Quick Presets
Ready-to-use configurations for common scenarios:- Fast Robot
- 3D Mapping
- Low Power
For high-speed navigation (drones, fast robots).
from neuronav import SensorConfig, SlamConfig
# Sensor: Low latency
sensor_config = SensorConfig(
rgb_width=640,
rgb_height=480,
fps=60, # High FPS
enable_imu=True
)
# SLAM: Fast processing
slam_config = SlamConfig(
custom_params={
"Rtabmap/DetectionRate": "2.0", # Process less frames
"Vis/MaxFeatures": "500", # Less features
"RGBD/LinearUpdate": "0.2", # Update less often
"RGBD/AngularUpdate": "0.2"
}
)
For high-quality 3D reconstruction.
# Sensor: High resolution
sensor_config = SensorConfig(
rgb_width=1920,
rgb_height=1080,
depth_width=1280,
depth_height=720,
fps=30
)
# SLAM: Maximum quality
slam_config = SlamConfig(
custom_params={
"Rtabmap/DetectionRate": "0", # Process all frames
"Vis/MaxFeatures": "2000", # More features
"Grid/3D": "true", # 3D occupancy grid
"Grid/CellSize": "0.01" # 1cm voxels
}
)
For battery-powered devices.
# Sensor: Minimal processing
sensor_config = SensorConfig(
rgb_width=640,
rgb_height=480,
fps=15, # Low FPS
enable_imu=False # Save power
)
# SLAM: Efficient settings
slam_config = SlamConfig(
enable_loop_closing=False, # Save CPU
custom_params={
"Rtabmap/MemoryThr": "300", # Limit memory
"Vis/MaxFeatures": "300"
}
)
Sensor Configuration
FullSensorConfig dataclass with all options:
from neuronav import SensorConfig
config = SensorConfig(
# Device selection
device_id="123456", # Camera serial number (optional)
# Resolution settings
rgb_width=1280, # Color image width
rgb_height=720, # Color image height
depth_width=640, # Depth image width
depth_height=480, # Depth image height
# Performance
fps=30, # Frames per second
# Features
enable_imu=True, # Use IMU if available
enable_ir=False, # IR projector/illuminator
# Advanced parameters
custom_params={
"exposure": "auto", # or specific value in microseconds
"gain": "16", # Sensor gain
"laser_power": "150", # 0-360 for RealSense
"temporal_filter": "true", # Smooth depth over time
"spatial_filter": "true", # Smooth depth spatially
"hole_filling": "true" # Fill depth holes
}
)
SLAM Configuration
FullSlamConfig dataclass explained:
from neuronav import SlamConfig
config = SlamConfig(
# ROS2 Topics (usually auto-configured)
rgb_topic="/camera/color/image_raw",
depth_topic="/camera/depth/image_raw",
camera_info_topic="/camera/color/camera_info",
imu_topic="/imu/data",
odom_topic="/odom",
# Frame IDs
robot_base_frame="base_link",
global_frame="map",
odom_frame="odom",
# Core features
enable_loop_closing=True, # Detect and close loops
enable_visualization=False, # RTAB-Map GUI
map_publish_frequency_ms=1000, # Map update rate
# Docker settings
use_gpu=False,
ros_domain_id=0,
# RTAB-Map parameters
custom_params={
# Detection
"Rtabmap/DetectionRate": "1.0", # Hz, 0=no limit
"Rtabmap/MemoryThr": "0", # Max nodes, 0=unlimited
# Visual features
"Vis/FeatureType": "6", # 6=ORB, 0=SURF, 11=SuperPoint
"Vis/MaxFeatures": "1000", # Features per image
# Loop closure
"Rtabmap/LoopThr": "0.11", # Loop closure threshold
"RGBD/ProximityBySpace": "true",
# Optimization
"Optimizer/Strategy": "1", # 0=TORO, 1=g2o, 2=GTSAM
"RGBD/OptimizeFromGraphEnd": "false"
}
)
Usage Examples
Fast Processing
sensor = RealSenseSensor(SensorConfig(
rgb_width=640,
rgb_height=480,
fps=30
))
slam = RTABMapSLAM(SlamConfig(
custom_params={
"Rtabmap/DetectionRate": "2.0",
"Vis/MaxFeatures": "500"
}
))
run_slam(sensor, slam)
High Accuracy
sensor = RealSenseSensor(SensorConfig(
rgb_width=1920,
rgb_height=1080,
fps=30
))
slam = RTABMapSLAM(SlamConfig(
custom_params={
"Rtabmap/DetectionRate": "0",
"Vis/MaxFeatures": "2000"
}
))
run_slam(sensor, slam)
Docker Deployment
# Build
./docker_build.sh
# Run
./docker_run.sh
# Or use docker-compose
docker-compose up
Visualization
Enable web-based 3D visualization:run_slam(sensor, slam, visualize=True)
# Open http://localhost:8765