Discrete-event simulator in javascript

js-simulator

https://github.com/chen0040/js-simulator

Inspired by the MASON Multiagent Simulation library in Java, js-simulator is a general-purpose discrete-event multiagent simulator for agent-based modelling and simulation. It was written entirely in Javascript.

 

Install

Run the following npm command to install

npm install js-simulator

Demo

The following HTML demo is available:

• Flocking Boids HTML DEMO
• Conway's Game of Life HTML DEMO
• School Yard HTML DEMO
• L-System (Fractal Plant) HTML DEMO
• L-System (Dragon Curve) HTML DEMO
• Ant System HTML DEMO
• Particle Swarm HTML DEMO
• Virus HTML DEMO
• (Spring) Balls HTML DEMO
• Solar System HTML DEMO
• Near Beer Game HTML DEMO

More demo and HTML GUI supports will be added in the subsequent releases.

Usage

Create and schedule discrete events or agents

The discrete-event simulator is managed via the Scheduler class, which can be created as shown below:

jssim = require('js-simulator');
var scheduler = new jssim.Scheduler();

The scheduler schedules and fires events based on their time and rank (i.e. the order of the event) spec.

To schedule the an event to fire at a particular time:

var rank = 1; // the higher the rank, the higher the priority assigned and the higher-rank event will be fired first for all events occurring at the same time interval
var evt = new jssim.SimEvent(rank);
evt.id = 20; 
evt.update = function(deltaTime) {
    console.log('event [' + this.id + '] with rank ' + this.rank + ' is fired at time ' + this.time);
    
    // the code below allows the evt to send message to another agent (i.e., the agent referred by receiver variable)
    /*
    var receiver_id = receiver.guid()
    this.sendMsg(receiver_id, {
        content: "Hello"
    });
    */
    
    // the code below allows the evt to process messages sent from another agent
    /*
    var messages = this.readInbox();
    for(var i = 0; i < messages.length; ++i){
        var msg = messages[i];
        var sender_id = msg.sender;
        var recipient_id = msg.recipient; // should equal to this.guid()
        var time = msg.time;
        var rank = msg.rank; // the messages[0] contains the highest ranked message and last messages contains lowest ranked
        var content = msg.content; // for example the "Hello" text from the sendMsg code above
    }
    */
};

var time_to_fire = 10; // fire this event at time = 10
scheduler.schedule(evt, time_to_fire);

The main logic for an event is defined in its update(deltaTime) method, as shown in the code above. Events with higher ranks and earlier time_to_fire will always be executed first by the scheduler.

An event can also be sheduled to fire at a later time from the current time (e.g., such an event can be fired within another event):

var delta_time_later = 10; // the event will be fired 10 time units from now, where now refers to the current scheduler time
scheduler.scheduleOnceIn(evt, delta_time_later);

In terms of multi-agent system, an event can be thought of as an agent. Such an agent may need to execute repeatedly. In the js-simulator, this is achieved by firing an event repeatedly at a fixed interval:

var interval = 2; // time interval between consecutive firing of the event
var start_time = 12; // time to fire the event for the first time
scheduler.scheduleRepeatingAt(evt, start_time, interval);

If the start_time is at from the start of the simulation, then the above scheduling can also be replaced by:

scheduler.scheduleRepatingIn(evt, interval);

Execute the scheduler loop for the main discrete-event simulation

After the events/agents are scheduled, they are not fired immediately but only fired when scheduler.update() method is called, each call to scheduler.update() to move the time forward. At each time forwarded, events with higher rank will be executed (by calling their update(delaTime) method) first. Also events with the same rank will be shuffled before execution.

