search

FranzAlbers / ouster_point_type_adapter

1 stars 1 forks source link

[Question]: why not use ` rclcpp::SensorDataQoS() ` #1

Open zymouse opened 1 month ago

zymouse commented 1 month ago

This is a great tool that helped me get up to speed on autoware quickly!

https://github.com/FranzAlbers/ouster_point_type_adapter/blob/594ec104d53096cafb868139f1289b4c0e6dff67/src/ouster_point_type_adapter_component.cpp#L16

Q1: I have a query about publishing the QOS for point cloud, why don't you use rclcpp::SensorDataQoS(), because rclcpp::SensorDataQoS() is CPU intensive?

Q2：I am four os1-32 lidars and find the CPU usage very high, do you have a good suggestion? image

Very much looking forward to your reply, thank you very much

FranzAlbers commented 1 month ago

Hello @zymouse,

Nice to hear this repo is helping you!

Q1: We currently only log our data to ROS2 bags and don't run any online calculations on the point clouds. I agree that SensorDataQos would be preferable if you intend to do that.

Q2: I'm not sure what your lidar processing pipeline looks like, but maybe setting up a container with the ouster driver, this package, and the autoware lidar processing pipeline as composable nodes with intra_process_communication enabled could reduce the CPU load?