render/mesh_renderer/build/test_device 0
  libEGL warning: DRI2: failed to create dri screen
  libEGL warning: DRI2: failed to create dri screen
  INFO: Unable to initialize EGL
  
  
  (base) ➜  ~ ./iGibson/igibson/render/mesh_renderer/build/test_device 1
  INFO: Loaded EGL 1.5 after reload.
  INFO: GL_VENDOR=Mesa/X.org
  INFO: GL_RENDERER=llvmpipe (LLVM 12.0.0, 256 bits)
  INFO: GL_VERSION=3.1 Mesa 21.2.6
  INFO: GL_SHADING_LANGUAGE_VERSION=1.40
  

Please let me know if I can share any more information that could be helpful in debugging this.

Thanks!

Hi,

I followed the instructions for Gibson2 installation. When I run the demo code test:

I get this error: GLSL 1.5.0 is not supported. Supported versions are ....

I did retry the installation with USE_GLAD set to FALSE in CMakeLists, but this resulted in the installation crashing.

Any ideas on the next steps I can take?

the demo is using ros1, turtlebot1 and python 2.7, which are all out of date. By using miniconda env based on python2.7, you even can not properly install igibson2!!

Thanks, your recent help was great! I am amazed by your support, thank you!

Here a few more points:

• I tried to use a different activity. Therefore I changed behavior_onboard_sensing.yaml by setting task: boxing_books_up_for_storage, but then I got an error message that ...fixed_furniture file can't be found. So I activated online_sampling. in the yaml-file. Does this randomize which objects are loaded and where they are placed?

But then I got:

Traceback (most recent call last):
  File "stable_baselines3_behavior_example.py", line 202, in <module>
    main()
  File "stable_baselines3_behavior_example.py", line 137, in main
    env = make_env(0)()
  File "stable_baselines3_behavior_example.py", line 129, in _init
    physics_timestep=1 / 300.0,
  File "/is/sg2/jmeier/behaviorchallenge/iGibson/igibson/envs/behavior_mp_env.py", line 108, in __init__
    automatic_reset=automatic_reset,
  File "/is/sg2/jmeier/behaviorchallenge/iGibson/igibson/envs/behavior_env.py", line 64, in __init__
    render_to_tensor=render_to_tensor,
  File "/is/sg2/jmeier/behaviorchallenge/iGibson/igibson/envs/igibson_env.py", line 60, in __init__
    render_to_tensor=render_to_tensor,
  File "/is/sg2/jmeier/behaviorchallenge/iGibson/igibson/envs/env_base.py", line 78, in __init__
    self.load()
  File "/is/sg2/jmeier/behaviorchallenge/iGibson/igibson/envs/behavior_env.py", line 175, in load
    self.load_task_setup()
  File "/is/sg2/jmeier/behaviorchallenge/iGibson/igibson/envs/behavior_env.py", line 164, in load_task_setup
    self.load_behavior_task_setup()
  File "/is/sg2/jmeier/behaviorchallenge/iGibson/igibson/envs/behavior_env.py", line 132, in load_behavior_task_setup
    online_sampling=online_sampling,
  File "/is/sg2/jmeier/behaviorchallenge/iGibson/igibson/activity/activity_base.py", line 92, in initialize_simulator
    self.initial_state = self.save_scene()
  File "/is/sg2/jmeier/behaviorchallenge/iGibson/igibson/activity/activity_base.py", line 98, in save_scene
    self.state_history[snapshot_id] = save_internal_states(self.simulator)
  File "/is/sg2/jmeier/behaviorchallenge/iGibson/igibson/utils/checkpoint_utils.py", line 38, in save_internal_states
    for name, obj in simulator.scene.objects_by_name.items():
AttributeError: 'NoneType' object has no attribute 'objects_by_name'

Can you help me such that I can load other activities? Do I have to take additional steps when I want to load my own activities besides placing them in bddl/activity_definitions/? Or would you recommend me to place it somewhere else?

• I would like to use the editor of behavior challenge to create a custom activity, but it seems not accessible. Can you already say, when we can use it again? https://behavior-annotations.herokuapp.com/. If the support for BehaviorChallenge's github is as quick as yours, I also don't mind to post it there ;-)

• A theoretical question: Is it possible to transport an object that can carry objects itself. e.g. is it possible to put an object into a bin and then transport the bin including the object in one hand?

