[kazkomiya]

Description

The aim of this series is to let the lateral controller run and be tested without a node, so that
it can later move to a package of its own. MPC and the classes it uses hold a logger and a
clock of their own, which means rclcpp::init() has to run first.
This PR removes those two dependencies: the package writes logs via a ROS-independent interface, and it takes
the time of the cycle as an argument.

commit1: remove two branches nothing can reach
commit2: publish the error derivatives and the solver time as diagnostics
Two commits above are trying to reduce redundant log messages.

commit3: write the log through an injected interface
LogWriter is the interface, NullLogWriter drops everything, and RosLogWriter writes to a
logger of ROS.
A plain interface would have been enough without throttling. The waiting time of a throttled
line has to live per place in the code, as RCLCPP_*_THROTTLE keeps it, and a function cannot
hold that state unless every caller passes it. That is why the AW_MPC_* macros exist. Since a
macro was needed anyway, they cover the five severities of ROS, so that the logic can keep them
after it moves out of this package. Apart from their namespace and the AW_MPC_ prefix, the
two headers hold nothing that belongs to this package.

commit4: take the ROS clock out of MPC
MPC and SteeringPredictor receive the time of the cycle as an argument. The caller already
passed the same time as the stamp of every message the cycle produces.
test_steering_predictor.cpp is new: the clock the class
held made a test impossible to write before.

Behaviour that changes

The diagnostic array grows from 27 to 32 entries. [0] to [26] keep their meaning. The four
error derivatives are written only by the dynamics vehicle model, as the removed DEBUG lines
were. The README lists the new entries.

Five logger names become one. The four names below built with rclcpp::get_logger() were
never registered with rosout, so their lines reached the terminal but not /rosout, and
~/config_logger could not set their level.This PR sends every line to the child logger of the node, which /rosout publishes and ~/config_logger can set.

logger name before	reached /rosout
<node>.lateral_controller	yes
mpc_logger	no
mpc_steer_predictor	no
mpc_utils	no
mpc_util	no

The steering prediction reads a different clock under simulated time. SteeringPredictor
holds a clock no time source feeds, so it reads the wall clock even with use_sim_time set.
It takes the stamp the node passes instead.

Three DEBUG lines and two ERROR lines are gone, as commit1 and commit2 describe.

Interface changes

DynamicsBicycleModel takes a const LogWriter & in its constructor, and
MPCUtils::calcTrajectoryYawFromXY takes one as its first argument. MPC::setLogger and
MPC::setClock give way to MPC::setLogWriter. MPC::calculateMPC and
SteeringPredictor::calcSteerPrediction take the time of the cycle. No caller of any of these
outside this package was found in this repository.

Related links

• test(autoware_mpc_lateral_controller): add characterization tests #13356: added the node tests this PR keeps unchanged

How was this PR tested?

process is properly launched by

ros2 run autoware_trajectory_follower_node controller_node_exe --ros-args \
  --params-file <trajectory_follower_node.param.yaml> --params-file <lateral/mpc.param.yaml> \
  --params-file <longitudinal/pid.param.yaml> --params-file <common/nearest_search.param.yaml> \
  --params-file <steer_offset.param.yaml> --params-file <vehicle_info.param.yaml> \
  -p lateral_controller_mode:=mpc -p longitudinal_controller_mode:=pid

The node comes up with the fifteen topics it had before, /diagnostics carries
controller: MPC_solve_checker, and ros2 component load reaches the constructor through the
class loader. With the five inputs published, ~/output/lateral_diagnostic carries 32 entries
and the warning of the solver appears once per second under the logger name
controller.lateral_controller. Setting that logger to DEBUG through ~/config_logger makes
the DEBUG lines of MPC appear and setting it back to INFO stops them.

unit test

• commands:

source /opt/ros/jazzy/setup.bash && source install/setup.bash && PKG=autoware_mpc_lateral_controller && colcon build --symlink-install --packages-select $PKG --cmake-args -DBUILD_TESTING=true -DCMAKE_BUILD_TYPE=Debug -DCMAKE_CXX_FLAGS='--coverage -O0 -g' -DCMAKE_C_FLAGS='--coverage -O0 -g' && colcon lcov-result --zero-counters --packages-select $PKG && colcon lcov-result --initial --packages-select $PKG --lcov-args=--ignore-errors=mismatch,gcov && colcon test --event-handlers console_cohesion+ --packages-select $PKG && colcon lcov-result --packages-select $PKG --lcov-args=--ignore-errors=mismatch,gcov,unused --filter "*/test/*" "/usr/*" "/opt/*" "*/install*/"

• results:

100% tests passed, 0 tests failed out of 5

Summary coverage rate:
  lines......: 88.2% (1978 of 2242 lines)
  functions..: 89.7% (200 of 223 functions)
  branches...: 47.0% (1898 of 4042 branches)

Evaluator (TIER IV internal)

Because Evaluator cannot be executed momentarily for some technical issues, I tried to locally execute scenario tests to be fulfilled, but failed due to administration to some datasets.
We agreed with the reviewers that evaluator shall be done after the logic is extracted in later PR and then this PR can be open and merged.
->
Confirmed success in the test scnearios below in evaluator
https://evaluation.ci.tier4.jp/evaluation/reports/de0e7bd1-50c9-5ade-8ce7-665e5d5d994d?project_id=autoware_dev
https://evaluation.ci.tier4.jp/evaluation/reports/e6be184a-a794-58e3-96bb-6dd0404a63e3?project_id=autoware_dev
https://evaluation.ci.tier4.jp/evaluation/reports/92aa3150-a3df-5bcd-89f8-53791537b57e?project_id=autoware_dev

[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

❌ Patch coverage is 76.66667% with 14 lines in your changes missing coverage. Please review.
✅ Project coverage is 37.44%. Comparing base (e0c3bec) to head (bf35e2f).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
...ontrol/autoware_mpc_lateral_controller/src/mpc.cpp	62.50%	6 Missing ⚠️
..._lateral_controller/src/mpc_lateral_controller.cpp	66.66%	6 Missing ⚠️
...ude/autoware/mpc_lateral_controller/log_writer.hpp	66.66%	1 Missing ⚠️
.../autoware_mpc_lateral_controller/src/mpc_utils.cpp	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #13451      +/-   ##
==========================================
- Coverage   37.56%   37.44%   -0.12%     
==========================================
  Files        2025     2033       +8     
  Lines      141946   142080     +134     
==========================================
- Hits        53319    53199     -120     
- Misses      88627    88881     +254     

Flag	Coverage Δ		*Carryforward flag
full-suite	37.42% <ø> (-0.14%)	⬇️	Carriedforward from eb0e8d0

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

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[interimadd]

Thanks for PR. I have approved with minor comment.

[Kotakku]

The rclcpp logger can record the calling function and line number along with the message string, but I believe that with RosLogWriter, all of these will end up pointing to the write function. Could you fix this if possible?