Skip to main content
The Config class manages configuration for your Donkeycar. It loads settings from Python files (config.py and myconfig.py) and makes them available as attributes.

​
Overview

Donkeycar uses a two-file configuration system:
  1. config.py - Default settings that come with Donkeycar templates
  2. myconfig.py - Your personal overrides that take precedence
This allows you to update Donkeycar without losing your custom settings. 
import donkeycar as dk

# Load config from current directory
cfg = dk.load_config()

# Access config values
print(cfg.CAMERA_TYPE)      # 'PICAM'
print(cfg.IMAGE_W)          # 160
print(cfg.DRIVE_LOOP_HZ)    # 20

​
Loading Configuration

​
load_config(config_path=None, myconfig="myconfig.py")

The primary function to load configuration.
​
config_path
str
default:"None"
Path to config.py file. If None, searches for config.py in the current working directory.
​
myconfig
str
default:"myconfig.py"
Filename of personal config overrides. Looked for in the same directory as config.py.
Returns:
​
cfg
Config
Config instance with all settings loaded as attributes
Example: Default Loading 
import donkeycar as dk

# Load from current directory
# Reads ./config.py and ./myconfig.py
cfg = dk.load_config()
Example: Specify Config Path 
import donkeycar as dk

# Load from specific directory
cfg = dk.load_config(config_path='/home/pi/mycar/config.py')
Example: Custom Override File 
import donkeycar as dk

# Use a different override file
cfg = dk.load_config(myconfig='myconfig_dev.py')
Example: From manage.py 
# In your car's manage.py
if __name__ == '__main__':
    args = docopt(__doc__)
    cfg = dk.load_config(myconfig=args['--myconfig'])
    
    if args['drive']:
        drive(cfg, model_path=args['--model'])

​
Config Class

​
from_pyfile(filename)

Load configuration from a Python file.
​
filename
str
required
Path to the Python file containing configuration variables
Returns:
​
success
bool
True if file was loaded successfully
from donkeycar.config import Config

cfg = Config()
cfg.from_pyfile('/home/pi/mycar/config.py')

# Access loaded values
print(cfg.CAMERA_TYPE)
Only variables in UPPER_CASE are loaded from the file. This convention separates configuration from code.

​
from_object(obj)

Load configuration from a Python object (module, class, etc).
​
obj
object
required
Object containing configuration attributes
from donkeycar.config import Config
import types

# Create a module-like object
my_config = types.ModuleType('my_config')
my_config.CAMERA_TYPE = 'WEBCAM'
my_config.IMAGE_W = 160
my_config.IMAGE_H = 120

cfg = Config()
cfg.from_object(my_config)

print(cfg.CAMERA_TYPE)  # 'WEBCAM'

​
from_dict(d, keys=[])

Update configuration from a dictionary.
​
d
dict
required
Dictionary of configuration key-value pairs
​
keys
list
default:"[]"
Optional list of keys to update. If empty, all UPPER_CASE keys are updated.
from donkeycar.config import Config

cfg = Config()
cfg.CAMERA_TYPE = 'PICAM'

# Override with dictionary
cfg.from_dict({
    'CAMERA_TYPE': 'WEBCAM',
    'IMAGE_W': 320,
    'IMAGE_H': 240
})

print(cfg.CAMERA_TYPE)  # 'WEBCAM'
Example: Selective Override 
cfg = Config()
cfg.from_pyfile('config.py')

# Only override specific keys
cfg.from_dict(
    {'CAMERA_TYPE': 'WEBCAM', 'IMAGE_W': 320, 'IMAGE_H': 240},
    keys=['CAMERA_TYPE']  # Only update CAMERA_TYPE
)

​
show()

Print all configuration values to console. 
import donkeycar as dk

cfg = dk.load_config()
cfg.show()

# Output:
# CAMERA_TYPE : PICAM
# IMAGE_W : 160
# IMAGE_H : 120
# DRIVE_LOOP_HZ : 20
# ...

​
__str__()

Get configuration as a string (list of tuples). 
import donkeycar as dk

cfg = dk.load_config()
print(str(cfg))

# Output:
# [('CAMERA_TYPE', 'PICAM'), ('IMAGE_W', 160), ('IMAGE_H', 120), ...]

​
to_pyfile(path)

