What's the exact goal of LiDAR perception module?

[jiachens]

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 18.04):
• Apollo version (5.0):

1. I am not sure whether the goal of LiDAR perception is to:
   a. percept all the obstacles which have impacts on the driving behavior
   b. or only percept the vehicle, non-vehicle, pedestrian, these 3 classes

For example, if there is a rock on the road which definitely has impacts on the driving behavior, is LiDAR perception module's responsibility to detect it as an unknown class? or just ignore it?

The same example generalizes to small animals on the road, should LiDAR perception module detect it or not? Note that I only consider LiDAR, no other sensors.

Thanks in advance!

[techoe]

Currently lidar is used for mapping/localization/perception. In perception, we facilitates two lidar-based object detection approaches: One is to detect multiple objects (pedestrian, cyclist, motorcycle, sedan, truck, and more) using CNN and the other is to detect and segment objects using min-cut which detects any object in lidar data.

[jiachens]

Currently lidar is used for mapping/localization/perception. In perception, we facilitates two lidar-based object detection approaches: One is to detect multiple objects (pedestrian, cyclist, motorcycle, sedan, truck, and more) using CNN and the other is to detect and segment objects using min-cut which detects any object in lidar data.

Thanks a lot for the reply!
I found codes about NCUT which could be Normalized-Cut. Can you give me a hint of where the code is used? Since when I use the previous Apollo version, there is no NCUT or min-CUT in the codebase

[jiachens]

Currently lidar is used for mapping/localization/perception. In perception, we facilitates two lidar-based object detection approaches: One is to detect multiple objects (pedestrian, cyclist, motorcycle, sedan, truck, and more) using CNN and the other is to detect and segment objects using min-cut which detects any object in lidar data.

Also, I just wonder that is the CNN segmentor responsible for detecting all the objects on the road, or it is responsible for only detecting the classified obstacles?

[apollodevcommunity]

得到障碍物的位置，速度，加速度等信息。具体的处理流程请参考https://github.com/ApolloAuto/apollo/blob/master/docs/specs/3d_obstacle_perception_cn.md ——回复参考来自Apollo开发者社区issue达人赛 https://apollo.auto/developer/questiondata_cn.html?target=395