Mapping Unreal/Airsim Environment using ROS packages

[thejeshk]

Hello Guys,

Is there any possible way to map the custom Unreal/Airsim environment using ROS packages?

Thank you :)
Thejesh

[manthan99]

ORB SLAM works fine with Airsim if you want to create a feature-based map using monocular camera data. The calibration parameters are pretty straight forward. Since you already have the ground truth odometry, you may try the common ROS mapping packages.
Also have a look at these issues-
#2369 #2396

[Yueting-Li]

Hi,
I am running px4 sitl. And I want to use ORB SLAM to map Unreal envionment. I start airsim_ros_wrapper node to get images from unreal envionment, then I use xbox to control drone but unfortunately the uav suffers great instability. If I close airsim node, the drone can return to stable state. Can anybody give me some suggestions?
Thank you!

[manthan99]

Hi @Yueting-Li
Try changing the flight mode to position control mode using the QGround Control App. It will be much more stable than the default stabilized flight mode

[Yueting-Li]

Thank you! @manthan99
Actually I have changed to position control mode. If I just run px4, everything is ok. But if I run airsim_node.launch in the meantime, the drone would fly without control. It confused me for a long while.

[manthan99]

@Yueting-Li , What settings.json are you using? Have you changed the settings according to PX4 instead of simple Flight controller?

[Yueting-Li]

@manthan99,
My settings.json is as follows:
{
"SeeDocsAt": "https://github.com/Microsoft/AirSim/blob/master/docs/settings.md",
"SettingsVersion": 1.2,
"SimMode": "Multirotor",
"Vehicles": {
"mav": {
"VehicleType": "PX4Multirotor",
"UseSerial": false,
"LocalHostIp": "192.168.1.1",
"SitlIp": "192.168.1.2",
"SitlPort": 14556,
"UdpIp": "192.168.1.2",
"UdpPort": 14560,
"QgcHostIp": "192.168.1.2",
"QgcPort": 14550,
"params": {
"NAV_RCL_ACT": 0,
"NAV_DLL_ACT": 0,
"LPE_LAT": 47.641468,
"LPE_LON": -122.140165,
"COM_OBL_ACT": 1
},
"Sensors": {
"Barometer": {
"SensorType": 1,
"Enabled": true
},
"Imu": {
"SensorType": 2,
"Enabled": true
},
"Gps": {
"SensorType": 3,
"Enabled": true
},
"Magnetometer": {
"SensorType": 4,
"Enabled": true
},
"Distance": {
"SensorType": 5,
"Enabled": true
},
"LidarCustom": {
"SensorType": 6,
"Enabled": true,
"NumberOfChannels": 16,
"PointsPerSecond": 10000,
"X": 0,
"Y": 0,
"Z": -1,
"DrawDebugPoints": true
}
},
"Cameras": {
"front_center_custom": {
"CaptureSettings": [
{
"PublishToRos": 1,
"ImageType": 0,
"Width": 672,
"Height": 376,
"FOV_Degrees": 90,
"TargetGamma": 1.5
},
{
"PublishToRos": 1,
"ImageType": 2,
"Width": 672,
"Height": 376,
"FOV_Degrees": 90,
"TargetGamma": 1.5
}
],
"X": 0.50,
"Y": 0,
"Z": 0.10,
"Pitch": 0.0,
"Roll": 0.0,
"Yaw": 0.0
}
},
"X": 2,
"Y": 0,
"Z": 0,
"Pitch": 0,
"Roll": 0,
"Yaw": 0
}
},
"SubWindows": [
{"WindowID": 0, "ImageType": 0, "CameraName": "front_center_custom", "Visible": true},
{"WindowID": 1, "ImageType": 3, "CameraName": "front_center_custom", "Visible": true},
{"WindowID": 2, "ImageType": 5, "CameraName": "front_center_custom", "Visible": true}
]
}

PS: "192.168.1.1" is the ip address where I run airsim&UE4(on windows), "192.168.1.2" is the ip address where I run PX4 & ROS(on ubuntu).

[manthan99]

@Yueting-Li Sorry for the late reply. Seems fine. I was using almost similar settings and was getting a quite stable flight in the position-controlled mode. I assume that you are using the "airsim_node.launch" file and not the other launch files (since some of them are for controlling the drone using ros commands)
Let me know if you find a solution

[catproof]

does anyone in this thread have experience running active SLAM in airsim using ROS? as in, using a SLAM algorithm to decide what the next control input is for flying the drone based off the current state estimate? I am wondering how feasible it is. there aren't any tutorials or guides on how to do this, seems like there is a lot of troubleshooting and scavenger-hunting awaiting me if I embark on this task.

[xxEoD2242]

@NickPerezCarletonUniversity that is what we are currently doing. It is possible but there is a lot you will need to do that is not in the guide. Can discuss offline if you'd like.

[catproof]

that would be awesome. send me a message on linkedin?... https://www.linkedin.com/in/nicolas-perez-is-awesome/

[catproof]

ORB SLAM works fine with Airsim if you want to create a feature-based map using monocular camera data. The calibration parameters are pretty straight forward. Since you already have the ground truth odometry, you may try the common ROS mapping packages. Also have a look at these issues- #2369 #2396

how do you set up the config file for mono-inertial in Orb Slam? I figured out how to do it for just mono, but I'm not sure how to populate the imu fields in the yaml file for mono-inertial.

the settings for imu in airsim have different names:

"Imu": {
    "SensorType": 2,
    "Enabled" : true,
    "AngularRandomWalk": 0.3,
    "GyroBiasStabilityTau": 500,
    "GyroBiasStability": 4.6,
    "VelocityRandomWalk": 0.24,
    "AccelBiasStabilityTau": 800,
    "AccelBiasStability": 36

compare to the settings in the yaml file for orb slam:

https://github.com/UZ-SLAMLab/ORB_SLAM3/blob/master/Examples/Monocular-Inertial/EuRoC.yaml

[catproof]

@manthan99 tagging you so you can see my last comment

[Erickrk]

Hello everyone, any updates on this?