Tutorial

First, you need to know that Py4bot is just a framework, not a ready-to-use software. So it is up to you to write the application matching your robot. But don't worry, it is very easy.

The framework contains all base code to setup a working multi-legs robot. All we have to do is overwrite a few classes, and implement virtual methods. It means that you need to know how to code in Python, and what classes / methods are... If it is not the case, have a look ​here.

Before we dive into the tutorials, we need to install Py4bot, as described in the InstallationGuide. It is strongly suggested to install from git, so we can keep it up-to-date very easily.

Once it is done, we don't need to edit/modify the source code anymore; we will write our own code in a dedicated directory.

Vocabulary

First, lets define a few words. I tried to use commonly used vocabulary, and hope I did understand them correctly. Feel free to contact me if you find mistakes.

• IK: Inverse Kinematic

Tutorials

• Tutorial 1: First robot
• Tutorial 1: Using the 3D simulator
• Tutorial 3: Servos calibration
• Tutorial 4: Add gamepad support

Tutorial 2: Using the 3D simulator

Files for this tutorial are available in ​py4bot/examples/tutorial_2/

Here, we will see how to use the 3D simulation stuff.

As before, we create a new empty dir in our home dir:

$ mkdir tutorial_2
$ cd tutorial_2

#!/usr/bin/env python
# -*- coding: utf-8 -*-

import math

import visual

from py4bot.api import *

import settings

Nothing fancy, here.

class Hexapod3D(Robot3D):
    def _createBody(self):
        return Body(settings.LEGS_ORIGIN)

    def _createLegs(self):
        legs = {}
        legIk = {}
        for legIndex in settings.LEGS_INDEX:
            legs[legIndex] = Leg3Dof(legIndex, {'coxa': Coxa(), 'femur': Femur(), 'tibia': Tibia()}, settings.FEET_NEUTRAL[legIndex])
            legIk[legIndex] = Leg3DofIk(settings.LEGS_GEOMETRY[legIndex])

        return legs, legIk

Same as before, except that we inherits from Robot3D, instead of Robot.

    def _createActuatorPool(self):
        driver = Driver3D()
        pool = ActuatorPool(driver)

        num = 0
        for leg in self._legs.values():

            # Create joints actuators
            coxa3D = Coxa3D(leg.coxa, num, settings.LEGS_GEOMETRY[leg.index]['coxa'])
            pool.add(coxa3D)
            num += 1

            femur3D = Femur3D(leg.femur, num, settings.LEGS_GEOMETRY[leg.index]['femur'])
            pool.add(femur3D)
            num += 1

            tibia3D = Tibia3D(leg.tibia, num, settings.LEGS_GEOMETRY[leg.index]['tibia'])
            pool.add(tibia3D)
            num += 1

            # Init 3D view
            coxa3D.frame = self._body3D
            femur3D.frame = coxa3D
            tibia3D.frame = femur3D

            coxa3D.pos = (settings.LEGS_ORIGIN[leg.index]['x'], 0, -settings.LEGS_ORIGIN[leg.index]['y'])
            femur3D.pos = (settings.LEGS_GEOMETRY[leg.index]['coxa'], 0, 0)
            tibia3D.pos = (settings.LEGS_GEOMETRY[leg.index]['femur'], 0, 0)

            coxa3D.rotate(angle=math.radians(settings.LEGS_ORIGIN[leg.index]['gamma0']), axis=(0, 1, 0))

        return pool

Here, we use specific 3D actuators: Coxa3D, Femur3D, Tibia3D; they all use the ​VPython library to represent themselves on the screen.

Note that we don't use a custom actuator mapping; nums are automatically incremented.

We also need to configure the hierarchy, and the relative positions of the actuators.

Note: this may change in the future; I'm working on a simpler way to do all this.

class Gamepad(RemoteControl):
    def _createFrontend(self):
        return Thrustmaster(settings.THRUSTMASTER_PATH)

    def _buildComponents(self):
        self._addConfig("walk")
        self._addConfig("body")

        self._addComponent(Button, command=self.selectNextConfig, key="button_008", trigger="hold")
        self._addComponent(Button, command=self.selectPreviousConfig, key="button_009", trigger="hold")

        self._addComponent(Button, command=self.robot.incBodyPosition, key="button_004", mapper=MapperSetValue(dz=+5))
        self._addComponent(Button, command=self.robot.incBodyPosition, key="button_005", mapper=MapperSetValue(dz=-5))
        self._addComponent(Joystick, configs="walk", command=GaitSequencer().walk, keys=("analog_02", "analog_03", "analog_00"), mapper=MapperWalk())

        self._addComponent(Analog, configs="body", command=self.robot.setBodyExtraPosition, key="analog_00", mapper=MapperSetMultiply('yaw', coef=-15))
        self._addComponent(Analog, configs="body", command=self.robot.setBodyExtraPosition, key="analog_03", mapper=MapperSetMultiply('pitch', coef=-15))
        self._addComponent(Analog, configs="body", command=self.robot.setBodyExtraPosition, key="analog_02", mapper=MapperSetMultiply('roll', coef=15))

