Open mgcho0608 opened 1 year ago
You can start from the following 2 points
cyber_monitor
to check whether the topic frequency is stableThanks for reply.
I used cyber_monitor and get this results.
I can't sure but maybe the green number of the right upside is the frame rate.
It seems communication with learning_data topic was not stable. But, how can I fix it? I have observed that the car operates normally when it is not in E2E mode.
This is the result of the cyber_monitor when it was run in NOT_LEARNING mode, not E2E_TEST, with the same thing as the other one. At this time, the car drove straight along the lane.
It seems communication with learning_data topic was not stable.
@mgcho0608 learning_data is send by Apollo? how it works?
@daohu527 I have confirmed that the learning_data topic (which is only usable in RL_TEST mode) not sending messages is normal in E2E_TEST mode. The biggest issue is that in E2E_TEST mode, the messages from the planning topic, which normally maintains a frame ratio of about 8-9 during normal operation (NO_LEARNING mode), are close to 0.00, indicating that messages are hardly being transmitted from the planning topic. To investigate this, I plan to use ADEBUG logging to determine exactly where in the planning process a significant amount of time is being spent. If anyone has any good ideas related to my issue, please let me know.
Hello @mgcho0608 How did you get semantic_map for single road HD-map?
System information
Steps to reproduce the issue:
Hi, I have problems in working E2E planning mode of apollo with lgsvl simulator SingleLaneRoad map. When I ran it in E2E mode, the car moved strangely even though it was in a simple single lane environment (for example, it encroached on the lane with a zigzag pattern). The scenario type is 'LEARNING_MODEL_SAMPLE LEARNING_MODEL_RUN', so it's certain that it moved in E2E mode. Have other users experienced this problem when running in E2E mode?