Example map: GoMentum Station Digital Twin Lite

We are providing the GoMentum Station Digital Twin Lite (GoMentum DTL) map as an example of a point cloud based environment. The GoMentum Station area is a testing ground dedicated for autonomous vehicles located in Concord, California. We drove our data collection car and collected sensor data in this area for generating the GoMentum DTL map. Using the point cloud importer and point cloud renderer, the Simulator can dynamically load and unload the point clouds based on the location of the ego vehicles. Thus, there is no limitation on the size of the map with improved performance of simulations.

GoMentum DTL preview

Figure: Preview of GoMentum Station Digital Twin Lite map

Prerequisites top#

  • SVL Simulator

  • GoMentum DTL (Digital Twin Lite) map assetbundle

  • Apollo 5.0

Instructions for how to run with Apollo 5.0 top#

Follow these steps to run Apollo on GoMentum DTL:

  1. Launch SVL Simulator

  2. Click on the Open Browser button to launch the simulator Web UI in a web browser

    Open Browser to launch the SVL Simulator user interface

  3. In the Simulations section, locate Local-Random: GoMentum DTL (Apollo) simulation under the Available from Others tab and add it by clicking the Add button

    Import a simulation for Local-Random: GoMentum DTL (Apollo)

    1. Select your cluster from the cluster dropdown menu in the General tab

      Select cluster

    2. Select the Apollo 5.0 as your sensor configuration for the Lincoln2017MKZ vehicle in the Test case tab

      Select sensor configuration

    3. Select the Apollo 5.0 as your autopilot and enter your bridge IP address and port number in the Autopilot tab

      Select autopilot

    4. Click the Publish button to finish the simulation setup in the Publish tab

      Publish simulation

  4. (Optional) GoMentum DTL map and Lincoln2017MKZ vehicle should be automatically added to the Library. If not, please follow the instructions for Adding a map and Adding a vehicle to add them manually.

  5. In the Vehicles section under the Library, locate Lincoln2017MKZ vehicle and click it to go into a sensor configuration page

    Select the Lincoln2017MKZ

  6. Select the Apollo 5.0 sensor configuration

    Select the Apollo 5.0

  7. If you see a notification for any missing plugins, click the Add to Library button and make sure that you have all the sensor plugins listed in the configuration added to the Plugins section under the Library

    Add plugins to Library

  8. Locate Local-Random: GoMentum DTL (Apollo) simulation in the Simulations tab and click the Run Simulation button at the bottom to start simulation

    Select the simulation and click the play button

  9. Local-Random: GoMentum DTL (Apollo) simulation should now be up and running in the main window of the simulator

    First screen of GoMentum DTL in the main window of SVL Simulator

  10. Finally, launch Apollo 5.0 alongside SVL Simulator

Straight lane in Dreamview and SVL Simulator

Figure: Apollo dreamview and SVL Simulator detecting NPCs on the road with sensor visualizations enabled

Intersection in Dreamview and SVL Simulator

Figure: Apollo dreamview and SVL Simulator detecting an NPC in an intersection from top-down view

Point cloud visualizations in rviz and SVL Simulator

Figure: Rviz showing point clouds published to ROS from SVL Simulator

Known issues top#

Overall, Apollo will be able to drive around most areas of GoMentum DTL with no issues, but there are some known issues in GoMentum DTL as this is an early access version. Here is a list of known issues for the current release that will be fixed soon in future releases.

Dynamic objects top#

Some other cars were captured while collecting sensor data from the field and remain as dummy noises on some roads. Later, we will have a Dynamic object removal feature in the Digital Twin Lite pipeline which removes dynamic objects such as cars or pedestrians from point clouds during the data processing.

Dynamic objects in GoMentum DTL

Figure: Streaks of blur on the road due to a moving car captured in ROSBAG

Baked-in shadows top#

Colorized point clouds have shadows baked in already (e.g., under trees) because shadows were captured in camera images during a data collection phase. In the future release, we’ll remove the baked-in shadows from images and use simulated shadows instead based on the sun position in the simulator.

Baked in shadows in GoMentum DTL

Figure: Baked-in shadows from trees and buildings on the road