search

czjaixuexi / carla_ros_bridge_pnc

通过carla-ros-bridge在carla上实现自动驾驶planning and control。
142 stars 50 forks source link

planning demo can not start planning path #2

Closed hankerbit closed 1 year ago

hankerbit commented 1 year ago

Hi, thanks for sharing the demo code. What are the parameters in "controller_config.yaml" to get good performance. Can you share them?

czjaixuexi commented 1 year ago

Hello, the parameters in the "controller_config.yaml" file in the current repository are the ones I adjusted and they work well based on the scenario in "controller_demo.launch". The map used is "Town05" and the vehicle speed is 8m/s. If you want to try other scenarios, here are my suggestions for adjustments: For "LQR", you can adjust "Q_ed" to speed up or slow down the convergence of lateral error, and adjust "Q_ephi" to speed up or slow down the convergence of heading angle error. Increasing the "R_value" will make the system more stable but slower in response, while decreasing the "R_value" will make the response faster but the system may become unstable. For "PurePursuit", increasing "k_pure" will increase the lookahead distance and improve system stability, but it will also increase the steady-state error and exhibit early turn-in phenomena. For "Stanley", increasing "k_cte" will accelerate the convergence of lateral error but may cause system instability.

hankerbit commented 1 year ago

Hello, the parameters in the "controller_config.yaml" file in the current repository are the ones I adjusted and they work well based on the scenario in "controller_demo.launch". The map used is "Town05" and the vehicle speed is 8m/s. If you want to try other scenarios, here are my suggestions for adjustments: For "LQR", you can adjust "Q_ed" to speed up or slow down the convergence of lateral error, and adjust "Q_ephi" to speed up or slow down the convergence of heading angle error. Increasing the "R_value" will make the system more stable but slower in response, while decreasing the "R_value" will make the response faster but the system may become unstable. For "PurePursuit", increasing "k_pure" will increase the lookahead distance and improve system stability, but it will also increase the steady-state error and exhibit early turn-in phenomena. For "Stanley", increasing "k_cte" will accelerate the convergence of lateral error but may cause system instability.

It is detail information, thanks, I will try.