Skip to content

RUNNING.md

Latest commit

 

History

History
114 lines (100 loc) · 5.46 KB

RUNNING.md

File metadata and controls

114 lines (100 loc) · 5.46 KB
Raw

Running VDSim

A run book for every way to drive the simulator. Configuration layout (catalog, scenes, blueprints): CATALOG_AND_PHYSICS. Legacy simconfig notes: SIM_CONFIG_ARCH. This page is how to execute.

0. Install

pip install .            # Python API (scikit-build-core builds the vdsim extension)
# or full C++ tree:
cmake -G Ninja -B build -DVDSIM_BUILD_PYTHON=ON && cmake --build build -j

1. Author the configs (web tool)

python3 builder/server.py --port 8200          # http://<host>:8200

Tabs save validated YAML: Vehicle → catalog parts / blueprints (or legacy experiment vehicle YAML), Sensors (suite) → configs/sensors/, Map (driving line + surface) → configs/maps/, Comms (data routing) → configs/comms/, Scenario (composes all + run mode) → configs/experiments/. (See builder/README.)

2. Run an experiment — four modes (same scenario)

A. Embedded API (synchronous, no network) — for experiment code & notebooks

import sys; sys.path.insert(0, "python")
from vdsim_lab import Simulation
sim = Simulation("yongin_lap")                 # configs/experiments/yongin_lap.yaml
while not sim.done():
    sim.set_control(steer=.., throttle=.., brake=..)   # omit -> scenario autopilot
    sim.step()
    gnss = sim.get_data("gnss")                # per-sensor measurement
    gt   = sim.state()                         # ground truth (x,y,vx,Fz,slip,…)

Or assemble inline with the fluent builders and collect a time-series:

from vdsim_lab import Experiment, Vehicle, Road, Maneuver, Sensors
res = (Experiment(level="L3").vehicle(Vehicle.preset("sedan"))
       .road(Road.preset("belgian_pave")).maneuver(Maneuver.step_steer(v=20, steer=.03))
       .sensors(Sensors().gnss().imu()).run(8.0))
res.to_csv("run.csv"); print(res.summary())

B. Real-time comms (UDP; SIL/HIL nodes in the loop)

The C++ vdsim_realtime is the single comms layer: it runs a real-time SimSession, receives CMD packets on one port, and emits STATE on another. The wire format is the canonical VDS1 binary protocol (cosim/cosim_protocol.hpp, CRC32, 76B CMD / 220B STATE).

build/bin/vdsim_realtime --scene=configs/scenes/two_vehicle_race.yaml \
    --cmd-port=7001 --state-ip=127.0.0.1 --state-port=7002 --rate=200
# browser 3D monitor for the run's lifetime: add --monitor -> http://<host>:8790/
#   (guides/LIVE_MONITOR.md "When the monitor is on")
# LuGre tire (override YAML default): add --lugre or use scene lugre_grade_demo.yaml
build/bin/vdsim_realtime --scene=configs/scenes/lugre_grade_demo.yaml
python3 examples/lugre_demo.py
# L5 terrain / stunt scenes (v0.5): heightmap hill, banked plane, banked turn, loop
build/bin/vdsim_realtime --scene=configs/scenes/terrain_hill_demo.yaml
build/bin/vdsim_realtime --scene=configs/scenes/banked_oval.yaml
# bake your own analytic hill heightmap (.bin) for a terrain scene:
python3 tools/bake_synthetic_hill.py --out assets/terrain/hill_demo.bin

v0.5 ships terrain + L5 for the headless / batch / cosim paths. GUI terrain load and L5 Play (the in-browser viewer) are a v0.5.2 item — see design/V0.5_TERRAIN_L5.md. Python participants (wheel/pedal clients, viewer bridge, HIL harnesses) encode CMD / decode STATE through cosim/protocol.py — one definition of the wire format, byte-compatible with the server. (See cosim/test_udp_client.py.)

C. Batch / campaign (headless, parallel)

python3 tools/vdsim_batch.py run campaign.yaml          # run
python3 tools/vdsim_batch.py run campaign.yaml --dry    # list expanded runs

A campaign expands explicit scenarios + parameter sweep (cartesian grid) + monte_carlo into runs, executes them in a process pool, and writes per-run CSV + summary.csv (params + metrics). Metrics: cte_rms, cte_max, peak_ay, max_Fz, vmax, dist, lap_time, rms_slip. (See BATCH_RUNNER.)

D. Interactive GUI (real-time viewer, drive-by-feel)

python3 gui/server.py --port 8100              # browser viewer + live sim
  • Browser: 3D view (Orbit/Chase/Top/Cockpit), road/terrain, telemetry; 🎮 Wheel/pedals (Gamepad API), 🔊 Sound (engine/tire/wind).
  • Force feedback (real wheel, on the client PC) — UDP to the GUI at http-port+1: 
    python3 tools/wheel_ffb_sdl.py --server <host> --udp-port 8101   # Windows+Linux (pysdl2)
python3 tools/wheel_ffb.py     --server <host> --udp-port 8101   # Linux evdev
    Autostart units in tools/autostart/. FFB torque = VDSim aligning moment (rack_torque). (See tools/autostart/README.)

3. Industrial / interop

  • FMI 2.0: build vdsim_l2.fmu / vdsim_l3.fmu (fmi_export/build_*.sh), import any CS FMU via fmi_export/fmu_master.py. See theory ch16.
  • CARLA bridge: carla_integration/ (VDSim physics ↔ CARLA render/sensors).
  • ISO maneuver validation: python3 apps/validation/run_validation.py → ISO 7401/4138/3888-2 metrics + plots. Credibility: VALIDATION.

4. Which mode?

want mode
drive a controller/estimator from your own code A. API
connect external SIL/HIL/perception over the network B. UDP comms
sweep params / Monte Carlo / regression metrics C. batch
watch it, or drive with a wheel + FFB D. GUI
plug VDSim physics into a co-sim tool FMI