• Is it possible to do all the 100 activities in the discrete action space? If so, how would I remove dust e.g.?

The Inverse Kinematics example script does not work out-of-the-box, and will error out with a message about control_freq being specified in Fetch's configuration file.

When this error is bypassed by commenting out the assertion, errors still occur. Fetch does not have a "robot_body" attribute, so

fetch.robot_body.reset_position([0, 0, 0])

should become

fetch.reset_position([0, 0, 0])

which is the standard in the functioning examples.

Similarly, it seems that

fetch.get_end_effector_position()

should become

fetch.links["gripper_link"].get_position()

RobotLink does not have a body_part_index, so

robot_id, fetch.links["gripper_link"].body_part_index, [x, y, z], threshold, maxIter

should become something like

robot_id, fetch.links["gripper_link"].(body/link)_id, [x, y, z], threshold, maxIter

After all of these changes, the example wildly flails around Fetch's arm, which I wouldn't imagine is the intended purpose of the example.

This script is fairly important for outlining the usage of IK in iGibson. If I fix it, I will submit a PR. Just wanted to outline the issue here as well.

Hi, I was trying to train pointnav agent using the given example 'stable_baselines3_example.py' but It gives me a memory error(attached). I solve this by reducing 'num_environments' from 8 to 1. But it isn't converging. I also attached the tensorboard logs. Do I need to change any other parameters (e.g. learning rate etc) to make it work with 1 environment.?

Hi @fxia22 and @ChengshuLi, I tried to download the Gibson2 Room Dataset from https://docs.google.com/forms/d/e/1FAIpQLScWlx5Z1DM1M-wTSXaa6zV8lTFkPmTHW1LqMsoCBDWsTDjBkQ/viewform, and I couldn't access the cloud storage because of the following issue.

This XML file does not appear to have any style information associated with it. The document tree is shown below. UserProjectAccountProblem User project billing account not in good standing.

Could you please check if the payment was properly made?

Hi there,

I am unable to get either docker or pip installation to run with GUI on a remote server (Ubuntu 18.04.5 LTS). nvidia-smi shows NVIDIA-SMI 450.80.02 Driver Version: 450.80.02 CUDA Version: 11.0 With a GeForce RTX 2080 SUPER

After installing docker according to these direction: https://docs.docker.com/engine/install/ubuntu/
sudo docker run hello-world runs successfully I cloned the repository

git clone [email protected]:StanfordVL/iGibson.git cd iGibson ./docker/pull-images.sh

docker images shows that I have these repositories download: igibson/igibson-gui latest f1609b44544a 6 days ago 8.11GB igibson/igibson latest e2d4fafb189b 6 days ago 7.48GB

But sudo ./docker/headless-gui/run.sh elicits this error: Starting VNC server on port 5900 with password 112358 please run "python simulator_example.py" once you see the docker command prompt: docker: Error response from daemon: could not select device driver "" with capabilities: [[gpu]].

sudo ./docker/base/run.sh also elicits: docker: Error response from daemon: could not select device driver "" with capabilities: [[gpu]].

One guess is that something is wrong with OpenGL, but I don't know how to fix it. If I run glxinfo -B, I get name of display: localhost:12.0 libGL error: No matching fbConfigs or visuals found libGL error: failed to load driver: swrast display: localhost:12 screen: 0 direct rendering: No (If you want to find out why, try setting LIBGL_DEBUG=verbose) OpenGL vendor string: Intel Inc. OpenGL renderer string: Intel(R) Iris(TM) Plus Graphics 655 OpenGL version string: 1.4 (2.1 INTEL-14.7.8)

Note: I can successfully run xeyes on the server and have it show up on my local machine. And glxgears shows the gears image but the gears are not rotating. (and returns this error: libGL error: No matching fbConfigs or visuals found libGL error: failed to load driver: swrast )

I also tried the steps from the trouble shooting page: ldconfig -p | grep EGL yields libEGL_nvidia.so.0 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libEGL_nvidia.so.0 libEGL_nvidia.so.0 (libc6) => /usr/lib/i386-linux-gnu/libEGL_nvidia.so.0 libEGL_mesa.so.0 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libEGL_mesa.so.0 libEGL.so.1 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libEGL.so.1 libEGL.so (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libEGL.so And I checked that /usr/lib/x86_64-linux-gnu/libEGL.so -> libEGL.so.1.0.0

I also do not appear to have any directories such as /usr/lib/nvidia-vvv (I only have /usr/lib/nvidia, /usr/lib/nvidia-cuda-toolkit, and /usr/lib/nvidia-visual-profiler)

Any help would be very much appreciated! Thank you so much.

