OpenCDA API Overview

Complete API Documentation

For comprehensive OpenCDA API documentation, visit the Official OpenCDA Documentation.

Component	Purpose	Status
CavWorld	Global CAV registry & shared ML models	✅ Enhanced
ScenarioManager	Simulation setup & vehicle spawning	✅ Stable
VehicleManager	Individual vehicle coordination	✅ Enhanced
PerceptionManager	Sensor fusion & object detection	✅ YOLOv8
LocalizationManager	Position tracking & filtering	✅ Stable
BehaviorAgent	Path planning & decision making	✅ Stable
V2XManager	Vehicle-to-vehicle communication	✅ Stable

Key Updates

ML ImprovementsSystem Changes

• YOLOv8 Integration: ~20% performance improvement with automatic YOLOv5 fallback
• Thread Safety: Fixed GIL issues in multi-threaded scenarios
• Memory Efficiency: Shared model instances across all vehicles
• Compatibility: Maintained existing camera-lidar fusion interfaces

• Version Management: Centralized in opencda/__init__.py
• CARLA Version: Fixed to 0.9.15 for consistency
• Package Manager: Pixi-based environment management
• Documentation: Enhanced API structure with examples

---

Core Workflow

What it does: Orchestrates simulation setup, vehicle management, and execution loop

Basic SetupVehicle CreationSimulation LoopDesign Principles

from opencda.scenario_testing.utils.customized_map_api import customized_map_helper

# Load configuration and create CAV world
scenario_params = load_yaml(config_yaml)
cav_world = CavWorld(apply_ml=True)

# Initialize scenario manager
scenario_manager = ScenarioManager(
    scenario_params, apply_ml=True,
    xodr_path=xodr_path, cav_world=cav_world
)

# Create individual CAVs and platoons 
single_cavs = scenario_manager.create_vehicle_manager(['platooning']) 
platoons = scenario_manager.create_platoon_manager()

# Add background traffic
traffic_manager, bg_vehicles = scenario_manager.create_traffic_carla()

while True: 
    scenario_manager.tick()

    # Update platoons
    for platoon in platoons:
        platoon.update_information()
        platoon.run_step()

    # Update individual CAVs
    for cav in single_cavs:
        if not cav.v2x_manager.in_platoon():
            cav.update_info()
            control = cav.run_step()
            cav.vehicle.apply_control(control)

Principle	Implementation	Benefits
Manager Pattern	Dedicated manager for each subsystem	Clear responsibilities, easy testing
Configuration-Driven	YAML configs with inheritance	Flexible scenarios, easy experimentation
Plugin Architecture	Custom algorithms in opencda/customize/	Extensible without core changes

---

Integration Guide

Quick Start

Get started with OpenCDA in 4 steps: - Installation

# Install with Pixi 
pixi install 
pixi shell 

- Quick Test

# Verify installation 
pixi run quick-test 

- Basic Scenario

# Run simple scenario 
python opencda.py -t single_2lanefree_carla 

- Custom Development

# Extend with custom algorithms 
from opencda.customize.ml_libs import MyCustomDetector 
cav_world = CavWorld(apply_ml=True, custom_ml=MyCustomDetector()) 

Integration Points: CARLA 0.9.15, SUMO co-simulation, V2X communication, PyTorch ML models