Regarding the high CPU usage of the ekf_localizer node

[yuanchao16]

Use official data and commands：
ros2 launch autoware_launch logging_simulator.launch.xml map_path:=$HOME/autoware_map/sample-map-rosbag vehicle_model:=sample_vehicle sensor_model:=sample_sensor_kit

When the ekf_localizer node is started together in the autoware project and the data packet is played, the CPU usage is as follows:

When I move the ekf_localizer node out of the Autoware project and compile and run it separately, the CPU usage is as follows:

Could you please let me know if any optimizations have been made to reduce CPU usage in the Autoware project? thanks

[takam5f2]

@yuanchao16

I think that top and htop are not good tools for investigating high CPU usage.
Could you share a profiling of EKF localizer with Linux perf and flamegraph?

Please access the following links to get a user manuals for Linux perf and flamegraph.

• https://www.swift.org/documentation/server/guides/linux-perf.html
• https://www.brendangregg.com/flamegraphs.html

After preparing the tools, please execute following commands while Autoware and ekf_localizer are executed.

perf record -p <PID of ekf_localizer> --call-graph dwarf -- sleep 10

This command generates perf.out file. If you find this, please create Flamegraph with the following commands.

perf script | /path/to/flamegraph-repos/stackcollapse-perf.pl > out.perf-folded
/path/to/flamegraph-repos/flamegraph.pl out.perf-folded > perf.svg

Please attach the perf.svg file at this issue. Best,

[yuanchao16]

@takam5f2 Thank you very much for your response!

The following is the Flamegraph when the ekf_localizer node is launched together in the Autoware project and the data bag is played:

The following is the Flamegraph when I move the ekf_localizer node out of the Autoware project and compile and run it separately：

Could you please help me check if it's an issue with Eigen or something else?

[takam5f2]

@yuanchao16
Thank you for providing the flamegraphs.
The shared library of Eigen seems strange from these graphs, but I can't figure out why Eigen consumes much CPU time.

Could you tell me how you produce this situation? I'd like to reproduce this situation as well when I am available.

[yuanchao16]

@takam5f2 Thank you very much for your reply and for taking the time to review the flamegraphs.

I realized that when I moved the ekf_localizer node out of the Autoware project and compiled it separately, I did not manually specify the Release mode as required by the official documentation. After specifying the Release mode, the issue was resolved.
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release

Thanks again for your support!

[takam5f2]

@yuanchao16
Thank you for reply. Please let me close this issue.

[liuXinGangChina]

what is your imu and odom's output rate? 100hz or ?