Same as before...

def main():
    def addGait(gaitClass, gaitName):
        group = settings.GAIT_LEGS_GROUPS[gaitName]
        params = settings.GAIT_PARAMS['default']
        for key1, value1 in settings.GAIT_PARAMS[gaitName].items():
            for key2, value2 in value1.items():
                params[key1][key2] = value2
        gait = gaitClass(gaitName, group, params)
        GaitManager().add(gait)

    scene = visual.display(width=settings.SCENE_WIDTH, height=settings.SCENE_HEIGHT)
    ground = visual.box(axis=(0, 1, 0), pos=(0, 0, 0), length=0.1, height=500, width=500, color=visual.color.gray(0.75), opacity=0.5)

    addGait(GaitTripod, "tripod")
    addGait(GaitTetrapod, "tetrapod")
    addGait(GaitRiple, "riple")
    addGait(GaitWave, "metachronal")
    addGait(GaitWave, "wave")
    GaitManager().select("riple")

    robot = Hexapod3D()

    remote = Gamepad(robot)

    GaitSequencer().start()
    remote.start()

    robot.setBodyPosition(z=30)

    robot.mainLoop()

    remote.stop()
    remote.join()
    GaitSequencer().stop()
    GaitSequencer().join()

if __name__ == "__main__":
    main()

The only difference, here, is the scene size init, and the ground addition (optional).

Settings

# -*- coding: utf-8 -*-

from py4bot.common import config


# Legs index
LEGS_INDEX = ('RF', 'RM', 'RR', 'LR', 'LM', 'LF')

# Legs geometry
LEGS_GEOMETRY = {
    'RM': {'coxa': 25, 'femur': 45, 'tibia': 65},
    'RF': {'coxa': 25, 'femur': 45, 'tibia': 65},
    'LF': {'coxa': 25, 'femur': 45, 'tibia': 65},
    'LM': {'coxa': 25, 'femur': 45, 'tibia': 65},
    'LR': {'coxa': 25, 'femur': 45, 'tibia': 65},
    'RR': {'coxa': 25, 'femur': 45, 'tibia': 65}
}

# Legs origin
LEGS_ORIGIN = {
    'RM': {'x':  35., 'y':   0., 'gamma0' :   0.},
    'RF': {'x':  35., 'y':  65., 'gamma0' :  30.},
    'LF': {'x': -35., 'y':  65., 'gamma0' : 150.},
    'LM': {'x': -35., 'y':   0., 'gamma0' : 180.},
    'LR': {'x': -35., 'y': -65., 'gamma0' : 210.},
    'RR': {'x':  35., 'y': -65., 'gamma0' : 330.},
}

# Legs feet neutral position
FEET_NEUTRAL = {
    'RM': LEGS_GEOMETRY['RM']['coxa'] + LEGS_GEOMETRY['RM']['femur'],
    'RF': LEGS_GEOMETRY['RF']['coxa'] + LEGS_GEOMETRY['RF']['femur'],
    'LF': LEGS_GEOMETRY['LF']['coxa'] + LEGS_GEOMETRY['LF']['femur'],
    'LM': LEGS_GEOMETRY['LM']['coxa'] + LEGS_GEOMETRY['LM']['femur'],
    'LR': LEGS_GEOMETRY['LR']['coxa'] + LEGS_GEOMETRY['LR']['femur'],
    'RR': LEGS_GEOMETRY['RR']['coxa'] + LEGS_GEOMETRY['RR']['femur'],
}

# Gaits
GAIT_LEGS_GROUPS = {
    'tripod':      (('RM', 'LF', 'LR'), ('RF', 'LM', 'RR')),
    'tetrapod':    (('RR', 'LM'), ('RF', 'LR'), ('RM', 'LF')),
    'riple':       (('RR',), ('LM',), ('RF',), ('LR',), ('RM',), ('LF',)),
    'metachronal': (('RR',), ('LM',), ('RF',), ('LR',), ('RM',), ('LF',)),
    'wave':        (('RR',), ('RM',), ('RF',), ('LR',), ('LM',), ('LF',))
}

