Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

chore(ekf_localizer): increase z_filter_proc_dev for large gradient road #211

Merged

Conversation

KYabuuchi
Copy link
Contributor

@KYabuuchi KYabuuchi commented Feb 21, 2025

Description

I think the current z_filter_proc_dev is too small, so I propose increasing it.

Recently, I encountered an issue where localization failed in tunnel with a gradient.
In areas with a steep gradient, even if NDT scan matching outputs a correct pose, the 1D filter in the EKF localizer suppresses the variation in the altitude position. And the suppression lead to the NDT's mis-matching.

image

In the tunnel data I tested, this change prevented localization failure, and I did not observe any performance degradation in other environments.

The gradient of the undersea tunnel I verified had a gradient of 60m down at 600m. 📉

Related links

How was this PR tested?

  • I conducted an automatic driving experiment in the seaside area.
  • I manually drove through the undersea tunnel to verify that localization and perception did not break down.

Notes for reviewers

autoware_launch PR: autowarefoundation/autoware_launch#1337

Interface changes

None.

Effects on system behavior

None.

Signed-off-by: Kento Yabuuchi <[email protected]>
Copy link

github-actions bot commented Feb 21, 2025

Thank you for contributing to the Autoware project!

🚧 If your pull request is in progress, switch it to draft mode.

Please ensure:

@SakodaShintaro
Copy link
Contributor

SakodaShintaro commented Feb 25, 2025

The performance tests have been started using these branches.
(TIER IV INTERNAL)

Copy link
Contributor

@SakodaShintaro SakodaShintaro left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I visualized the test results, which show a slight decrease in performance for only two data points. However, after looking into them, I found that the decline is unrelated to the changes in this PR, as the tests have some randomness.

compare_result_masked

LGTM

@KYabuuchi
Copy link
Contributor Author

@SakodaShintaro Thank you for testing! 😄

@KYabuuchi KYabuuchi added the run:build-and-test-differential Mark to enable build-and-test-differential workflow. (used-by-ci) label Feb 26, 2025
@KYabuuchi KYabuuchi enabled auto-merge (squash) February 26, 2025 06:11
Copy link

codecov bot commented Feb 26, 2025

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 51.07%. Comparing base (4cb18f5) to head (ea5996e).
Report is 54 commits behind head on main.

Additional details and impacted files
@@             Coverage Diff             @@
##             main     #211       +/-   ##
===========================================
- Coverage   78.75%   51.07%   -27.69%     
===========================================
  Files          11       35       +24     
  Lines         193     1627     +1434     
  Branches       73      609      +536     
===========================================
+ Hits          152      831      +679     
- Misses         11      648      +637     
- Partials       30      148      +118     
Flag Coverage Δ
differential 51.07% <ø> (?)
total ?

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@KYabuuchi KYabuuchi merged commit ef8d6a4 into autowarefoundation:main Feb 26, 2025
31 of 33 checks passed
@KYabuuchi KYabuuchi deleted the chore/increase_z_filter_dev branch February 26, 2025 06:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
run:build-and-test-differential Mark to enable build-and-test-differential workflow. (used-by-ci)
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants