About The Project

This library implements a programmable simulation and top-down visualization for a simple land vehicle. The library was created as teaching material in the Internet of Things (IOT) project for and by the students of Berlin School of Design and Communication.

Motivation

In the IOT project students build a general purpose rover platform. Some usage scenarios students realize are the autonomous search for land mines or long-term research observations in earth environments with rough climate conditions such as (ice) deserts.

A central challenge in this project is the implementation of routing and navigation in the presence of static obstacles too small to be marked on maps (e.g. stones and rocks). Testing and debugging these algorithms in realtime on hardware is inevitable, but also slow and expensive. For this reason we were searching for a simulation environment with the following features:

• fully programmable control loop
• vehicle with either chain drive or six steerable wheels
• open world
• simulated location sensor providing real latitude and longitude values
• simulated proximity sensor
• simulated compass
• simulated measurement "errors" on all sensor values
• implementation with an easy to learn script language, preferably JavaScript/Typescript

After some research we came to the conclusion our requirements are too special and created our own simulation. Here it is.

Built With

The rover simulation uses the following 3rd party libraries:

• Geodesy
• P2

P2 seems not to be actively maintained and is also not really suited for simulation with natural gravity on Z-axis. We could replace this dependency sooner or later.

Installation

To use the simulation create a new web project including a bundler like Webpack and install the package from github.

with npm

npm install github:davidlinner/rover

Proper NPM package about to follow...

Usage

Basic example:

import { ControlLoop, Simulation, LocationOfInterest, RoverType } from 'rover';

// Create a simple control function which only tells the engines to go forward, left a bit slower, so we drive in a circle
// The control function will be called periodically in the control loop
const loop: ControlLoop = ({ location, heading, clock }, { engines, steering }) => {
	return {
		engines: [0.8, 0.8],
	};
};

// Create a new simulation
const simulation = new Simulation({
	// add above loop
	loop,

	// type of vehicle (tank or rover); tank expects two engine argument left and right
	vehicleType: VehicleType.tank,

	// define the origin offset of the rover on the planet,
	origin: {
		latitude: 52.477050353132384,
		longitude: 13.395281227289209,
	},

	// select the DOM element the visualization canvas should be appended to as child
	element: document.querySelector('main'),

	// define a list of static points to be rendered in the visualization (nice for waypoints)
	locationsOfInterest: [
		{
			latitude: 52.47880703639255,
			longitude: 13.395281227289209,
			label: 'A',
		},
	],

	// set rending options like size and colors
	// consult the API docs for all switches and params
	renderingOptions: {
		width: 500, // width of created canvas
		height: 500, // height of created canvas
	},
});

// as soon as the simulation is created the visualization is rendered,
// but have to explicitely start the control loop
simulation.start();

Adding measurement errors

Additionally, you can configure a more realistic behavior for the simulation by setting the physicalContraints property. Three constraint collections, AUTHENTICITY_LEVEL0, AUTHENTICITY_LEVEL1, AUTHENTICITY_LEVEL2 are predefined, while the first means no errors at all.

AUTHENTICITY_LEVEL1, in contrast, adds random measurement errors with equal distribution to heading and location and a little, random unbalance to the two engines. Since the random seeds are different for each run of the simulation error correction has to be added to the control loop.

AUTHENTICITY_LEVEL2 is similar to AUTHENTICITY_LEVEL1, but adds errors with a more realistic gaussian distribution.

Example with some environmental error:

// Import main Simulation class and error factory function
import { Simulation, AUTHENTICITY_LEVEL2, /*...*/ } from 'rover';
//...

// Create a new simulation
const simulation = new Simulation({
	//...
	physicalConstraints: AUTHENTICITY_LEVEL2,
    //...
});

Adding obstacles

Similar to locations of interest, obstacles can be added in advance to the simulation. Obstacles all have a circular shape, but the radius can be configured in meters. To create more irregularly shaped obstacles, create overlapping circles.

const simulation = new Simulation({
    //...
    obstacles: [{
        radius: 1,
        latitude: 52.477000353132384,
        longitude: 13.395281227289209
    }],
    //...
});

Please consult the API Docs for more details.

• API Docs

License

Distributed under the MIT License.