liuXinGangChina commented 3 months ago

Checklist

[X] I've read the contribution guidelines.
[X] I've searched other issues and no duplicate issues were found.
[X] I've agreed with the maintainers that I can plan this task.

Description

Since AWF’s milestone for the half of the year focus on urban odd，we try to test autoware in real urban odd. firstly go with localization stability

Purpose

Test localization stability in real urban environment

Possible approaches

[x] scan a map in a particular area of the city
[ ] collecting data under different traffic situation -- rush-hour, noon etc
[ ] evaluate the localization pipeline using above data

Definition of done

[ ] test report

liuXinGangChina commented 2 months ago

Test condition

Autoware Version: 2024.07.17
Docker Version: Official Docker 2024.08
Launch file: Logging_simulator.launch (only localization and related module been pulled up)
Rosbag: Lidar Ouster os1-64-U (mode 1024x10)
Traffic status: Rush hour (6.00 pm)
Traffic Participants: e-bike, bike, motorcycle, bus, truck, car, pedestrian
Test route Distance: 5km
Duration: 16 minutes
Route elements: more than 10 crossroads and zebra crossing, pass through hybrid road which shared by bike and car

Test result

Localization pipeline works well under different scenarios, localization failure or crush not detected
Path filtered by ekf is smooth and continuous
NDT‘s exec time:
1. Average: 3ms
2. Max: 10ms
Rviz screen cast: including localization visualization, ndt-exec-time and ndt-iteration-num click here to watch the video

liuXinGangChina commented 1 month ago

At the end of September，we successfully collected data under rain We use that data for lidar only mode localization(ndt) test, the result show that even in rainy env autoware's localization pipeline still work well

For detailed localization report please take a look at this discussion

autowarefoundation / autoware.universe

Test Autoware in real urban environment #8071