FCND - Backyard Flyer Project

In this project, you’ll set up a state machine using event-driven programming to autonomously flying a drone. You will be using flying a quadcopter in Unity simulator. After completing this assignment, you’ll be familiar with sending commands and receiving incoming data from the drone.

The python code you write is similar to how the drone would be controlled from a ground station computer or an onboard flight computer. Since communication with the drone is done using MAVLink, you will be able to use your code to control an PX4 quadcopter autopilot with very little modification!

Step 1: Download the Simulator

If you haven’t already, download the version of the simulator that’s appropriate for your operating system from this repository.

Step 2: Set up your Python Environment

If you haven’t already, set up your Python environment and get all the relevant packages installed using Anaconda following instructions in this repository

Step 3: Clone this Repository

git clone https://github.com/udacity/FCND-Backyard-Flyer

Task

The required task is to command the drone to fly a 10 meter box at a 3 meter altitude. You’ll fly this path in two ways: first using manual control and then under autonomous control.

Manual control of the drone is done using the instructions found with the simulator.

Autonomous control will be done using an event-driven state machine. First, you will need to fill in the appropriate callbacks. Each callback will check against transition criteria dependent on the current state. If the transition criteria are met, it will transition to the next state and pass along any required commands to the drone.

Drone API

To communicate with the simulator (and a real drone), you will be using the UdaciDrone API. This API handles all the communication between Python and the drone simulator. A key element of the API is the Drone superclass that contains the commands to be passed to the simulator and allows you to register callbacks/listeners on changes to the drone’s attributes. The goal of this project is to design a subclass from the Drone class implementing a state machine to autonomously fly a box. A subclass is started for you in backyard_flyer.py

Drone Attributes

The Drone class contains the following attributes that you may find useful for this project:

self.armed: boolean for the drone’s armed state
self.guided: boolean for the drone’s guided state (if the script has control or not)
self.local_position: a vector of the current position in NED coordinates
self.local_velocity: a vector of the current velocity in NED coordinates

For a detailed list of all of the attributes of the Drone class check out the UdaciDrone API documentation.

Registering Callbacks

As the simulator passes new information about the drone to the Python Drone class, the various attributes will be updated. Callbacks are functions that can be registered to be called when a specific set of attributes are updated. There are two steps needed to be able to create and register a callback:

Create the callback function:

Each callback function you may want needs to be defined as a member function of the BackyardFlyer class provided to you in backyard_flyer.py that takes in only the self parameter. You will see in the template provided to you in backyard_flyer.py three such callback methods you may find useful have already been defined. For example, here is the definition of one of the callback methods:

class BackyardFlyer(Drone):
    ...

    def local_position_callback(self):
        """ this is triggered when self.local_position contains new data """
        pass

In order to have your callback function called when the appropriate attributes are updated, each callback needs to be registered. This registration takes place in you BackyardFlyer’s __init__() function as shown below:

class BackyardFlyer(Drone):

    def __init__(self, connection):
        ...

        # TODO: Register all your callbacks here
        self.register_callback(MsgID.LOCAL_POSITION, self.local_position_callback)

Since callback functions are only called when certain drone attributes are changed, the first parameter to the callback registration indicates for which attribute changes you want the callback to occur. For example, here are some message id’s that you may find useful (for a more detailed list, see the UdaciDrone API documentation):

MsgID.LOCAL_POSITION: updates the self.local_position attribute
MsgID.LOCAL_VELOCITY: updates the self.local_velocity attribute
MsgID.STATE: updates the self.guided and self.armed attributes

Outgoing Commands

The UdaciDrone API’s Drone class also contains function to be able to send commands to the drone. Here is a list of commands that you may find useful during the project:

connect(): Starts receiving messages from the drone. Blocks the code until the first message is received
start(): Start receiving messages from the drone. If the connection is not threaded, this will block the code.
arm(): Arms the motors of the quad, the rotors will spin slowly. The drone cannot takeoff until armed first
disarm(): Disarms the motors of the quad. The quadcopter cannot be disarmed in the air
take_control(): Set the command mode of the quad to guided
release_control(): Set the command mode of the quad to manual
cmd_position(north, east, down, heading): Command the drone to travel to the local position (north, east, down). Also commands the quad to maintain a specified heading
takeoff(target_altitude): Takeoff from the current location to the specified global altitude
land(): Land in the current position
stop(): Terminate the connection with the drone and close the telemetry log

These can be called directly from other methods within the drone class:

self.arm() # Seends an arm command to the drone

Manual Flight

To log data while flying manually, run the drone.py script as shown below:

python drone.py

Run this script after starting the simulator. It connects to the simulator using the Drone class and runs until tcp connection is broken. The connection will timeout if it doesn’t receive a heartbeat message once every 10 seconds. The GPS data is automatically logged.

