Jump to content

Connect SuperML | Leeroopedia MCP: Equip your AI agents with best practices, code verification, and debugging knowledge. Powered by Leeroo — building Organizational Superintelligence. Contact us at founders@leeroo.com.

Principle:CARLA simulator Carla Sensor Blueprint Configuration

From Leeroopedia
Knowledge Sources
Domains Simulation, Perception, Sensor_Fusion
Last Updated 2026-02-15 00:00 GMT

Overview

Sensor blueprint configuration is the process of selecting a sensor type from CARLA's blueprint library and parameterizing its physical and operational characteristics before spawning it into the simulation.

Description

CARLA models sensors as actors defined by blueprints, which are templates specifying the sensor type and its configurable attributes. The blueprint library contains entries for every supported sensor modality, each identified by a hierarchical string ID (e.g., sensor.camera.rgb, sensor.lidar.ray_cast, sensor.other.radar). Once a blueprint is retrieved, its attributes can be modified to control the sensor's behavior.

Each sensor modality exposes different configurable parameters:

  • RGB Camera (sensor.camera.rgb): image_size_x, image_size_y, fov, sensor_tick, lens_k, lens_kcube, lens_x_size, lens_y_size, bloom_intensity, exposure_mode, shutter_speed, iso.
  • Depth Camera (sensor.camera.depth): image_size_x, image_size_y, fov, sensor_tick.
  • Semantic Segmentation Camera (sensor.camera.semantic_segmentation): image_size_x, image_size_y, fov, sensor_tick.
  • LiDAR (sensor.lidar.ray_cast): channels, range, points_per_second, rotation_frequency, upper_fov, lower_fov, horizontal_fov, atmosphere_attenuation_rate, dropoff_general_rate, dropoff_intensity_limit, dropoff_zero_intensity, noise_stddev, sensor_tick.
  • Radar (sensor.other.radar): horizontal_fov, vertical_fov, points_per_second, range, sensor_tick.
  • GNSS (sensor.other.gnss): noise_alt_bias, noise_alt_stddev, noise_lat_bias, noise_lat_stddev, noise_lon_bias, noise_lon_stddev, noise_seed, sensor_tick.
  • IMU (sensor.other.imu): noise_accel_stddev_x/y/z, noise_gyro_stddev_x/y/z, noise_gyro_bias_x/y/z, noise_seed, sensor_tick.

All attributes are set as strings via set_attribute(), and CARLA performs type conversion internally.

Usage

Blueprint configuration is used whenever creating a new sensor for data collection. It is the first step after deciding which sensor modalities to include in a data collection rig. Configuration should match the target deployment sensor specifications (e.g., matching the resolution and FOV of a real camera) or be set to collect data at the desired quality level for training or evaluation purposes.

Theoretical Basis

Sensor Parameterization Theory: Each sensor modality is characterized by a set of intrinsic parameters that define its measurement model. For cameras, these include the pinhole camera model parameters (focal length derived from FOV and image dimensions), resolution, and lens distortion coefficients. For LiDAR, parameters define the scanning geometry: number of vertical channels, angular field of view, point density (points_per_second / rotation_frequency), and maximum range.

Blueprint Pattern: CARLA uses the Prototype/Blueprint design pattern where each sensor blueprint serves as a configurable template. The blueprint defines default values for all attributes, and the user selectively overrides those relevant to their application. This separates sensor type selection from sensor parameterization, enabling flexible composition of sensor suites.

sensor_tick Parameter: Every sensor supports a sensor_tick attribute that controls its data generation rate independently of the simulation tick. When set to 0.0 (default), the sensor produces data every simulation tick. When set to a positive value, it produces data only when the specified interval has elapsed. This enables mixed-rate sensor suites (e.g., cameras at 20 Hz, LiDAR at 10 Hz) within the same synchronous simulation.

Noise Models: GNSS and IMU sensors support configurable Gaussian noise models with bias and standard deviation parameters. These simulate real-world sensor imperfections and are essential for testing perception algorithms under realistic conditions. The noise_seed parameter enables reproducible noise sequences.

Related Pages

Implemented By

Page Connections

Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment