lgsvl / lanefollowing Goto Github PK

Dear Team of the LGSVL Project,

Is there an alternative to Docker? Or is it necessary to use it? Is it possible to use the cpu instead of gpu?
About us:
We are Researchers from University RWTH Aachen Germany. Me and my PhD Supervisor are working on an big autnomous driving project about self driving busses.

Best Regards,
qsengshu

I have LGSVL 3.5 running fine.

I have the repository and pulled the docker image. When I get to the 3rd line which is "docker-compose up build" I get the error shown in the image. Could you please inform me why I am getting this error.

Thank you

Hi,
I tried to build the ROS2 packages, but it failed because of a GPG error. I've read the old key has been revoked as part of the measures to deal with a recent security incident with build.ros.org.
Do you have any suggestions how to fix this?

Thanks for your help.

$ docker-compose up build
[...]

Get:17 http://archive.ubuntu.com/ubuntu bionic/universe amd64 Packages [11.3 MB]
Err:7 http://packages.ros.org/ros/ubuntu bionic InRelease
  The following signatures couldn't be verified because the public key is not available: NO_PUBKEY F42ED6FBAB17C654
Err:8 http://packages.ros.org/ros2/ubuntu bionic InRelease
  The following signatures couldn't be verified because the public key is not available: NO_PUBKEY F42ED6FBAB17C654

[...]

Reading package lists...
W: GPG error: http://packages.ros.org/ros/ubuntu bionic InRelease: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY F42ED6FBAB17C654
E: The repository 'http://packages.ros.org/ros/ubuntu bionic InRelease' is not signed.
W: GPG error: http://packages.ros.org/ros2/ubuntu bionic InRelease: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY F42ED6FBAB17C654
E: The repository 'http://packages.ros.org/ros2/ubuntu bionic InRelease' is not signed.
ERROR: Service 'build' failed to build: The command '/bin/sh -c apt-get update && apt-get install --no-install-recommends -y     ${PYTHON}-rosdep     ${PYTHON}-rosinstall     ${PYTHON}-vcstools     && rm -rf /var/lib/apt/lists/*' returned a non-zero code: 100

Dear LGSVL,

Currently our Team is working on a implementation for lanefollowing with reinforcement learning considering the following 3 tasks:
We want to reward the XE_LGSVL Vehicle if it follows the lane correctly. Later on it also gets rewarded if it avoid cars. Additionally if the vehicle collides with an object it should get automatically resetted.

Unfortunately we have problems with localizing a single lane. We recognized that the streets are complex structures for example "SanFrancisco->Environment->SF-Roads->SFRoads001"

1. Are the single lanes of a street available as gameobjects so that we can read out the so called "Transform" or position data from a single lane? Cause with that we could tell the car to stay in its lane.

2. If only the complex street structure is available: How can we get the position data from the boundaries of the streetobject?

3. Do the NPC Vehicle follow a special steering policy? For example following an invisible node structure on the lanes? Or are they just steering arbitrarily left and right during moving forward?

Why there is all NONE?
selfubuntu:~/ros2_dashing/install/rclpy/lib/python3.6/site-packages/rclpy$ ros2 run lane_following collect
[INFO] [collect]: Starting Collect node...
Directory is ready: /home/hanyh/lgsvl/AD_stack_project/ros2_src/install/lane_following/lib/python3.6/site-packages/train/data
Directory is ready: /home/hanyh/lgsvl/AD_stack_project/ros2_src/install/lane_following/lib/python3.6/site-packages/train/data/img
[INFO] [collect]: Center camera topic: None
[INFO] [collect]: Left camera topic: None
[INFO] [collect]: Right camera topic: None
[INFO] [collect]: Canbus topic: None
[INFO] [collect]: Up and running...

I want to used this simulation as data generation for deep learning project. While looking though all available node in current ROS connection there isn't any node for velocity data of ECO-vehicle. Is there any instruction to build new publishing node inside simulation to publish other type of data aside from steering angle? Or if there any editing need to be done inside the simulation script, could you guide me to it location?

