AV Simulation

Test autonomous driving algorithms in photorealistic virtual environments with realistic traffic, weather, and sensor simulation.

Why Simulation is Essential

Autonomous vehicles cannot be safely tested at scale on public roads. Simulation enables millions of miles of virtual testing, including rare and dangerous scenarios that would be impossible to reproduce safely in the real world. Companies like Waymo simulate billions of miles annually.

AV Simulators

Getting Started with CARLA

import carla
import random

# Connect to CARLA server
client = carla.Client('localhost', 2000)
world = client.get_world()

# Spawn an ego vehicle
bp_lib = world.get_blueprint_library()
vehicle_bp = bp_lib.find('vehicle.tesla.model3')
spawn_point = random.choice(world.get_map().get_spawn_points())
vehicle = world.spawn_actor(vehicle_bp, spawn_point)

# Attach sensors
camera_bp = bp_lib.find('sensor.camera.rgb')
camera_bp.set_attribute('image_size_x', '1920')
camera_bp.set_attribute('image_size_y', '1080')
camera_transform = carla.Transform(carla.Location(x=1.5, z=2.4))
camera = world.spawn_actor(camera_bp, camera_transform, attach_to=vehicle)

# Add LiDAR
lidar_bp = bp_lib.find('sensor.lidar.ray_cast')
lidar_bp.set_attribute('channels', '64')
lidar_bp.set_attribute('range', '100')
lidar = world.spawn_actor(lidar_bp, camera_transform, attach_to=vehicle)

# Enable autopilot for traffic
vehicle.set_autopilot(True)

Scenario Testing

Effective AV simulation requires testing across diverse scenarios:

Metrics for Evaluation

• Safety: Collision rate, time-to-collision, minimum distance to obstacles
• Comfort: Jerk (rate of acceleration change), lateral acceleration, steering smoothness
• Efficiency: Trip time, distance traveled, fuel/energy consumption
• Rule compliance: Speed limit adherence, traffic signal compliance, lane violations