The scheduler can be executed in the following loop:

while(scheduler.hasEvents()) {
    var evts_fired = scheduler.update();
}

The above will run until no more events to fire in the scheduler, to stop the scheduler at a particular instead, use the following loop:

while(scheduler.current_time < 20) { // stop the scheduler when current scheduler time is 20
    scheduler.update();
}

The current scheduler time can be obtained by calling (this is useful if we want to know the current time inside the scheduler):

var current_scheduler_time = scheduler.current_time;

Sample Codes

Flocking behavior Demo

The source code below shows how to create a flocking of 15 boids (12 preys and 3 predators) that demonstrate the flocking principles:

Firstly we will declare a Boid class the inherits from the jsssim.SimEvent class, which defines the behavior of a single boid:

var jssim = require('js-simulator');


var Boid = function(id, initial_x, initial_y, space, isPredator) {
    var rank = 1;
    jssim.SimEvent.call(this, rank);
    this.id = id;
    this.space = space;
    this.space.updateAgent(this, initial_x, initial_y);
    this.sight = 75;
    this.speed = 12;
    this.separation_space = 30;
    this.velocity = new jssim.Vector2D(Math.random(), Math.random());
    this.isPredator = isPredator;
    this.border = 100;
};

Boid.prototype = Object.create(jssim.SimEvent);
Boid.prototype.update = function(deltaTime) {
    var boids = this.space.findAllAgents();
    var pos = this.space.getLocation(this.id);

    if(this.isPredator) {
        var prey = null;
        var min_distance = 10000000;
        for (var boidId in boids)
        {
            var boid = boids[boidId];
            if(!boid.isPredator) {
                var boid_pos = this.space.getLocation(boid.id);
                var distance = pos.distance(boid_pos);
                if(min_distance > distance){
                    min_distance = distance;
                    prey = boid;
                }
            }
        }

        if(prey != null) {
            var prey_position = this.space.getLocation(prey.id);
            this.velocity.x += prey_position.x - pos.x;
            this.velocity.y += prey_position.y - pos.y;
        }
    } else {
        for (var boidId in boids)
        {
            var boid = boids[boidId];
            var boid_pos = this.space.getLocation(boid.id);
            var distance = pos.distance(boid_pos);
            if (boid != this && !boid.isPredator)
            {
                if (distance < this.separation_space)
                {
                    // Separation
                    this.velocity.x += pos.x - boid_pos.x;
                    this.velocity.y += pos.y - boid_pos.y;
                }
                else if (distance < this.sight)
                {
                    // Cohesion
                    this.velocity.x += (boid_pos.x - pos.x) * 0.05;
                    this.velocity.y += (boid_pos.y - pos.y) * 0.05;
                }
                if (distance < this.sight)
                {
                    // Alignment
                    this.velocity.x += boid.velocity.x * 0.5;
                    this.velocity.y += boid.velocity.y * 0.5;
                }
            }
            if (boid.isPredator && distance < this.sight)
            {
                // Avoid predators.
                this.velocity.x += pos.x - boid_pos.x;
                this.velocity.y += pos.y - boid_pos.y;
            }
        }
    }


    // check speed
    var speed = this.velocity.length();
    if(speed > this.speed) {
        this.velocity.resize(this.speed);
    }

    pos.x += this.velocity.x;
    pos.y += this.velocity.y;

    // check boundary
    var val = this.boundary - this.border;
    if (pos.x < this.border) pos.x = this.boundary - this.border;
    if (pos.y < this.border) pos.y = this.boundary - this.border;
    if (pos.x > val) pos.x = this.border;
    if (pos.y > val) pos.y = this.border;
        
    console.log("boid [ " + this.id + "] is at (" + pos.x + ", " + pos.y + ") at time " + this.time);
};

Once the boid is defined we can then create and schedule the flocking event simulator using the code below:

var scheduler = new jssim.Scheduler();
scheduler.reset();

var space = new jssim.Space2D();
for(var i = 0; i < 15; ++i) {
    var is_predator = i > 12;
    var boid = new Boid(i, 0, 0, space, is_predator);
    scheduler.scheduleRepeatingIn(boid, 1);
}

while(scheduler.current_time < 20) {
  scheduler.update();
}

Conway's Game of Life

The sample code below shows how to create the game of life simulation:

var jssim = require('js-simulator');

var CellularAgent = function(world) {
  jssim.SimEvent.call(this);
  this.world = world;
}; 

