Apollo: Carla 0.9.x Apollo bridge

[volprjir]

Hi everyone.

Is your feature request related to a problem? Please describe.
There is missing connection to Carla simulator or documentation how to do it.

Describe the solution you'd like
I would like to connect Carla with Baidu Apollo project. Solution which can use Carla ROS bridge (https://github.com/carla-simulator/ros-bridge).

Describe alternatives you've considered
There used to be some kind of bridge for older version of Carla from nemodrive (https://nemodrive-docs.readthedocs.io/projects/nemodrive-apollo/en/latest/index.html) but even this one didn't work perfectly.

Thank you

[zhongjunni]

Hi @volprjir , can you tell us the version of Apollo you used?
From Apollo 3.5.0, we have changed the dependency from ApolloROS To CyberRT.
So, the ros-bridge may not work for CyberRT.

Thank you!

[volprjir]

Hi @gruminions, to be honest, I'm not sure how to find out the version as when I click on Check Version button in Dreamview I get apolloauto/apollo:dev-x86_64-20180906_2002. But I used this repo:

https://github.com/nemodrive/apollo

I also tried to use official repo, but I'm not able to use the bridge there. Now I understand why as you mentioned CyberRT. Anyway, can you please help me how to connect the simulator with the last Apollo version, please?

Thank you

[volprjir]

@gruminions: I'm currently in touch with person from Carla development team and he will support this connection from their side. Can you please let me know where can I find some kind of expected data format or data types so we can map Carla data to Apollo? Thank you

[zhongjunni]

@gruminions: I'm currently in touch with person from Carla development team and he will support this connection from their side. Can you please let me know where can I find some kind of expected data format or data types so we can map Carla data to Apollo? Thank you

Hi, maybe you can refer to rosbag_to_record, a tool for converting data from rosbag to record.

You can know the format of most of the data from this tool.

[luqiang21]

@volprjir Hi, you can check LGSVL simulator which already supports connecting with Apollo 3.5. You can see the demo here. You can try it with their Apollo 3.5 fork.

[volprjir]

Ok, now I know that I can use cyber_monitor for listening data. If I combine this tool with cyber Apollo demo I can se channels and messages. I'm wondering how can I send data to this connection with python script. How should I start?

[volprjir]

@gruminions Is it possible to run two separate containers of 3.5 version on one computer for testing purposes? I suceeded to change ports, but it seems that both containers are still somehow connected, so if I run bootstrap.sh in one container, it tries to run bootstrap.sh in second container too (which fails)

[volprjir]

@luqiang21 Thank you for the link. I'm able to send speed or percentage of steering wheel by python script from the apollo docker . I'm just trying to run this script out of docker so I'm able to sed the commands from Carla. Can you please help me with that?

[luqiang21]

@volprjir To send commands from outside docker, maybe you need to compile cyber outside which will be complicated. I don't have successful experience on this.

[daohu527]

@luqiang21 Thanks for your help! I also want to run with carla.
@volprjir
I have run a python example outside of apollo docker. I run a script outside the docker:

python cyber/python/examples/listener.py

then start a talker inside the docker:

python cyber/python/examples/talker.py 

I can success get connected but I receive this error below:

WARNING: Logging before InitGoogleLogging() is written to STDERR
I0423 11:29:35.165530 24978 global_data.cc:153] [cyber_py] host ip: 192.168.1.102
========================================================================================================================
E0423 11:29:35.766544 25002 segment.cc:233] [cyber_py] get shm failed.
E0423 11:29:35.766624 25002 segment.cc:85] [cyber_py] init failed, can't read now.
E0423 11:29:36.767777 25002 segment.cc:233] [cyber_py] get shm failed.
E0423 11:29:36.767860 25002 segment.cc:85] [cyber_py] init failed, can't read now.
E0423 11:29:37.769208 25002 segment.cc:233] [cyber_py] get shm failed.
E0423 11:29:37.769312 25002 segment.cc:85] [cyber_py] init failed, can't read now.
E0423 11:29:38.770581 25002 segment.cc:233] [cyber_py] get shm failed.
E0423 11:29:38.770668 25002 segment.cc:85] [cyber_py] init failed, can't read now.

I will try to solve it and try again.

add:
I finaly solve the problem follow this:

1. client in docker set the CYBER_IP=docker_ip_address
2. client outside the docker set the CYBER_IP=ifconfig_address

Then it works well, and can receive message.

[fpasch]

Could we summarize the options to connect Apollo to Carla?

1. Cyber Python API (examples)
2. LGSVL Bridge

Is there any other option?

[xmyqsh]

1. [Cyber C++ API] + [Carla C++ API]
   This way should be better for debugging. We could move on Cyber Python API after this.
   Because the attempt on Cyber Python API has not been successful. #7957

2. [Cyber C++ API] + [Carla python API], the server-client way like LGSVL Bridge, with a python wrapped C++ client or server at the Carla side.

[volprjir]

@gruminions We decided to start with the bridge with Apollo 3.0 - using ROS. So we have a script which passes data from CARLA simulator to Apollo. There are speed, brake, throttle. GPS, Velodyne64 and Camera L modules are shinning green. Although we are not able to let Apollo drive car in CARLA. I'm pretty sure, there is a problem that CAN module is in UNKNOWN state. If I try to run Localization module, I'll get this in data/log/localization.out

E0502 13:22:14.038951 5268 msf_localization.cc:152] map: /apollo/modules/map/data/demo/local_map E0502 13:22:14.039369 5268 msf_localization.cc:153] lidar_extrin: modules/localization/msf/params/velodyne_params/velodyne64_novatel_extrinsics_example.yaml E0502 13:22:14.039381 5268 msf_localization.cc:154] lidar_height: modules/localization/msf/params/velodyne_params/velodyne64_height.yaml E0502 13:22:14.040738 5268 msf_localization.cc:221] Ant imu lever arm file: modules/localization/msf/params/gnss_params/ant_imu_leverarm.yaml F0502 13:22:14.044245 5268 localization_lidar.cc:58] Reflectance map folder is invalid! Check failure stack trace: [ WARN] [1556796132.973815682]: transport file: /home/tmp/ros/etc/ros/transport_mode.yaml is not map [ WARN] [1556796132.973893887]: Parse transport_mode file failed

and for perception this:

[NVBLAS] NVBLAS_CONFIG_FILE environment variable is NOT set : relying on default config filename 'nvblas.conf'
[NVBLAS] Cannot open default config file 'nvblas.conf'
[NVBLAS] Config parsed
[NVBLAS] CPU Blas library need to be provided
F0502 13:23:18.427935 9710 syncedmem.hpp:18] Check failed: error == cudaSuccess (38 vs. 0) no CUDA-capable device is detected
Check failure stack trace:
planning.ERROR
E0502 14:15:06.538234 29635 reference_line_provider.cc:604] Failed to create reference line from routing
E0502 14:15:06.538251 29635 reference_line_provider.cc:198] Fail to get reference line
E0502 14:15:06.538362 29635 pnc_map.cc:515] failed to get lane from point x: -81.769532174508768 y: 30.216548593355142 z: 0
E0502 14:15:06.539947 29635 pnc_map.cc:180] Failed to get waypoint from routing with point: (-81.7695, 30.2165, 0)
E0502 14:15:06.540004 29635 pnc_map.cc:434] Failed to update vehicle state in pnc_map
E0502 14:15:06.540022 29635 reference_line_provider.cc:550] Failed to extract segments from routing

Can you please help us how to let Apollo drive?