digital_twin

Digital Twin — Data Generation

This folder contains the code that produced the dataset in ../data/. It is the Python end of the digital-twin pipeline:

DEM (Copernicus COP30)
  → MATLAB / RoadRunner R2024a   (HD map from the GU-186 terrain)
  → CARLA Sim 0.9.16             (autonomous traversal, 20 Hz kinematic logging)
  → energy model in Python       (this folder)

CARLA's Traffic Manager drives one vehicle over the truncated GU-186 segment while a Python client logs the kinematic trace; the energy variables (SOC, power, current, voltage, regeneration, battery temperature, …) are then computed tick-by-tick by the model here and written to a CSV.

Files

File	Purpose
ev_model.py	Longitudinal EV energy model with weather coupling, plus EV_LIBRARY, a dictionary of parameterized vehicles (Tesla Model 3 RWD, Audi e-tron 55 quattro, and a Cybertruck entry).
carla_acquisition.py	CARLA acquisition script. Holds the 12-scenario factorial catalog (SCENARIOS); running a scenario is a matter of setting ACTIVE_SCENARIO. Writes datos_grid_<ID>.csv.

ev_model.py

ElectricVehicleModel integrates the longitudinal force balance (grade, rolling resistance, aerodynamic drag, inertia) and maps wheel power to battery power through motor/regeneration efficiencies, discharge/regen limits, an internal-resistance thermal model, and SOC-dependent pack voltage. Optional weather coupling applies three effects:

1. HVAC auxiliary power vs ambient temperature (hvac_aux_power),
2. a battery-capacity factor vs ambient temperature (battery_capacity_factor), and
3. an effective rolling-resistance coefficient vs pavement condition (rolling_coefficient).

The module has no CARLA dependency and can be imported on its own (it is also reused by the federated experiments' physical baseline). The full vehicle- and weather-parameter tables, and their literature sources, are documented in the dataset README.

carla_acquisition.py

The script defines all 12 scenarios of the 2 × 2 × 3 grid (vehicle × initial SOC × weather) in the SCENARIOS constant. To produce one log:

1. set ACTIVE_SCENARIO to the desired identifier (e.g. 'S05');
2. start CARLA 0.9.16 with the GU-186 HD map loaded;
3. run the script.

A run is one-way and ends when the vehicle has accumulated ≥ 10 km of physical travel (a stuck detector and a safety timeout are the fallback stop conditions). The output file is named datos_grid_<ID>.csv and matches the schema documented in the dataset README.

Note on road grade. The OpenDRIVE network exported from RoadRunner has flat lane elevation (z = 0), while the visual mesh and physical colliders keep the DEM elevation. The model therefore reads the grade from the vehicle's instantaneous pitch angle, not from the waypoint network — see the dataset README for details.

Requirements

• CARLA Sim 0.9.16 and its Python API (only for carla_acquisition.py).
• Python 3.12 with numpy and opencv-python (cv2, used for the live HUD).

ev_model.py on its own needs only the Python standard library.