CellularAgent.prototype = Object.create(jssim.SimEvent.prototype);
CellularAgent.prototype.update = function (deltaTime) {
  var width = this.world.width;
  var height = this.world.height;
  var past_grid = this.world.makeCopy();
  for(var i=0; i < width; ++i) {
      for(var j = 0; j < height; ++j) {
          var count = 0;
          for(var dx = -1; dx < 2; ++dx) {
              var x = i + dx;
              if (x >= width) {
                  x = 0;
              }
              if (x < 0) {
                  x = width - 1;
              }
              for(var dy = -1; dy < 2; ++dy) {
                var y = j + dy;
                  if(y >= height) {
                      y = 0;
                  }
                  if(y < 0) {
                      y = height - 1;
                  }
                  count += past_grid.getCell(x, y);
              }
          }
          if (count <= 2 || count >= 5) {
              this.world.setCell(i, j, 0); // dead
          }
          if (count == 3) {
              this.world.setCell(i, j, 1); // live
          }
      }
  }
};

var scheduler = new jssim.Scheduler();
var grid = new jssim.Grid(640, 640);

scheduler.reset();
grid.reset();

grid.setCell(1, 0, 1);
grid.setCell(2, 0, 1);
grid.setCell(0, 1, 1);
grid.setCell(1, 1, 1);
grid.setCell(1, 2, 1);
grid.setCell(2, 2, 1);
grid.setCell(2, 3, 1);

scheduler.scheduleRepeatingIn(new CellularAgent(grid), 1);

while(scheduler.current_time < 20) { // this assumes that we want to terminate at time 20
  scheduler.update();
}

School Yard Demo

The sample code below shows the school yard demo:

var Student = function(id, yard, network) {
   jssim.SimEvent.call(this);
   this.id = id;
   this.yard = yard;
   this.network = network;
   this.MAX_FORCES = 3.0;
   this.forceToSchoolMultiplier = 0.01;
   this.randomMultiplier = 0.1;
};
Student.prototype = Object.create(jssim.SimEvent.prototype);
Student.prototype.update = function(deltaTime) {
   var students = this.yard.findAllAgents();


   var me = this.yard.getLocation(this.id);

    var sumForces = new jssim.Vector2D(0, 0);
    var forceVector = new jssim.Vector2D(0, 0);
    var edges = this.network.adj(this.id);
    var len = edges.length;
    for (var buddy = 0; buddy < len; ++buddy)
    {
        var e = edges[buddy];
        var buddiness = e.info;


        var him = this.yard.getLocation(e.other(this.id));

        if (buddiness >= 0)
        {
            forceVector.set((him.x - me.x) * buddiness, (him.y - me.y) * buddiness);
            if (forceVector.length() > this.MAX_FORCES)
            {
                forceVector.resize(this.MAX_FORCES);
            }
        }
        else
        {
            forceVector.set((me.x - him.x) * buddiness, (me.y - him.y) * buddiness);
            if (forceVector.length() > this.MAX_FORCES)
            {
                forceVector.resize(0);
            }
            else if(forceVector.length() > 0)
            {
                forceVector.resize(this.MAX_FORCES - forceVector.length());
            }
        }

        sumForces.addIn(forceVector);
    }

    sumForces.addIn(
        new jssim.Vector2D((this.yard.width * 0.5 - me.x) * this.forceToSchoolMultiplier, (this.yard.height * 0.5 - me.y) * this.forceToSchoolMultiplier)
        );

    sumForces.addIn(
        new jssim.Vector2D(this.randomMultiplier * (Math.random() * 1.0 - 0.5), this.randomMultiplier * Math.random() * 1.0 - 0.5));

    sumForces.addIn(me);

    me.x = sumForces.x;
    me.y = sumForces.y;

    console.log("Student " + this.id + " is at (" + me.x + ", " + me.y + ") at time " + this.time);
};

var scheduler = new jssim.Scheduler();
var yard = new jssim.Space2D();
var network = new jssim.Network(30);
yard.width = 50;
yard.height = 50;

scheduler.reset();
yard.reset();
network.reset();
for(var i=0; i < 30; ++i) {
   var student = new Student(i, yard, network);
   yard.updateAgent(student, Math.random() * yard.width, Math.random() * yard.height);
   scheduler.scheduleRepeatingIn(student, 1);
}

