CLI Overview

Introduction to the Donkeycar command-line interface

The Donkeycar framework provides a command-line tool called donkey to help you manage your autonomous vehicle projects. This CLI tool offers commands for creating new car projects, calibrating hardware, training models, and analyzing data.

Installation

The donkey command is automatically available after installing Donkeycar:

pip install donkeycar[pc]

Usage

The general syntax for the donkey command is:

donkey <command> [options]

Available Commands

The Donkeycar CLI provides the following commands:

Command	Description
createcar	Create a new car application with folder structure and templates
findcar	Find your car’s IP address on the network
calibrate	Calibrate steering servo and throttle ESC
train	Train a neural network model from recorded data
makemovie	Create a video from recorded tub data
tubplot	Plot model predictions against actual data
tubhist	Generate histograms of tub data
ui	Launch the graphical user interface
createjs	Create joystick/controller configuration
cnnactivations	Visualize CNN layer activations
models	Display model database information
update	Update car templates to latest version

Getting Help

To see the list of available commands:

donkey

To get help for a specific command:

donkey <command> --help

Common Workflows

Setting Up a New Car

1. Create a new car project:

   donkey createcar --path ~/mycar
   

2. Calibrate the servos:

   cd ~/mycar
   donkey calibrate --channel 0
   

3. Drive and collect data, then train a model:

   donkey train --tub ./data --model ./models/pilot.h5
   

Analyzing Data

1. View data distribution:

   donkey tubhist --tub ./data/tub_1_20-01-01
   

2. Plot model predictions:

   donkey tubplot --tub ./data/tub_1_20-01-01 --model ./models/pilot.h5
   

3. Create a video:

   donkey makemovie --tub ./data/tub_1_20-01-01 --out drive_video.mp4
   

Configuration

Many commands accept a --config option to specify a custom configuration file:

donkey train --config ./custom_config.py --tub ./data

The default configuration file is config.py in the current directory.

Next Steps

• Create your first car
• Calibrate your hardware
• Train a model