Save current configuration to a Python file.
​
path
str
required
Path where the configuration file should be written
import donkeycar as dk

cfg = dk.load_config()

# Modify some values
cfg.CAMERA_TYPE = 'WEBCAM'
cfg.IMAGE_W = 320

# Save to new file
cfg.to_pyfile('/home/pi/mycar/config_backup.py')
Generated file format: 
# config_backup.py
CAMERA_TYPE = "WEBCAM"
IMAGE_W = 320
IMAGE_H = 120
DRIVE_LOOP_HZ = 20
# ... all other config values

​
Configuration System

​
config.py vs myconfig.py

The two-file system separates defaults from customization: config.py (template defaults) 
# Default configuration from template
CAMERA_TYPE = "PICAM"
IMAGE_W = 160
IMAGE_H = 120
DRIVE_LOOP_HZ = 20
JOYSTICK_MAX_THROTTLE = 0.5
myconfig.py (your overrides) 
# Your personal overrides
CAMERA_TYPE = "WEBCAM"  # Override: using webcam
IMAGE_W = 320           # Override: higher resolution
IMAGE_H = 240
# DRIVE_LOOP_HZ not specified, uses default of 20
# JOYSTICK_MAX_THROTTLE not specified, uses default of 0.5
Result when loaded: 
import donkeycar as dk
cfg = dk.load_config()

print(cfg.CAMERA_TYPE)           # 'WEBCAM' (from myconfig.py)
print(cfg.IMAGE_W)               # 320 (from myconfig.py)
print(cfg.DRIVE_LOOP_HZ)         # 20 (from config.py)
print(cfg.JOYSTICK_MAX_THROTTLE) # 0.5 (from config.py)

​
Common Configuration Options

​
Camera Settings

# Camera type
CAMERA_TYPE = "PICAM"  # PICAM, WEBCAM, CVCAM, CSIC, V4L, MOCK

# Image resolution
IMAGE_W = 160
IMAGE_H = 120
IMAGE_DEPTH = 3  # 3=RGB, 1=Grayscale

# Camera options
CAMERA_FRAMERATE = 20
CAMERA_VFLIP = False
CAMERA_HFLIP = False

​
Drive Train Settings

# Drive train type
DRIVE_TRAIN_TYPE = "PWM_STEERING_THROTTLE"

# PWM Calibration values (from 'donkey calibrate')
PWM_STEERING_THROTTLE = {
    "PWM_STEERING_PIN": "PCA9685.1:40.1",
    "PWM_THROTTLE_PIN": "PCA9685.1:40.0",
    "STEERING_LEFT_PWM": 460,
    "STEERING_RIGHT_PWM": 290,
    "THROTTLE_FORWARD_PWM": 500,
    "THROTTLE_STOPPED_PWM": 370,
    "THROTTLE_REVERSE_PWM": 220,
}

​
Vehicle Loop Settings

# Drive loop frequency
DRIVE_LOOP_HZ = 20  # Updates per second

# Max loops for testing
MAX_LOOPS = None  # None = run forever

​
Joystick/Controller Settings

# Enable joystick by default
USE_JOYSTICK_AS_DEFAULT = True

# Controller type
CONTROLLER_TYPE = 'xbox'  # ps3, ps4, xbox, nimbus, etc.

# Throttle limiting
JOYSTICK_MAX_THROTTLE = 0.5  # 50% max speed
JOYSTICK_STEERING_SCALE = 1.0
JOYSTICK_DEADZONE = 0.01

​
Training Settings

# Model type
DEFAULT_MODEL_TYPE = 'linear'  # linear, categorical

# Training parameters
BATCH_SIZE = 128
MAX_EPOCHS = 100
TRAIN_TEST_SPLIT = 0.8
LEARNING_RATE = 0.001

# Early stopping
USE_EARLY_STOP = True
EARLY_STOP_PATIENCE = 5

​
Recording Settings

# Auto-record when throttle > 0
AUTO_RECORD_ON_THROTTLE = True

# Record during AI driving (careful!)
RECORD_DURING_AI = False

# Create new directory each session
AUTO_CREATE_NEW_TUB = False

# Data storage
DATA_PATH = os.path.join(CAR_PATH, 'data')
MODELS_PATH = os.path.join(CAR_PATH, 'models')

​
AI/Autopilot Settings

# Throttle multiplier in AI mode
AI_THROTTLE_MULT = 1.0

# Launch control
AI_LAUNCH_DURATION = 0.0  # Seconds
AI_LAUNCH_THROTTLE = 0.0  # Boost throttle
AI_LAUNCH_ENABLE_BUTTON = 'R2'