var buddies = {};
for (var i = 0; i < 30; ++i)
{
    var student = i;
    var studentB = i;
    do
    {
        studentB = Math.floor(Math.random() * 30);
    } while (student == studentB);
    var buddiness = Math.random();
    if(!network.connected(student, studentB)){
        network.addEdge(new jssim.Edge(student, studentB, buddiness));
    }

    var studentB = i;
    do
    {
        studentB = Math.floor(Math.random() * 30);
    } while (student == studentB);

    buddiness = Math.random();
    if(!network.connected(student, studentB)){
        network.addEdge(new jssim.Edge(student, studentB, buddiness));
    }
}

while (scheduler.current_time < 2) {
   scheduler.update();
}

Open Source: MLP using Back-propagation learning in Java

Package provides java implementation of multi-layer perceptron neural network with back-propagation learning algorithm

https://github.com/chen0040/java-ann-mlp

Install

Add the following dependency to your POM file:

<dependency>
  <groupId>com.github.chen0040</groupId>
  <artifactId>java-ann-mlp</artifactId>
  <version>1.0.1</version>
</dependency>

Usage

MLP Regression

The training of mlp regression is done by calling:

mlpRegression.fit(trainingData);

The regression of mlp is done by calling:

double predicted = mlpRegression.transform(dataRow);

The sample code below shows how to use the MLP regression to predict the (-1, +1) numerical output of the heart-scale sample from libsvm.

The training data is loaded from a data frame connected to the "heart_scale" libsvm file (please refer to here for more example on how to create a data frame).

InputStream inputStream = new FileInputStream("heart_scale");

DataFrame dataFrame = DataQuery.libsvm().from(inputStream).build();

System.out.println(dataFrame.head(10));

TupleTwo<DataFrame, DataFrame> miniFrames = dataFrame.shuffle().split(0.9);

DataFrame trainingData = miniFrames._1();
DataFrame crossValidationData = miniFrames._2();

MLPRegression regression = new MLPRegression();
regression.setHiddenLayers(8);
regression.setEpoches(1000);
regression.fit(trainingData);

for(int i = 0; i < crossValidationData.rowCount(); ++i){
 double predicted = regression.transform(crossValidationData.row(i));
 double actual = crossValidationData.row(i).target();
 logger.info("predicted: {}\texpected: {}", predicted, actual);
}

MLP Multi-Class Classifier

The training of mlp classifier is done by calling:

mlpClassifier.fit(trainingData);

The classification of mlp is done by calling:

String predicted = mlpClassifier.classify(dataRow);

The sample code below shows how to use the MLP classifier to predict the labels of the heart-scale sample from libsvm.

The training data is loaded from a data frame connected to the "heart_scale" libsvm file (please refer to here for more example on how to create a data frame).

As the heart-scale data frame has (-1, +1) numerical output, the codes first coverts the (-1, +1) as categorical output label "category-label".

InputStream inputStream = new FileInputStream("heart_scale");

DataFrame dataFrame = DataQuery.libsvm().from(inputStream).build();

dataFrame.unlock();
for(int i=0; i < dataFrame.rowCount(); ++i){
 DataRow row = dataFrame.row(i);
 row.setCategoricalTargetCell("category-label", "" + row.target());
}
dataFrame.lock();

MLPClassifier mlpClassifier = new MLPClassifier();
mlpClassifier.setHiddenLayers(6); // one hidden layer, to set two or more hidden layer call mlpClassifier.setHiddenLayer([layer1NeuronCunt], [layer2NeuronCunt], ...);
mlpClassifier.setEpoches(1000);
mlpClassifier.setLearningRate(0.2);
mlpClassifier.fit(dataFrame);

int correctnessCount = 0;
for(int i = 0; i < dataFrame.rowCount(); ++i){
 DataRow row = dataFrame.row(i);



 String predicted_label = mlpClassifier.classify(row);
 correctnessCount += (predicted_label.equals(row.categoricalTarget()) ? 1 : 0);

 if(i < 10) {
    System.out.println(row);
    System.out.println("predicted: " + predicted_label + "\tactual: " + row.categoricalTarget());
 }
}

System.out.println("Prediction Accuracy: "+(correctnessCount * 100 / dataFrame.rowCount()));