perf(velocity_smoother): use ProxQP for faster optimization

[takayuki5168]

Description

Use ProxQP instead of OSQPfor the velocity optimization.
Enabling warm start makes several scenarios fail due to the optimization failure as follows, so the warm start is disabled.
https://evaluation.tier4.jp/evaluation/reports/76156fc6-7529-5edf-b8f8-de8589118b37?project_id=prd_jt

Related links

How was this PR tested?

scenario test: https://evaluation.tier4.jp/evaluation/reports/6b10df3c-2ef3-5b69-9955-aee37d8b6c2b?project_id=prd_jt

psim
Before (with OSQP) (with warm start)

⏰ Worst Case Execution Time ⏰
onCurrentTrajectory: 56.20 [ms]
    ├── calcExternalVelocityLimit: 0.00 [ms]
    ├── updateDataForExternalVelocityLimit: 0.01 [ms]
    └── calcTrajectoryVelocity: 55.86 [ms]
        ├── applyExternalVelocityLimit: 0.01 [ms]
        ├── applyStopApproachingVelocity: 0.00 [ms]
        └── smoothVelocity: 55.84 [ms]
            ├── calcInitialMotion: 0.00 [ms]
            ├── applyLateralAccelerationFilter: 1.16 [ms]
            ├── applySteeringRateLimit: 0.13 [ms]
            ├── resampleTrajectory: 3.61 [ms]
            ├── apply: 50.68 [ms]
            │   ├── forwardJerkFilter: 0.12 [ms]
            │   ├── backwardJerkFilter: 0.22 [ms]
            │   │   └── forwardJerkFilter: 0.18 [ms]
            │   ├── mergeFilteredTrajectory: 0.02 [ms]
            │   ├── resample: 0.74 [ms]
            │   ├── resample: 0.61 [ms]
            │   ├── resample: 0.63 [ms]
            │   ├── resample: 0.62 [ms]
            │   ├── initOptimization: 0.45 [ms]
            │   └── optimize: 47.12 [ms]
            ├── overwriteStopPoint: 0.03 [ms]
            └── insertBehindVelocity: 0.15 [ms]

After with OSQP (with warm start)

⏰ Worst Case Execution Time ⏰
onCurrentTrajectory: 35.54 [ms]
    ├── calcExternalVelocityLimit: 0.00 [ms]
    ├── updateDataForExternalVelocityLimit: 0.01 [ms]
    └── calcTrajectoryVelocity: 35.27 [ms]
        ├── applyExternalVelocityLimit: 0.01 [ms]
        ├── applyStopApproachingVelocity: 0.00 [ms]
        └── smoothVelocity: 35.23 [ms]
            ├── calcInitialMotion: 0.00 [ms]
            ├── applyLateralAccelerationFilter: 1.13 [ms]
            ├── applySteeringRateLimit: 0.18 [ms]
            ├── resampleTrajectory: 3.26 [ms]
            ├── apply: 30.32 [ms]
            │   ├── forwardJerkFilter: 0.16 [ms]
            │   ├── backwardJerkFilter: 0.19 [ms]
            │   │   └── forwardJerkFilter: 0.15 [ms]
            │   ├── mergeFilteredTrajectory: 0.03 [ms]
            │   ├── resample: 1.79 [ms]
            │   ├── resample: 0.87 [ms]
            │   ├── resample: 0.71 [ms]
            │   ├── resample: 0.71 [ms]
            │   ├── initOptimization: 0.22 [ms]
            │   └── optimize: 25.49 [ms]
            ├── overwriteStopPoint: 0.03 [ms]
            └── insertBehindVelocity: 0.14 [ms]

After with ProxQP (without warm start)

⏰ Worst Case Execution Time ⏰
onCurrentTrajectory: 34.26 [ms]
    ├── calcExternalVelocityLimit: 0.00 [ms]
    ├── updateDataForExternalVelocityLimit: 0.01 [ms]
    └── calcTrajectoryVelocity: 33.88 [ms]
        ├── applyExternalVelocityLimit: 0.01 [ms]
        ├── applyStopApproachingVelocity: 0.00 [ms]
        └── smoothVelocity: 33.83 [ms]
            ├── calcInitialMotion: 0.00 [ms]
            ├── applyLateralAccelerationFilter: 2.26 [ms]
            ├── applySteeringRateLimit: 0.32 [ms]
            ├── resampleTrajectory: 12.33 [ms]
            ├── apply: 18.60 [ms]
            │   ├── forwardJerkFilter: 0.21 [ms]
            │   ├── backwardJerkFilter: 0.36 [ms]
            │   │   └── forwardJerkFilter: 0.31 [ms]
            │   ├── mergeFilteredTrajectory: 0.04 [ms]
            │   ├── resample: 1.60 [ms]
            │   ├── resample: 1.20 [ms]
            │   ├── resample: 1.29 [ms]
            │   ├── resample: 1.33 [ms]
            │   ├── initOptimization: 0.44 [ms]
            │   └── optimize: 11.99 [ms]
            ├── overwriteStopPoint: 0.07 [ms]
            └── insertBehindVelocity: 0.12 [ms]

Notes for reviewers

None.

Interface changes

None.

Effects on system behavior

None.

[github-actions]

Thank you for contributing to the Autoware project!

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

Please ensure:

• You've checked our contribution guidelines.
• Your PR follows our pull request guidelines.
• All required CI checks pass before marking the PR ready for review.

[codecov]

Codecov Report

Attention: Patch coverage is 88.23529% with 2 lines in your changes missing coverage. Please review.

Project coverage is 29.11%. Comparing base (5093495) to head (b9ece7f).
Report is 13 commits behind head on main.

Files	Patch %	Lines
...y_smoother/src/smoother/jerk_filtered_smoother.cpp	50.00%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8028      +/-   ##
==========================================
- Coverage   29.24%   29.11%   -0.13%     
==========================================
  Files        1600     1605       +5     
  Lines      117738   117823      +85     
  Branches    50720    50699      -21     
==========================================
- Hits        34427    34308     -119     
- Misses      74120    74333     +213     
+ Partials     9191     9182       -9     

Flag	Coverage Δ		*Carryforward flag
differential	18.83% <88.23%> (?)
total	29.17% <ø> (-0.07%)	⬇️	Carriedforward from 5093495

*This pull request uses carry forward flags. Click here to find out more.

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

[maxime-clem]

Is it okay to only do the change for the jerk filtered smoother ?
What about the L_2 and L_inf pseudo jerk smoother ?

[takayuki5168]

@maxime-clem
Thank you for the comment.

First of all, it's a bit hard to check whether the behavior won't change with this change.
JerkFiltered can be tested in the simulation/experiment and widely used in TIER IV's projects, so we can easily check it. But for the L2 and Linf, since there is no baseline and the algorithms are not so used, I kept the algorithms as they are.