Hi,Team

I was wondering how to acquire your original training set and test set used for training the model
lanefollowing/ros2_wc/src/lane_following/model/model.h5

I did see it in the repository.

Thanks

After cloning the repo and pulling the latest lgsvl Docker image, when running the set up steps:

docker-compose up build
docker-compose up drive
docker-compose up drive_visual

I get an error shown below. The build step seems to work, but not drive and drive_visual. It seems as if the error is identical for each step, occurring at line 165 of events.js. How can I fix this and start collecting data? thank you.

xiayu@xiayu-Lenovo-Legion-R7000P2021H:~/Desktop/lanefollowing$ docker-compose up drive
Starting lanefollowing_drive ... done
Attaching to lanefollowing_drive
drive_1 | terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injectorboost::system::system_error >'
drive_1 | what(): bind: Address already in use
drive_1 | /lanefollowing/scripts/drive.sh: line 14: 2729 Aborted lgsvl_bridge

Hello !!!

I recently came across the post regarding the lgsvl lanefollowing example lgsvl/simulator#698

Did someone able to set it up successfully?

I tried the same example with the latest simulator version (2020.06), I am unable to collect the data. Also when i set bridge type as ROS2 for the vehicle the bridge status is disconnected, If i choose the bridge type as ROS then the bridge is getting connected (i am using Lexus2016RXHybrid (Autoware) vehicle),But vehicle command count is not changing.

Can someone please help me in fixing this issue...

Many thanks in advance

Trying to drive lane forwarding but, no reactions in lg simulator.

I have such output in by docker compose drive:

Starting lanefollowing_drive_visual ... done
Attaching to lanefollowing_drive_visual
drive_visual_1      | (node:1301) Warning: N-API is an experimental feature and could change at any time.
drive_visual_1      | The web socket server started on ws://localhost:9090
drive_visual_1      | 2020-02-11 12:58:58.831753: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
drive_visual_1      | 2020-02-11 12:58:59.116185: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:898] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
drive_visual_1      | 2020-02-11 12:58:59.116693: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1356] Found device 0 with properties: 
drive_visual_1      | name: GeForce GTX 1050 major: 6 minor: 1 memoryClockRate(GHz): 1.493
drive_visual_1      | pciBusID: 0000:01:00.0
drive_visual_1      | totalMemory: 3.95GiB freeMemory: 1.64GiB
drive_visual_1      | 2020-02-11 12:58:59.116804: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1435] Adding visible gpu devices: 0
drive_visual_1      | 2020-02-11 12:58:59.532366: I tensorflow/core/common_runtime/gpu/gpu_device.cc:923] Device interconnect StreamExecutor with strength 1 edge matrix:
drive_visual_1      | 2020-02-11 12:58:59.532467: I tensorflow/core/common_runtime/gpu/gpu_device.cc:929]      0 
drive_visual_1      | 2020-02-11 12:58:59.532529: I tensorflow/core/common_runtime/gpu/gpu_device.cc:942] 0:   N 
drive_visual_1      | 2020-02-11 12:58:59.532668: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1053] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 1401 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1050, pci bus id: 0000:01:00.0, compute capability: 6.1)
drive_visual_1      | [INFO] [drive]: Model loaded: /lanefollowing/ros2_ws/src/lane_following/model/model.h5

In simulator I can see BridgeStatus:connected, but no activity, nothing changes.
Is it because of my GPU performance not enough? Or I should wait for some training step?
Or I should choose some another map?
I used BorregasAve (with Autoware)

When I try to execute "docker-compose up drive", I get this error: "rosbridge: command not found" Is this a problem from ROS2 "dashing" or rather lgsvl docker?

I have already contacted ros answers and received this reply: "Almost definitely an issue with their docker image. It's a wild guess, but it seems quite plausible that this is related to their latest commit, and in particular this change."

Dear LGSVL Team,

In following i got this 3 Questions and i appreciate your help, also me and my team like your work very much!

