artificialintelligencetoolkit / jyrobot Goto Github PK

View Code? Open in Web Editor NEW

0.0 4.0 2.0 11.33 MB

Python 99.96% Makefile 0.04%

jyrobot's Introduction

jyrobot

A lightweight Python robot simulator for Jupyter Lab, Notebooks, and other environments.

Goals

A lightweight mobile robotics simulator
Usable in the classroom, research, or exploration
Explore wheeled robots with range, cameras, and light sensors
Operate quickly without a huge amount of resources
Create reproducible experiments
Designed for exposition, experimentation, and analysis
Sensors designed for somewhat realistic problems (such as image recognition)
Especially designed to work easily with Machine Learning and Artificial Intelligence systems

Examples

There are pre-designed simulations ready to run, like this:

import jyrobot
import random

world = jyrobot.load("two-scribblers")

for robot in world.robots:
    # Give each robot a desired speed:
    robot.forward(1)

def control(world):
    for robot in world.robots:
        if robot.stalled:
	    # If stuck, just reverse:
            robot.reverse()
	# Turn randomly:
        robot.turn(1 - random.random() * 2)

# Watch the robots move in real time, or faster:
world.watch()
world.run(control, show=True, real_time=False)
# Press Control+C or interrupt the kernel to stop

You can also easily assemble your own simulations, robots, and sensors.

import jyrobot

world = jyrobot.World(width=100, height=100)
world.watch()

robot = jyrobot.Scribbler()

world.add_robot(robot)

robot.add_device(jyrobot.RangeSensor())
robot.add_device(jyrobot.Camera())

world.save_as("world-1")

Installation

For the core operations, you will need to install just jyrobot:

pip install jyrobot

For just image processing on top of of the core, you will need:

Pillow - Python Image Library (PIL)

For the full set of options, you will need:

Pillow - Python Image Library (PIL)
ipywidgets
IPython
bqplot

There are three different backends:

"pil" - requires Pillow (Python Image Library, PIL), the default; best tested
"canvas" - requires ipycanvas and numpy; some issues
"svg" - requires svgwrite and cairosvg (for backend.take_picture()); some issues

You can install all of the above with conda or pip.

To use the Jupyter enhancements, you'll also need the browser-based extensions. You can install those with:

jupyter labextension install @jupyter-widgets/jupyterlab-manager ipycanvas bqplot

If not in a conda environment, then you will also need to:

jupyter nbextension enable --py widgetsnbextension

jyrobot's People

Watchers

Forkers

hashb tce

jyrobot's Issues

enable look up of robot devices by user-defined names

I have a robot with 2 RangeSensors and 2 LightSensors, here's what the info() method returns:
robot.info()
device[0 or 'ir']: <RangeSensor 'left-ir' direction=0.0, range=20, width=57.3>
device[1 or 'ir']: <RangeSensor 'right=ir' direction=0.0, range=20, width=57.3>
device[2 or 'light']: <LightSensor 'left-light' position=(6, -6)>
device[3 or 'light']: <LightSensor 'right-light' position=(6, 6)>

Could we allow lookup by the user-defined name?

world.steps not quitting simulation when a controller returns True

world.steps prints stop message after each execution

In doing nnet experiments with robots, the typical workflow will be to call world.steps repeatedly, these stop messages become unwieldy in the jupyter notebook. For example:

Simulation stopped at: 00:00:01.0; speed 18.83 x real time
Simulation stopped at: 00:00:02.0; speed 71.27 x real time
Simulation stopped at: 00:00:03.0; speed 76.6 x real time
Simulation stopped at: 00:00:04.0; speed 59.33 x real time
Simulation stopped at: 00:00:05.0; speed 76.48 x real time
Simulation stopped at: 00:00:06.0; speed 21.17 x real time
...

World.reset() should update the elapsed time

Resetting the world should reset the time to 0.

RangeSensor should show it's position

When you print a RangeSensor it only shows it's angle, range, and width. But the position is important to determining which of the possible multiple range sensors this represents,

robot.stalled not getting reset

After one controller ends with a stalled robot, if you do a set_pose to a new location, the robot.stalled remains True.

robot.get_pose() returning directions that are out of the range [0,360]

Should the direction of the robot always be reported in the range [0,360]?

Triggered error, robot not in world, is incorrect

When executing the following code I get an error message: This robot is not in the world

world = jyrobot.World(width=250, height=100)
robot = jyrobot.Scribbler()
robot.add_device(jyrobot.RangeSensor(width=0, max=100))
robot.set_pose(120, 50, 180)
world.add_robot(robot)

But the x, y, direction of the set_pose command are within the boundaries of the world

message from world.set_seed is misleading

Always says "reusing" even if this is the first time using that seed. Would be better to say "using seed ..."

world.watch() doesn't show robot until world.reset() is done

Whenever I try to watch a world, the robot doesn't show up until I do world.resetI(). Here's an example world definition:

Create a world with one wall and a robot

world = jyrobot.World(width=200, height=200)
world.add_wall("blue", 80, 50, 100, 150)
robot = jyrobot.Scribbler(x=50, y=100, direction=0)
robot.add_device(jyrobot.RangeSensor(position=(6,-6),width=57.3,max=20,angle=90,name="left-ir"))
robot.add_device(jyrobot.RangeSensor(position=(6,6),width=57.3,max=20,angle=90,name="right=ir"))
world.add_robot(robot)
world.update()
world.save()

robot.speak continues from old controller to new controller

If you have multiple controllers in the same notebook, if the previous one was speaking, then the next one will continue to say the last message. Should reset the speak message at the start of every control session.

RangeSensor should have getter for max

Add a get_max method to the RangeSensor. Knowing the max is useful when testing when to respond to the sensor.