To stop logging data, stop the simulator first and the script will automatically terminate after approximately 10 seconds.

Alternatively, the drone can be manually started/stopped from a python/ipython shell:

from udacidrone import Drone
from udacidrone.connection import MavlinkConnection
conn = MavlinkConnection('tcp:127.0.0.1:5760', threaded=False)
drone = Drone(conn,tlog_name="TLog-manual.txt")
drone.start()

If threaded is set to False, the code will block and the drone logging can only be stopped by terminating the simulation. If the connection is threaded, the drone can be commanded using the commands described above, and the connection can be stopped (and the log properly closed) using:

drone.stop()

When starting the drone manually from a python/ipython shell you have the option to provide a desired filename for the telemetry log file (such as “TLog-manual.txt” as shown above). This allows you to customize the telemetry log name as desired to help keep track of different types of log files you might have. Note that when running the drone from python drone.py for manual flight, the telemetry log will default to “TLog-manual.txt”.

Message Logging

The telemetry data is automatically logged in “Logs\TLog.txt” or “Logs\TLog-manual.txt” for logs created when running python drone.py. Each row contains a comma seperated representation of each message. The first row is the incoming message type. The second row is the time. The rest of the rows contains all the message properties. The types of messages relevant to this project are:

MsgID.STATE: time (ms), armed (bool), guided (bool)
MsgID.GLOBAL_POSITION: time (ms), longitude (deg), latitude (deg), altitude (meter)
MsgID.GLOBAL_HOME: time (ms), longitude (deg), latitude (deg), altitude (meter)
MsgID.LOCAL_POSITION: time (ms), north (meter), east (meter), down (meter)
MsgID.LOCAL_VELOCITY: time (ms), north (meter), east (meter), down (meter)

Reading Telemetry Logs

Logs can be read using:

t_log = Drone.read_telemetry_data(filename)

The data is stored as a dictionary of message types. For each message type, there is a list of numpy arrays. For example, to access the longitude and latitude from a MsgID.GLOBAL_POSITION:

# Time is always the first entry in the list
time = t_log['MsgID.GLOBAL_POSITION'][0][:]
longitude = t_log['MsgID.GLOBAL_POSITION'][1][:]
latitude = t_log['MsgID.GLOBAL_POSITION'][2][:]

The data between different messages will not be time synced since they are recorded at different times.

Autonomous Control State Machine

After getting familiar with how the drone flies, you will fill in the missing pieces of a state machine to fly the drone autonomously. The state machine is run continuously until either the mission is ended or the Mavlink connection is lost.

The six states predefined for the state machine:

MANUAL: the drone is being controlled by the user
ARMING: the drone is in guided mode and being armed
TAKEOFF: the drone is taking off from the ground
WAYPOINT: the drone is flying to a specified target position
LANDING: the drone is landing on the ground
DISARMING: the drone is disarming

While the drone is in each state, you will need to check transition criteria with a registered callback. If the transition criteria are met, you will set the next state and pass along any commands to the drone. For example:

def state_callback(self):
	if self.state == States.DISARMING:
    	if !self.armed:
        	self.release_control()
        	self.in_mission = False
        	self.state = States.MANUAL

This is a callback on the state message. It only checks anything if it’s in the DISARMING state. If it detects that the drone is successfully disarmed, it sets the mode back to manual and terminates the mission.

Running the State Machine

After filling in the appropriate callbacks, you will run the mission:

python backyard_flyer.py

Similar to the manual flight, the GPS data is automatically logged to the specified log file.

Reference Frames

Two different reference frames are used. Global positions are defined [longitude, latitude, altitude (pos up)]. Local reference frames are defined [North, East, Down (pos down)] and is relative to a nearby global home provided. Both reference frames are defined in a proper right-handed reference frame . The global reference frame is what is provided by the Drone’s GPS, but degrees are difficult to work with on a small scale. Conversion to a local frame allows for easy calculation of m level distances. Two convenience function are provided to convert between the two frames. These functions are wrappers on utm library functions.

# Convert a local position (north, east, down) relative to the home position to a global position (lon, lat, up)
def local_to_global(local_position, global_home):

# Convert a global position (lon, lat, up) to a local position (north, east, down) relative to the home position
def global_to_local(global_position, global_home):

Submission Requirements

Filled in backyard_flyer.py

Flying-Cars

Contains Flying Car Nanodegree Projects