# Gaits parameters
GAIT_PARAMS = {
    'default': {
        'length': {
            'min': 5.,
            'max': 40.
        },
        'angle': {
            'min': 0.5,
            'max': 5.
        },
        'height': {
            'min': 20.,
            'max': 40.
        },
        'speed': {
            'min': 25.,
            'max': 250.
        }
    },
    'tripod': {},
    'tetrapod': {},
    'riple': {},
    'metachronal': {},
    'wave': {}
}

# Gamepad path
THRUSTMASTER_PATH = "/dev/input/by-id/usb-Mega_World_Thrustmaster_dual_analog_3.2-event-joystick"

# VPython scene default size
SCENE_WIDTH = 640
SCENE_HEIGHT = 480

We removed the servos mapping/calibration, as we don't need them anymore, and we just added the scene default size.

Tutorial 3: Servos calibration

There are different reasons why we need to calibrate servos, depending on:

• how servos are attached to the robot
• what servos we actually use
• how accurate servos are

py4bot-gui-servocal.py script can be used to fine tune our servos calibration, and generate the SERVOS_CALIBRATION dict.

Most servos have a range of 180° (more or less), and their neutral position (mid-range) is around 1500µs. It is important to mount the servos on the robot so the we can reach the joints positions ranges we want.

The following picture shows the usual joints positions when servos are at neutral, for a 3 DoF leg:

Looking how Py4bot internally manages angles (UserGuideGit#a3DoFleg), we can see that matching joints positions are:

• coxa = 0°
• femur = 180°
• tibia = 90°

This means that there will be some offsets between the servo angles and the framework angles.

Another thing to tune is the servo neutral position, to correct mechanical errors.

Last thing to tune is the ratio between pulse variation and real servo displacement. This ratio varies a lot from a brand to another. And even for a same model, it varies from a servo to another. This is especially true for low cost servos.

Offsets and ratios are defined in the SERVO_CALIBRATION table.

Here is how to use py4bot-gui-servocal.py.

We first need to enable the servo, by clicking the Enable checkbox. Then, we need to check if the servo rotates the right direction: increasing the pulse value should increase the joint angle (= make in turn in the trigonometric sense in the leg coordinate system). If it is not the case, we just click the Invert checkbox (usually, coxa joints are all the same, and other joints should be inverted for legs of one side, but it depends if there is a symmetry in the mechanics or not).

Then, we can set the offset as shown above by moving the Offset slider. This has no impact on the servo position, but changes the table offset.

Next, we need to fine tune the servo neutral position. To do this, we move the Neutral slider until we mechanically reach the correct neutral position.

Once this is done, we can adjust the ratio. This step is a little bit harder, as we need to measure the angle the joint really moves. The best way is to move up to +-90°, as it is visually easy to see, but it is not always mechanically possible, and we can restrict the range to +-45°. But we have to keep in mind that the larger the angle is, the better the accuracy will be.

So, we move the Test Angle slider to 45°. Then, we move the Ratio slider so the joint really reaches 45°. Finally, we can check if going to -45° moves the joint to the symmetrical position.

We of course need to repeat the entire procedure for each servo/joint, which can be tedious for a 4 DoF hexapod :o/ But after that, our robot should walk better ;o)

Note that if we already have a settings.py module, we can launch py4bot-gui-servocal.py from the directory containing that module in order to reload the previous values. And if the module contains a LEGS_SERVOS_MAPPING, the script will show the legs/joints names. So, it is better to start writing this param, before launching the script.

Tutorial 4: Add gamepad support

Files for this tutorial are available in ​py4bot/examples/tutorial_4/.

Here, we will see how to create a custom input frontend for a new gamepad.

Again, we create a new empty dir in our home dir:

$ mkdir tutorial_4
$ cd tutorial_4

TODO

Conclusion

Thanks for reading these tutorials! There are many additional things to do, and final implementation may change, according to feedback/suggestions I will get. But the core is there. Again, the goal of this framework is to provide a high level tool to build complete and powerfull applications to control multi-legs robots.

Have a look at ​py4bot/examples/cronos/, which contains the code for my 4DoF hexapod, and closely follows Py4bot devs.