Apollo: The real-time relative map can't be displayed correctly in Dreamview for Apollo master branch (May 8, 2019)

Created on 9 May 2019  ·  6Comments  ·  Source: ApolloAuto/apollo

Describe the bug
A bag file (csc_hw_right.bag) was recorded using the rosbag command. The bag file was converted to a record file (csc_hw_right.record) using the rosbag_to_record csc_hw_right.bag csc_hw_right.record command . The real-time relative map was generated using this record file. However, the map is displayed in chaos in Dreamview and the web page shows nothing when the F5 key is pressed to refresh the display. The source code was downloaded from the latest master branch (May 8, 2019, commit id 2f308aac31e13a4c234e3b12010116333af398e6) based on the Cyber RT framework. The /apollo/relative_map topic is displayed properly through the cyber_monitor tool.

To Reproduce
Steps to reproduce the behavior:

  1. download the latest master branch source code.
git pull --rebase upstream master
git reset --hard git reset --hard 2f308aac31e13a4c234e3b12010116333af398e6
  1. start and enter the docker.
bash docker/scripts/dev_start.sh
bash docker/scripts/dev_into.sh
  1. build the apollo project.
bash apollo.sh build -j 8
  1. convert the bag file to a record file.
rosbag_to_record data/bag/csc_hw_right.bag data/bag/csc_hw_right.record
  1. modify the modules/common/data/global_flagfile.txt file as follows:
    change
--use_navigation_mode=false

to

--use_navigation_mode=true

and add a line:

--local_utm_zone_id=49

global_flagfile_config

  1. modify the modules/localization/conf/localization.conf file as follows:
    change
--local_utm_zone_id=10

to

--local_utm_zone_id=49

localization_config

  1. start Dreaview:
bash scripts/bootstrap.sh
  1. On the Dreamview web page, enable the Relative Map module and confirm that it is currently in Navigation Mode.
    dreamview_in_relative_map_and_navigation_mode

  2. Replay the record file:

cyber_recorder play -f data/bag/csc_hw_right.record -l
  1. Send a navigation line:
bash scripts/navigator.sh data/bag/csc_hw_right.smoothed

Expected behavior
The real-time relative map with one lane is correctly displayed in Dreamview and the web page can be updated correctly when the F5 key is pressed to refresh the display.

Actual results
The real-time relative map with one lane is displayed in chaos in Dreamview and the web page shows nothing when the F5 key is pressed to refresh the display.

Screenshots
dreamview_in_chaos
The real-time relative map with one lane is displayed in chaos

dreamview_in_chaos_after_refresh
The web page shows nothing when the F5 key is pressed to refresh the display.

relative_map_topic
relatvie_map_topic_in_detail
The /apollo/relative_map topic is displayed properly through the cyber_monitor tool.

Desktop (please complete the following information):

  • OS: Ubuntu 16.04
  • Browser: Chrome 68.0.3440.106 (Official Build) (64-bit)
  • Apollo Version: master branch (May 8, 2019, commit id 2f308aac31e13a4c234e3b12010116333af398e6)

Data file
URL: https://pan.baidu.com/s/1qyXyR94Yhhug28ahzkZtuw
Key: t95z

Simulation & Dreamview Bug Help wanted

All 6 comments

@natashadsouza Thank you.
It's a problem not for Apollo 2.5 but for Apollo master branch (above Apollo 3.5). In fact, the real-time relative map is displayed correctly with Dreamview for Apollo prior to 3.0. However, there are some displaying problems with Dreamview for Apollo master branch (above Apollo 3.5).

@davidhopper2003 noted. Thank you very much for explaining the requirements so clearly. Waiting on @vlin17 for a response.

@natashadsouza Sure. Thanks once more.

Thank you for always providing reproduction steps in details, that definitely speed up the debugging process. The code was assuming map_geometry's CurveSegment always has start_position defined, we have updated our code to consider the situation where no start_position is given. Please try again after https://github.com/ApolloAuto/apollo/pull/8276 is delivered.

@vlin17 Thank you very much. You are awesome. The contribution of the Apollo team is gorgeous and it is a great honour to submit a bug for your team.

Thank you @davidhopper2003 for your encouraging words. We truly appreciate your support in reporting new bugs to us!
Closing this issue currently.

Was this page helpful?
0 / 5 - 0 ratings