Autoware highway localization capabilitity demonstration

[liuXinGangChina]

Hello everyone. I'm delighted to share with you the results of our road testing of the latest branch of Autoware for highway localization.

1.Test Cases

1.Ego vehicle perform lane centering driving on highway
2.Ego vehicle perform lane changing driving on highway
3.Ego vehicle perform ramp in from common road to highway

2.Test Environment

Item	Description	Additions
Test infrastructure	close test field which include high way（multi lane，ramp）
Test Conditions	speed range 100 km/h ~ 120 km/h
Ego sensor for localization	1.camera 2MP 120° 2.Odometer 3.Trimble BD992 without RTK 4.IMU
Ground Truth	Asensing 571-INS with rtk

3.Test Result

Test case	Result	Additions(redline-groundtruth, blueline-ekf)	Deviation ( blue-square root, green-longitudinal, red-lateral)
High Way cruise Automatic initial pose estimation Lane changing	Failed , estimated pose is in adjacent lane
High Way cruise Manual pose initialization Lane changing	Successful , after ego pose converge with ground truth it stay with ground truth closely even during lane change
High Way cruise Automatic initial pose estimation Lane changing Ramp	Successful , after ego pose converge with ground truth it stay with ground truth closely even during lane change , however sometimes it may fail caused by wrong heading estimation
High Way cruise Manual pose initialization Lane changing Ramp	Successful , after ego pose converge with ground truth it stay with ground truth closely even during lane change

4.Video captured during test

Video will be uploaded soon.

5.Summary

Autoware's Camera based localization Pipeline （Yabloc）can achieve lane-level localization accuracy in Highway scenarios.