We are subscribing to following topics marked in the red box to get our images from the simulation to train our neural network.

Picture(1)

1. Question Picture(1): Can i understand the topic as a path? If yes unfortunately i can't find the path in my simulator folder.
2. Question Picture(1): If i want to subscribe to the sensors how would my topic look like? where can i look up the commands?

Picture(2)

3. Question Picture(2): Also i don't understand the step after training the neural network. The neural network will give me an steering cmd output which we are going to publish under the topic 'lanefollowing/steering_cmd'. But i cant find the part of the code where the simulation is subscribing to this topic. Where does the simulation subscribes to the predicted steering command?

Hi LGSVL team

I am trying to put a project at San Jose State university, showcasing bounding boxes. I am trying to create a docker drive container with visual and connecting it with LG simulator, so I can see how ROS2 is going to draw the lane following curve.

When I provide command docker-compose up drive_visual
everything works fine. and I can see constant prediction.

But when I give docker-compose up drive_visual, i see the following logs.

Starting lanefollowing_drive_visual ... done
Attaching to lanefollowing_drive_visual
drive_visual_1      | (node:1301) Warning: N-API is an experimental feature and could change at any time.
drive_visual_1      | The web socket server started on ws://localhost:9090
drive_visual_1      | 2019-04-26 08:02:42.072975: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
drive_visual_1      | 2019-04-26 08:02:42.205908: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:898] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
drive_visual_1      | 2019-04-26 08:02:42.206326: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1356] Found device 0 with properties: 
drive_visual_1      | name: Tesla P100-PCIE-16GB major: 6 minor: 0 memoryClockRate(GHz): 1.3285
drive_visual_1      | pciBusID: 0000:00:04.0
drive_visual_1      | totalMemory: 15.90GiB freeMemory: 12.78GiB
drive_visual_1      | 2019-04-26 08:02:42.206359: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1435] Adding visible gpu devices: 0
drive_visual_1      | 2019-04-26 08:02:42.610532: I tensorflow/core/common_runtime/gpu/gpu_device.cc:923] Device interconnect StreamExecutor with strength 1 edge matrix:
drive_visual_1      | 2019-04-26 08:02:42.610607: I tensorflow/core/common_runtime/gpu/gpu_device.cc:929]      0 
drive_visual_1      | 2019-04-26 08:02:42.610617: I tensorflow/core/common_runtime/gpu/gpu_device.cc:942] 0:   N 
drive_visual_1      | 2019-04-26 08:02:42.610968: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1053] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 12378 MB memory) -> physical GPU (device: 0, name: Tesla P100-PCIE-16GB, pci bus id: 0000:00:04.0, compute capability: 6.0)
drive_visual_1      | [INFO] [drive]: Model loaded: /lanefollowing/ros2_ws/src/lane_following/model/model.h5
drive_visual_1      | Unable to init server: Could not connect: Connection refused
drive_visual_1      | 
drive_visual_1      | (LGSVL End-to-End Lane Following:1319): Gtk-WARNING **: 08:02:45.419: cannot open display: 
lanefollowing_drive_visual exited with code 137

I get this connection refused to something, which i am unable to figure out. I can connect from LG simulator to ROS2 but i was expecting a UI screen showing up. Can someone please suggest how I am suppose to bring up another screen to see lane following curve like shown in the video ?

I have an ubuntu instance 16.04 on GCP with 1 GPU Testla P100, 16 VCpus, CUDA 9.2

Hi,

I am following the instruction to try the lanefollowing.
I cannot find the XE_Rigged-lgsvl vehicle in the lgsvl simulator.
Is there any way I can add this vehicle, or edit the code so that the available vehicles can be used?

Thanks!

I have followed the instruction https://www.lgsvlsimulator.com/docs/create-ros2-ad-stack/

And try to build the package using colcon as it states in the tutorial.

source /opt/ros/crystal/setup.bash
cd ~/ros2_ws
colcon build --symlink-install

but my package build failed with error 'str' object has no attribute 'format__map'

Could you please help me look into this?