Lidar image problem

[dotzou]

Could anyone tell me why there is such a big difference between my LiDAR images and the author's?

[Tangzj2020]

Hello , Is your CARLA version 0.9.6? Looking at this difference may be that your version is not compatible.

[Morphlng]

Hello , Is your CARLA version 0.9.6? Looking at this difference may be that your version is not compatible.

Yes, this is a problem due to the Carla version. I've tested it on 0.9.6, it works well.

However, we would like to use it on the latest version since other part of our work is already done. I've modified some incompatible api to make the environment error-free, but this lidar display is still wrong (with 0.9.13, it's basically black, you can only see some edges of the pavement). While I'm trying to figure out what changed (data order? data meaning?）, I would like to know if somebody has already been through this.

[Tangzj2020]

Hi, I am sorry I have not met this problem as you said that ''with 0.9.13,it's basically black...", but I've successfully test the project on 0.9.13, from 0.9.6 to 0.9.13 version, I only add "client.get_trafficmanager().set_synchronous_mode(True)" in ''carla_env.py'' with comments ''Set fixed simulation step for sychronous mode'' this line. Good luck.

[Morphlng]

Hi, I am sorry I have not met this problem as you said that ''with 0.9.13,it's basically black...", but I've successfully test the project on 0.9.13, from 0.9.6 to 0.9.13 version, I only add "client.get_trafficmanager().set_synchronous_mode(True)" in ''carla_env.py'' with comments ''Set fixed simulation step for synchronous mode'' this line. Good luck.

Unfortunately, that doesn't help. The only part I've changed relating to rendering lidar data is changing to location.point.x/y/z at this line:

gym-carla/gym_carla/envs/carla_env.py

Lines 495 to 496 in c01ae77

	for location in self.lidar_data:
	point_cloud.append([location.x, location.y, -location.z])

By "basically black", I mean I can't see the red rings like this:

And I've noticed that it seems like the heading of lidar data is not right, it's facing right instead of up.

[YanlinQi]

Hi, I am sorry I have not met this problem as you said that ''with 0.9.13,it's basically black...", but I've successfully test the project on 0.9.13, from 0.9.6 to 0.9.13 version, I only add "client.get_trafficmanager().set_synchronous_mode(True)" in ''carla_env.py'' with comments ''Set fixed simulation step for synchronous mode'' this line. Good luck.

Unfortunately, that doesn't help. The only part I've changed relating to rendering lidar data is changing to location.point.x/y/z at this line:

gym-carla/gym_carla/envs/carla_env.py

Lines 495 to 496 in c01ae77

	for location in self.lidar_data:
	point_cloud.append([location.x, location.y, -location.z])

By "basically black", I mean I can't see the red rings like this:

And I've noticed that it seems like the heading of lidar data is not right, it's facing right instead of up.

I have the same problem here. Looking forward to solutions for this issue.

[Tangzj2020]

@YanlinQi @Morphlng Hi, you may try to change this line :
point_cloud.append([loaction.x, location.y, -location.z])
to the line:
point_cloud.append([loaction.x, location.y, location.z]) or point_cloud.append([loaction.point.x, location.point.y, location.point.z]),

I've successfully test my project in this setting.

[YanlinQi]

@ZuojinTang Thanks a lot for providing this solution. It indeed works! I have been wondering why there is a "-" before location.z, but didn't try it out. Many thanks to you!

[Morphlng]

try to change this line : point_cloud.append([loaction.x, location.y, -location.z]) to the line: point_cloud.append([loaction.x, location.y, location.z]) or point_cloud.append([loaction.point.x, location.point.y, location.point.z]),

By changing this line, the lidar image no longer seem black, yet the direction is still incorrect. The BEV and WayPoints is vertically displayed but the lidar image is horizontally displayed, a screenshot below can show this more directly.

I think there is still something wrong about those flip and rot90, I'm still trying to fix it.

[YanlinQi]

try to change this line : point_cloud.append([loaction.x, location.y, -location.z]) to the line: point_cloud.append([loaction.x, location.y, location.z]) or point_cloud.append([loaction.point.x, location.point.y, location.point.z]),

By changing this line, the lidar image no longer seem black, yet the direction is still incorrect. The BEV and WayPoints is vertically displayed but the lidar image is horizontally displayed, a screenshot below can show this more directly.

I think there is still something wrong about those flip and rot90, I'm still trying to fix it.

I just tried to change the rotation of y, and it works for the rotation problem:

• self.lidar_trans = carla.Transform(carla.Location(x=0.0, y=0.0, z=self.lidar_height), carla.Rotation(0, 90, 0))
  but the slip problem still exists.

[wzr6009]

i try to fix this problem, but it slit.

[wzr6009]

Brothers, I have solved this problem, you need to adjust the following parameters，my carla is 0.9.13
'd_behind': 16, # distance behind the ego vehicle (meter)
x_bins = np.arange(-(self.obs_range - self.d_behind), self.d_behind+self.lidar_bin, self.lidar_bin)
y_bins = np.arange(-self.obs_range/2, self.obs_range/2+0.01, self.lidar_bin)

[YanlinQi]

Brothers, I have solved this problem, you need to adjust the following parameters，my carla is 0.9.13 'd_behind': 16, # distance behind the ego vehicle (meter) x_bins = np.arange(-(self.obs_range - self.d_behind), self.d_behind+self.lidar_bin, self.lidar_bin) y_bins = np.arange(-self.obs_range/2, self.obs_range/2+0.01, self.lidar_bin)

Awesome! It works pretty well now. I am highly appreciated for that!

[wzr6009]

兄弟，我已经解决了这个问题，你需要调整以下参数，我的carla是0.9.13 'd_behind': 16, # ego vehicle后面的距离（米） x_bins = np.arange(-(self.obs_range - self .d_behind), self.d_behind+self.lidar_bin, self.lidar_bin) y_bins = np.arange(-self.obs_range/2, self.obs_range/2+0.01, self.lidar_bin)

惊人的！ 现在效果很好。 对此我深表感谢！

朋友你也复现他这个论文吗，要不要认识交流一下啊，我这里遇到一些问题

[YanlinQi]

兄弟，我已经解决了这个问题，你需要调整以下参数，我的carla是0.9.13 'd_behind': 16, # ego vehicle后面的距离（米） x_bins = np.arange(-(self.obs_range - self .d_behind), self.d_behind+self.lidar_bin, self.lidar_bin) y_bins = np.arange(-self.obs_range/2, self.obs_range/2+0.01, self.lidar_bin)

惊人的！ 现在效果很好。 对此我深表感谢！

朋友你也复现他这个论文吗，要不要认识交流一下啊，我这里遇到一些问题

好啊，怎么加你呢^_^

[wzr6009]

兄弟，我已经解决了这个问题，你需要调整以下参数，我的carla是0.9.13 'd_behind': 16, # ego vehicle后面的距离（米） x_bins = np.arange(-(self.obs_range - self .d_behind), self.d_behind+self.lidar_bin, self.lidar_bin) y_bins = np.arange(-self.obs_range/2, self.obs_range/2+0.01, self.lidar_bin)

惊人的！ 现在效果很好。 对此我深表感谢！

朋友你也复现他这个论文吗，要不要认识交流一下啊，我这里遇到一些问题

好啊，怎么加你呢^_^

加我微信吧 18734916009