​
Usage Examples

​
Example: Basic Drive Script

import donkeycar as dk
from donkeycar.parts.camera import PiCamera

def drive(cfg):
    V = dk.vehicle.Vehicle()
    
    # Use config values
    cam = PiCamera(
        image_w=cfg.IMAGE_W,
        image_h=cfg.IMAGE_H,
        image_d=cfg.IMAGE_DEPTH,
        vflip=cfg.CAMERA_VFLIP
    )
    V.add(cam, outputs=['cam/image_array'], threaded=True)
    
    # Start at configured frequency
    V.start(rate_hz=cfg.DRIVE_LOOP_HZ)

if __name__ == '__main__':
    cfg = dk.load_config()
    drive(cfg)

​
Example: Environment-Specific Configs

import os
import donkeycar as dk

# Load different config based on environment
env = os.getenv('ENV', 'production')

if env == 'development':
    cfg = dk.load_config(myconfig='myconfig_dev.py')
elif env == 'testing':
    cfg = dk.load_config(myconfig='myconfig_test.py')
else:
    cfg = dk.load_config()  # Use myconfig.py

print(f"Running in {env} mode")
print(f"Camera: {cfg.CAMERA_TYPE}")

​
Example: Runtime Overrides

import donkeycar as dk

# Load base config
cfg = dk.load_config()

# Override for this session
if using_simulator:
    cfg.CAMERA_TYPE = 'MOCK'
    cfg.DRIVE_TRAIN_TYPE = 'MOCK'
    cfg.DONKEY_GYM = True

# Override from command line
import sys
if '--debug' in sys.argv:
    cfg.MAX_LOOPS = 100
    cfg.VERBOSE_TRAIN = True

​
Example: Config Validation

import donkeycar as dk
import sys

def validate_config(cfg):
    """Check for common configuration errors"""
    errors = []
    
    # Check required settings
    if not hasattr(cfg, 'CAMERA_TYPE'):
        errors.append("CAMERA_TYPE not set")
    
    # Validate values
    if cfg.IMAGE_W <= 0 or cfg.IMAGE_H <= 0:
        errors.append("Invalid image dimensions")
    
    if cfg.DRIVE_LOOP_HZ <= 0 or cfg.DRIVE_LOOP_HZ > 100:
        errors.append("Invalid DRIVE_LOOP_HZ (must be 1-100)")
    
    # Check for deprecated settings
    if hasattr(cfg, 'OLD_SETTING'):
        errors.append("OLD_SETTING is deprecated, use NEW_SETTING")
    
    if errors:
        print("Configuration errors:")
        for error in errors:
            print(f"  - {error}")
        sys.exit(1)

cfg = dk.load_config()
validate_config(cfg)

​
Best Practices

​
Keep myconfig.py Minimal

Only override what you need to change: 
# myconfig.py - Good
CAMERA_TYPE = "WEBCAM"  # Only changed value
JOYSTICK_MAX_THROTTLE = 0.3

# myconfig.py - Avoid
# Don't copy entire config.py, only include changes
CAMERA_TYPE = "WEBCAM"
IMAGE_W = 160  # Unnecessary if same as default
IMAGE_H = 120  # Unnecessary if same as default
# ... 800 more lines ...

​
Document Your Overrides

Add comments explaining why you changed values: 
# myconfig.py

# Using webcam instead of Pi Camera
CAMERA_TYPE = "WEBCAM"

# Reduced max throttle for indoor testing
JOYSTICK_MAX_THROTTLE = 0.3

# Higher resolution for better obstacle detection
IMAGE_W = 320
IMAGE_H = 240

​
Version Control

Track myconfig.py in git to preserve your settings: 
# In your car directory
git add myconfig.py
git commit -m "Add custom config for webcam setup"

​
Backup Before Updates

Before updating Donkeycar: 
cp myconfig.py myconfig.py.backup

​
See Also

Build docs developers (and LLMs) love

Get started for freeTalk to us