The TwoWheelRobot seems to be incorrectly calculating the base angular velocity that is returned in the proprioception dictionary.

$\omega$ = angular velocity $\upsilon$ = linear velocity $V_r$ = right wheel velocity $V_l$ = left wheel velocity $R$ = wheel radius $l$ = wheel axle length

The incorrect formula can be found here and is

\omega=\frac{V_r-V_l}{l}

The equations to get wheel velocities from linear and angular velocities that are applied to a DD controller are here. These equations seem to be the source of truth that the proprioception calculation should match. The equations are the following:

V_l = \frac{\upsilon - \omega \cdot l/2}{R}

V_r = \frac{\upsilon + \omega \cdot l/2}{R}

Solving for $\omega$ and $\upsilon$ results in the following equations:

\omega = \frac{(V_l - V_r) \cdot R }{2 \cdot l}

\upsilon = \frac{(V_l + V_r) \cdot R}{2}

Ultimately, I think the angular velocity formula needs to be updated here to this $\omega = \frac{(V_l - V_r) \cdot R }{2 \cdot l}$

When running the ext_object scripts, I encountered an error in the mesh renderer. On line 1094 of mesh_renderer_cpu.py the code refers to an attribute of an InstanceGroup object called pose_rot. The actual attribute as defined in the InstanceGroup object is poses_rot. The line below is similarly effected with the pose_trans call needing to be poses_trans. My code works when I fix the typo on line 1094 but I wanted to let you know so you can fix it for others.

When I put both a Fetch robot and Behavior Robot in a VR environment (Behavior robot is the VR avatar) and have a vision sensor in environment YAML file, I get the issue below. I believe this may be a bug in mesh_renderer_cpu.py where it tries to get RGB data for all robots in the scene and fails when it reaches the Behavior Robot. I think it needs to skip the Behavior Robots. Is this in fact a bug or an issue on my end? Thanks.

Traceback (most recent call last):
  File "main.py", line 79, in <module>
    main()
  File "main.py", line 73, in main
    state, reward, done, _ = env.step(action)
  File "C:\Users\icaro\513-final-project\igibson\envs\igibson_env.py", line 360, in step
    state = self.get_state()
  File "C:\Users\icaro\513-final-project\igibson\envs\igibson_env.py", line 279, in get_state
    vision_obs = self.sensors["vision"].get_obs(self)
  File "C:\Users\icaro\513-final-project\igibson\sensors\vision_sensor.py", line 155, in get_obs
    raw_vision_obs = env.simulator.renderer.render_robot_cameras(modes=self.raw_modalities)
  File "C:\Users\icaro\513-final-project\igibson\render\mesh_renderer\mesh_renderer_cpu.py", line 1256, in render_robot_cameras
    frames.extend(self.render_single_robot_camera(robot, modes=modes, cache=cache))
  File "C:\Users\icaro\513-final-project\igibson\render\mesh_renderer\mesh_renderer_cpu.py", line 1270, in render_single_robot_camera
    for item in self.render(modes=modes, hidden=hide_instances):
TypeError: 'NoneType' object is not iterable

As show in above picture, I want use the example code‘ igibson/examples/demo/robot_example.py.’ to achieve the four robots will have a fun cocktail party. But I want replace one as a quadrotor, then I didn't found the quadrotor's yaml file in /igibson/examples/configs. What shoud I do in the next step?

The robot misaligns the body with the hands and head when the user_tracked_body parameter is false. Also, the body falls as if it is disconnected from the hands and head. Neither happens when user_tracked_body is true. The picture attached shows how the robot is rendered in the beginning. Do you know why this may be the case or is it a bug?

I am trying to have the robot move about the space using an Oculus joystick, so I assume that setting this parameter to false is required.