Visualizers

Several visualization options are available. For in-process rendering, Rayrai is the recommended option and runs inside the simulation process (no RaisimServer required). For server-based visualization, RaisimUnity and RaisimUnreal provide external viewers. A server-based visualizer is required to view simulations when using RaisimServer-based examples. This site provides documentation for Rayrai, RaisimUnity, and RaisimUnreal.

Visualizer Selection Guide (Summary)

  • For in-process/offscreen rendering or UI embedding: Rayrai.

  • For a cross-platform, user-friendly solution: RaisimUnity.

  • For integrated GPUs on Linux: RaisimUnityOpenGL.

  • For maximum versatility, publication-quality visuals, or RL vision data: RaisimUnreal.

Visualizer Selection Guide (Detailed)

Rayrai (In-Process/Offscreen)

Rayrai is a lightweight C++ visualizer executing within the simulation process. It renders to an OpenGL texture and is architected for integration with custom UIs (ImGui, Qt, etc.) or headless pipelines.

  • In-process rendering (no RaisimServer required).

  • Offscreen render targets via RayraiWindow.

  • Supports custom visuals, point clouds, coordinate frames, and object picking.

  • Optimal for tooling, camera sensor simulation, and custom pipelines.

  • Not a full-featured server visualizer (lacks built-in maps or UI tooling).

Build instructions are available in the Installation section. Detailed usage and API coverage are provided in the Rayrai section.

RaisimUnity

raisimUnity

  • Binaries are included in the raisimUnity directory.

  • User-friendly.

  • Lower frame rate but reduced system resource consumption compared to RaisimUnreal.

  • Executes as an independent process (via RaisimServer).

  • Closed-source.

  • Compatible with Linux, macOS, and Windows.

RaisimUnreal (Beta)

raisimUnity

  • Binaries are provided via GitHub releases.

  • Currently available only on Windows and Linux.

  • The most feature-rich visualizer. Only RaisimUnreal supports maps (refer to the RaisimUnreal section).

  • Utilizes multithreading for performance; however, it may impede RL training or simulation speed.

  • User-friendly.

  • Supports graphs and bar charts (see examples/map_atlas_charts.cpp).

  • Executes as an independent process (via RaisimServer).

  • Closed-source.

RaiSimUnity vs. RaiSimUnreal Comparison

The following comparison highlights differences between RaisimUnity and RaisimUnreal.

  • Graphics Quality

    • RaisimUnity: 7/10.

    • RaisimUnreal: 10/10 (Quality may be lower on Linux due to driver discrepancies).

  • Compatibility

    • RaisimUnity: Includes an OpenGL variant to ensure compatibility with legacy or integrated GPUs lacking Vulkan support on Linux.

    • RaisimUnreal: Experimental support; reporting issues via GitHub is encouraged.

  • GPU Utilization (based on reference system benchmarks)

    • RaisimUnity: 90%.

    • RaisimUnreal: 98% (Performance is significantly higher on Windows; Linux performance can be impacted by driver behavior and weather presets).

  • GPU Memory Usage (with RaiSim examples)

    • RaisimUnity: ~2 GB.

    • RaisimUnreal: ~2 GB.

  • Mesh Loading Time

    • RaisimUnity: Very fast.

    • RaisimUnreal: Considerably slower due to the lack of mesh instancing, resulting in redundant asset loading.

  • Support

    • RaisimUnity: Actively supported and maintained.

    • RaisimUnreal: Development focus has shifted to RaisimUnreal.

  • Graphs (Time Series and Bar Charts)

    • RaisimUnity: None.

    • RaisimUnreal: Uses Kantan Chart to visualize server-provided graphs; see examples/map_atlas_charts.cpp.

  • Video Recording

    • RaisimUnity: Functional on Linux.

    • RaisimUnreal: Functional on Linux and Windows.

  • Object Interactions

    • RaisimUnity: Unsupported.

    • RaisimUnreal: Supports force application, distance measurement, and object lifecycle management.