Linear Quadratic Regulator (LQR) Lateral Controller implementation for Autoware AWSIM - f1tenth. This package provides a robust and efficient method to control the lateral dynamics of autonomous vehicles within the Autoware ecosystem using LQR. This LQR controller is designed to ensure trajectory following for autonomous driving applications.
The controller works by minimizing the lateral deviation and heading error of the vehicle from a given trajectory, providing precise steering commands to maintain the desired path.
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release \
-DCMAKE_EXPORT_COMPILE_COMMANDS=On --packages-up-to f1tenth_launchTo ensure having that the vehicle speed is constant change src/universe/external/trajectory_loader/src/trajectory_loader_node.cpp line 80:
From: trajectory_point.longitudinal_velocity_mps = point[3];
To: trajectory_point.longitudinal_velocity_mps = 0.5;
Controller implementation repository.
The class that handles controller operation inherits from the base class LateralControllerBase and implements the class's virtual methods, which provide data needed for the control algorithm through input arguments (e.g. method LateralOutput run(InputData const & input_data)).
lqr.cpp/.hpp- controller logic implementation,lqr_lateral_controller.cpp/.hpp- autoware interface for running algorithm, input/output data handling,lqr_lateral_controller_node.cpp/.hpp- not used, the controller is launched from the base class, left for compatibility and unification of the appearance of packages,lqr_lateral_controller.launch.py- not used, the controller is launched from the base class, left for compatibility and unification of the appearance of packages.
The calculated control signal is returned as the output of the function LateralOutput run(InputData const & input_data).
The presented implementation of the LQR control algorithm was created based on the control method presented in the book Vehicle Dynamics and Control. This implementation assumes that the vehicle speed is constant and does not change.
Starting the algorithm requires setting the appropriate controller in the launch files of the autoware control system. Example launch from https://github.com/amadeuszsz/autoware.git:
source ~/autoware/intall/setup.bash
ros2 launch f1tenth_launch e2e_simulator.launch.py
In order to evaluate the obtained results, the operation of the algorithm was compared with others implemented in the repository. The results are summarized in a table:
| Algorithm | Time per lap [s] |
|---|---|
| lqr_lateral_controller | 113.558 |
| pure_pursuit | 108.289 |
| mpc | 112.148 |
- Vehicle Dynamics and Control
- Lane-Keeping Control of Autonomous Vehicles Through a Soft-Constrained Iterative LQR
- Improved Linear Quadratic Regulator Lateral Path Tracking Approach Based on a Real-Time Updated Algorithm with Fuzzy Control and Cosine Similarity for Autonomous Vehicles
- Linear quadratic regulator: Discrete-time finite horizon
- Lateral Control Using Parameter Self-Tuning LQR on Autonomous Vehicle
- LQR Control of an All-Wheel Drive Vehicle Considering Variable Input Constraint
Agnieszka Piórkowska
Natalia Wiśniewska
Miłosz Gajewski
Poznan University of Technology 2024