Skip to content

DIVP Automation Tool

Automation around DIVP simulation runs: it parses DIVP scenario XML, drives the MATLAB/Simulink models that render camera and LiDAR data for each agent, and generates OpenLABEL 1.0.0 annotations from the simulation outputs. This is the tool used to produce the RealSim-CP dataset.

Location: tools/divp-automation-tool/

Requires DIVP / MATLAB

This tool drives the DIVP V-Drive Simulink models through the MATLAB Engine, so it needs a licensed MATLAB and access to DIVP. It is published here for transparency and reproducibility; most users only consume the resulting dataset and do not need to run it. See Simulator (DIVP).

How it works

flowchart LR
  xml["DIVP scenario XML<br/>+ trajectory CSVs"] --> decomp["generate_scenario_xml()<br/>decompose into per-agent scenarios"]
  decomp --> sim["SimulationManager<br/>MATLAB Engine drives V-Drive"]
  sim --> render["CameraAutomation.slx<br/>LidarV20Automation.slx<br/>GroundTruthBboxAutomation.slx"]
  render --> out["per-agent images + point clouds + bbox"]
  out --> label["generate_label.py<br/>OpenLABEL 1.0.0 JSON"]
  • main.py — entry point. generate_scenario_xml() decomposes a multi-agent scenario into per-agent scenario XMLs; SimulationManager starts the MATLAB engine, loads the Simulink models, and simulates each sensor-equipped agent.
  • generate_label.py — assembles the OpenLABEL annotation file (streams, coordinate systems, per-frame transforms, object cuboids) from the outputs.
  • lib/calculate_matrix.py — rotation-matrix / Euler→quaternion helpers.
  • lib/utils.py — JSON I/O helpers.
  • *.slx, *.m, *_automation.json — the MATLAB/Simulink automation models and their render configs (camera + LiDAR V20).

Requirements

  • Python 3.11+
  • matlabengine==24.1.4, numpy>=2.3.4, pandas>=2.3.3 (see pyproject.toml)
  • A licensed MATLAB compatible with the matlabengine version, plus the DIVP V-Drive toolbox.

Install & run

cd tools/divp-automation-tool
pip install -e .

python main.py

By default main.py runs the automation_scenario example. The expected layout is a scenarios/<name>/ tree containing the DIVP scenario XML, OSI supplement JSON, and trajectory CSVs; outputs are written under scenarios/<name>/output_data/.

Scenario data not bundled

The large per-location scenario inputs/outputs are not included in this repository. The rendered, annotated result is the published RealSim-CP dataset.