The Goal

When children approach the house the Pumpkin Patch comes to life providing an interactive environment based on motion and sound detection.

Service Layer

The following technologies were used:

ActorSystem:

Using Akka with a Finite State Machine DSL that various actors can implement to transition their internal state.

These actors facilitate the state of the IoT devices involved.

FSM code | FSM grammar code | preview actor code

Using the DSL is very easy.

1) Implement the FSM trait which provide the actor with an instance of a Finite State Machine.

class Projector extends BaseActor implements FSM {}

2) Configure the FSM DSL durring the initializtion of your actor.

    @Override
    void configureFsmDsl() {
	    fsm.record().onCommands([Play]).fromState(Any).goToState(Playing).transition = { Command command ->
	        remoteDispatch(command)
	    }

	    fsm.record().onCommands([Off, Pause]).fromState(Any).goToState(Paused).transition = { Command command ->
	        remoteDispatch(command)
	    }

	    fsm.record().onCommands([Resume]).fromState(Paused).goToState(Playing).transition = { Command command ->

	        if (!currentVideo || currentVideo == Video.Name.NONE)
	            return new Guard(reason: "ineligible video for command:Resume")

	        remoteDispatch(command)

	    }
    }

3) Fire relevent messages to your Finite State Machine in your actors onRecieve method.

	@Override
    void onReceive(Object message) throws Exception {      
        if (message instanceof Command)
            fsm.fire(message)            
    }

Android:

Enables playing and stopping of videos as well as device commands useing subtitles "timedText" to have fine grain control of IoT devices... i.e: blow smoke at 1:01:02.
android code | actor code

Arduino:

Sound Detection. arduino code | actor code
Smoke Machine. arduino code | actor code

Philips Hue Bloom(s) with Bridge:

Enables control of Hue lights [color/luminance] via a rest API. preview code

Materials / Application:

This provides a great platform for arranging your Pumpkin Patch. Attach the tarps to the side and back of the popup so that luminance levels can be controlled. You will also find the aluminum supporting structures useful for cable management and hanging props.

Cover one side of the pumpkin with blue painters tape and draw your spooky face. The tape serves two purposes:

1. 1) provides a platform for drawing your design one or more times.
2. 2) exposes freshly cut areas for spray painting black; while protecting untouched orange areas.

Stagger different sized hays bales for which to arrange pumpkins on.

Weave in and out of hay bales

Hang the webs from the ceiling and sides of the Easy Popup

Getting my cheapo projector to start playing the pumpkins singing on demand turned out to be a bit challenging. The problem being that the projector doesn’t go into a projecting state when you give it power. A button must be pushed. Once that button is pushed the projector OS loads and then you must navigate to a movie screen and pick your SD card media. In order to have my pumpkins singing on demand, I decided to configure the projector to play the pumpkin singing video continuously on a loop while using a networked Arduino “Arduino Yun” with a servo to cover or lift a piece of paper from the projector lense. This creates the illusion of on demand pumpkin singing.

Load a program on your arduino to take commands via rest and translate them into commands to the servo

I built a box from supplies I picked up at Lowes. I left space in between side panels so the equipment inside doesn’t overheat.

The ability to detect current audio levels is valuable in creating an interactive experience. Not only does this allow for the lights to have a Color / Luminance Schema that matches your music, they will also react to sounds of trick or treaters. I’ve added a bonus that blows smoke and powers the strobe light when the sum of the last 5 received audio values exceeds a certain level.

Like the projector; the smoke machine requires physical interaction to have smoke blow on demand. The smoke machine must be powered on for about 5 minutes so that it can heat up enough to blow smoke when a switch is toggled. I decided to alway supply power so that it is in a constant ready state and mount a servo to the blow smoke switch. Now when the Halloween Actor System is in a automation on state smoke will blow every 30 seconds for 2 seconds

Load a program on your arduino to take commands via rest and translate them into commands to the servo

Github: https://github.com/yakoob/IoT

The following technologies were used:

ActorSystem:

Controls the Android tablet as well as the Arduiono-Relay Module for automation of lighting/music using Grails, Akka, MQTT, Async-REST.
preview code

LightController:

Arduino application controlling a relay module by exposing an API for various lighting methods via REST.
preview code

MediaController:

Enables playing and stopping of songs via MQTT using android sdk.
preview code

Clone from Github:

https://github.com/yakoob/christmas

A continuation of my previous post... This introduction to a Finite State Machine DSL will show you how I am able to better manage event based transitional logic.

Lets focus on the InboundChannelActor "incoming call".

