2021.2.1 Release Features
Pedestrians and NPCs are separated from simulator and can be loaded from asset bundles. User still has to build NPCs and Pedestrians as external assets with Simulator, as these are still not fully supported from web user interface. Now, users can build their own NPCs and pedestrians with animations more easily with the simulator. We have also added some new NPCs and Pedestrians like Scooters, Bicyclists and Animals. These are available through our binary builds (win, linux) only.
Bridges from simulator to Autonomous-Driving or Robotics stack are also separated from simulator and can be loaded as asset bundles from the web user interface. Users can now build their own bridge to SVL Simulator and upload it as an asset bundle to user's library on web user interface to include in the simulation. Following bridges are supported as default: ROS, ROS2, CyberRT and here is an example of a custom bridge implementation.
SVL Simulator's cloud based web user interface can now manage and serve content for different versions of SVL Simulator. User can now edit asset bundles for different simulator versions to already added assets on the web user interface instead of creating duplicate assets per version. The web user interface can confirm validity of asset versions against the simulator version under test before it is run, to avoid simulation failures due to mismatched asset versions.
We have made several bug fixes, stability and performance improvements to support running multiple ego vehicles in a distributed simulation environment.
With the 2021.2 release, we have significantly improved the user experience on our new 3D sensor configuration editor that helps you create or modify sensor configurations for your vehicles. Some improvements include: better zoom and centering of the vehicle, opacity and grid toggles, axes color updates, mouse based controls, and some critical bug fixes on parent-child sensor transform.
Spectator EGO is an empty Ego vehicle without colliders, geometry, and vehicle dynamics. Use this vehicle when there is no need for Ego to be present during simulation. You can still attach sensors or observe simulation.
Full rework of Lidar sensor - it will now use cubemap texture instead of a pool of tall, 2D textures. This change fixes artifacts due to tall textures and solid point cloud rendering. This change improves lidar scaling - high beam counts, scans per rotation and frequency, should yield significantly more FPS than previous versions.
Note: Performance improvement is not noticeable if CyberRT is used, due to massive conversion performance cost.
For a full list of bug fixes, improvements, and all other changes since the last release, please see the release notes.