GPULiDAR negative FOV start bug

[reza-shahriari]

Hi,
this is my settings.json file:

{
  "SettingsVersion": 2.0,
  "SimMode": "ComputerVision",
  "Vehicles": {
      "CVSensor": {
          "VehicleType": "ComputerVision",
          "AutoCreate": true,
          "Sensors": {
            "gpulidar1": {
              "SensorType": 8,
              "Enabled" : true
          }
          }
      }
  }
}

when I run the game I see this error:

changing settings and adding more details to GPU lidar (as mentioned here) not change the problem!
even using Car simulation leads to the same problem. when I Disable(Enabled: false) the GPU lidar sensor everything works perfectly

[WouterJansen]

Can you share you settings.json file please? It seems to be tring to load a material CSV file for the intensity generation of GPU LiDAR and something is going wrong there.

[reza-shahriari]

Can you share you settings.json file please?

this is my setting.json file:
{
"SettingsVersion": 2.0,
"SimMode": "ComputerVision",
"Vehicles": {
"CVSensor": {
"VehicleType": "ComputerVision",
"AutoCreate": true,
"Sensors": {
"gpulidar1": {
"SensorType": 8,
"Enabled" : true
}
}
}
}
}

for material.csv I have an empty csv file in the Documents/airsim folder

[WouterJansen]

remove the material.csv file or make sure it is not empty (or contains an empty line) and this will be no longer crashing. I will make a catch for this for next release.

[reza-shahriari]

Thanks.
deleting the material.csv solved my problem

[reza-shahriari]

The crash problem solved and its not crashing now but lidar GPU results is almost always just a point!

    def parse_gpulidarData(self):
        points = np.array(self.data.point_cloud, dtype=np.dtype('f4'))
        points = np.reshape(points, (int(points.shape[0] / 5), 5))
        xyz = points[:, 0:3]
        self.points = xyz

def visualize(self):
        print("visualizing data ...")
        
       # Convert NED (AirSim) to ENU coordinate system
        x = self.points[:, 0]
        y = -self.points[:, 1]  # Flip Y axis
        z = -self.points[:, 2]  # Flip Z axis

        # Verify the number of points
        print(f"Total points: {x.shape[0]}")  

        # RGB extraction (assuming the rgb array is already defined)
        # Normalize RGB values to the [0, 1] range for matplotlib
        rgb_normalized = self.rgb / 255.0

        # Plotting the Lidar data with colors
        fig = plt.figure()
        ax = fig.add_subplot(111, projection='3d')

        # Scatter plot with colors
        scatter = ax.scatter(x, y, z, c=rgb_normalized, marker='o')

        # Set labels
        ax.set_xlabel('X')
        ax.set_ylabel('Y')
        ax.set_zlabel('Z')

        # # Adjust the aspect ratio for better visualization
        # max_range = np.array([x.max() - x.min(), y.max() - y.min(), z.max() - z.min()]).max() / 2.0
        # mid_x = (x.max() + x.min()) * 0.5
        # mid_y = (y.max() + y.min()) * 0.5
        # mid_z = (z.max() + z.min()) * 0.5

        # ax.set_xlim(mid_x - max_range, mid_x + max_range)
        # ax.set_ylim(mid_y - max_range, mid_y + max_range)
        # ax.set_zlim(mid_z - max_range, mid_z + max_range)

        # Display the plot
        plt.show()

visualizing data ...
Total points: 32768

[reza-shahriari]

settings file:

{
  "SettingsVersion": 2.0,
  "SimMode": "ComputerVision",
  "Vehicles": {
      "Drone1": {
          "VehicleType": "ComputerVision",
          "AutoCreate": true,
          "Sensors": {
            "gpulidar1": {
            "SensorType": 8,
            "Enabled" : true,
            "Range":150,
            "NumberOfChannels": 16,
            "RotationsPerSecond": 10,
            "PointsPerSecond": 100000,
            "X": 0, "Y": 0, "Z": -1,
            "Roll": 0, "Pitch": 0, "Yaw" : 0,
            "VerticalFOVUpper": -15,
            "VerticalFOVLower": -25,
            "HorizontalFOVStart": -20,
            "HorizontalFOVEnd": 20,
            "DrawDebugPoints": true
          }
          }
      }
  }
}

[WouterJansen]

Hi, I get the same results. One thing, RotationsPerSecond is not an argument for the LiDARs in Cosys-AirSim. We use MeasurementsPerCycle to define the horizontal sample density. See the documentation on the LiDAR and GPULiDAR for all the arguments that can be used.

It seems you are running into an edge case that is causing a bug when to reset the pointcloud data. This is due to you using a negative HorizontalFOVStart argument. I will fix this in a future release. For now you can either switch to the normal LiDAR sensor which does not have this bug or do a little change to your settings to prevent having to use a negative HorizontalFOVStart. For example in your case you want to have a 40 degrees FOV from -20 degrees to 20 degrees, you have to set it to Yaw : -20, HorizontalFOVStart: 0 and_HorizontalFOVEnd_: 40.

[reza-shahriari]

Thanks for the answer.
changing setting to this worked nice for me:

{
  "SettingsVersion": 2.0,
  "SimMode": "ComputerVision",
  "Vehicles": {
    "Drone1": {
      "VehicleType": "ComputerVision",
      "AutoCreate": true,
      "Sensors": {
        "gpulidar1": {
          "SensorType": 8,
          "Enabled": true,
          "External": false,
          "NumberOfChannels": 50,
          "Range": 100,
          "MeasurementsPerCycle": 2048,
          "RotationsPerSecond": 10,
          "Resolution": 2048,
          "X": 0,
          "Y": 0,
          "Z": -0.5,
          "Roll": 0,
          "Pitch": 0,
          "Yaw": 0,
          "VerticalFOVUpper": 30,
          "VerticalFOVLower": -30,
          "HorizontalFOVStart": 0,
          "HorizontalFOVEnd": 360,
          "DrawDebugPoints": false,
          "DrawMode": 0,
          "IgnoreMarked": true,
          "GroundTruth": true,
          "InstanceSegmentation": true,
          "Annotation": "",
          "GenerateIntensity": false,
          "rangeMaxLambertianPercentage": 80,
          "rainMaxIntensity": 70,
          "rainConstantA": 0.01,
          "rainConstantB": 0.6,
          "DrawSensor": true
        }
      }
    }
  }
}