First we need to add an instance of a State Machine so that we can configure it when each inbound channel is initialized. I include a trait that allows me to have a new instance of a FiniteStateMachine, then I define the DSL which is added to the State Machine's transitions registry.

trait ChannelState implements Channel {
    public FiniteStateMachine fsm = FiniteStateMachine.newInstance(None)
}

class InboundChannelActor extends BaseActor implements InboundChannelState {

    public InboundChannelActor(String channelUuid, String destination) {

        this.channelUuid = channelUuid
        this.destination = destination

        // record a transition for Event: "Parked and All" from State: "None" to State: "Greeting"
        state_ChannelState__fsm.record().on([Parked, All]).from(None).to(None).act = {
            log.info("on Parked from None -> None")
            state_ChannelState__fsm.goToState(state_ChannelState__fsm.currentValue)
        }

        // record a transition for Event: "init" on a Greeting, 
        state_ChannelState__fsm.record().on([Init]).from(Greeting).to(Greeting).act = { event ->
            // play greeting
            audioActorService.actorRef.tell(new Shout(channelUuid: channelUuid, audioId: state_ChannelState__fsm.currentValue.greeting.id), context.self())
            // start timer for next action
            akkaService.system.scheduler().scheduleOnce(Duration.create(state_ChannelState__fsm.currentValue.nextStateSeconds, TimeUnit.SECONDS),
                new Runnable() {@Override public void run() {
		// go-to next state
                state_ChannelState__fsm.goToState(new Init(), state_ChannelState__fsm.currentValue.nextState)
            } }, akkaService.system.dispatcher())
        }

        // record a transition for Event: "Init" on a Menu
        state_ChannelState__fsm.record().on([Init]).from(Menu).to(Menu).act = { event ->
            // play menu greeting
            audioActorService.actorRef.tell(new Shout(channelUuid: channelUuid, audioId: state_ChannelState__fsm.currentValue.greeting.id), akkaService.actorNoSender())
        }

        // record a transition for Event: "KeyPressed" on a Menu,
        state_ChannelState__fsm.record().on([KeyPressed]).from(Menu).to(Menu).act = { event ->
	    // repeat the keys back to you... 
            speechActorService.actorRef.tell(new Speak(channelUuid: channelUuid, words: "dtmf ${event.digit}"), akkaService.actorNoSender())
        }

	// record a transition for Event: "Hungup" on All transitions... 
        state_ChannelState__fsm.record().on([Hungup]).from(All).to(All).act = { event ->
	    // stop the Actor
            // todo: stop child actors
            context.stop(context.self())
        }

        // Start the Finite State Machine with Configured Action
        state_ChannelState__fsm.goToState(VoiceState.findByE164(destination))
    }	
}

Lastly... add the Akka onRecieve() method for firing events of to the state machine where we check the transitions registry for a match and "act"

class InboundChannelActor extends BaseActor implements InboundChannelState {
	// akka's onRecieve method... fire events to the state machine for this inbound channel instance
	@Override
	void onReceive(Object message) throws Exception {
		state_ChannelState__fsm.fire(message, state_ChannelState__fsm.currentState)
	}
}

1. * Application built to scale up (concurrency)
2. * Application built to scale out (remoting)
3. * Avoid race conditions, deadlocks, live locks
4. * Fault tolerance.

So the main idea with Actor Model is instead of using shared memory, messages are sent to actors who's state can only be altered by its self. A actor has a mailbox for receiving its messages. When a actor sends a message to a mailbox it's non-blocking. When a actor receives a message in its mailbox its uninterrupted. Actors act on their mailbox messages in order retrospectively.

Lets dive into a real world example using my PBX demonstration. A call is parked on my freeSwitch. I will send a message to a SwitchBoardActor who's job is to setup up new calls or route existing calls.

class FreeSwitchConnectionService {
    switchboardActor.tell(parked, akkaService.actorNoSender())
}

My SwitchBoardActor receives the Parked() message and knows that Parked incoming messages should create a new InboundChannelActor informing it of the Parked message so the newly created InboundChannelActor can setup its' state to receive additional messages.

    class SwitchBoardActor extends UntypedActor implements ActorMaker {
        @Override
        void onReceive(Object message) throws Exception {
            if (message instanceof FreeSwitchEvent && message instanceof Parked) {
                makeInboundChannelActor(message.channelUuid, message.destination).tell(message, context.self())
            }
        }
    }

    class InboundChannelActor extends BaseActor implements ObjectBinding {
        @Override
        void onReceive(Object message) throws Exception {            
            if (message instanceof Parked && !message.variableActor && message.inbound {                
                // tell freeSwitch to set the variableActor on the freeSwitch channel so the next message get routed to the correct instance of InboundChannelActor
                freeSwitchActorService.actorRef.tell(copyObject(message, new SetVar(key: "actor", value: self.path().toString())), context.self())
            }
        }
    }

Lets have a look at subsequent messages that have the Actor Uri attached to it. i.e: DTMF events are sent into the Actor System. We decide if we want to do text to speech or play an audio.

    class FreeSwitchConnectionService {
        KeyPressed keyPressed = copyMap(fsEventMap, new KeyPressed())
        switchboardActor.tell(keyPressed, akkaService.actorNoSender())
    }

    class SwitchBoardActor extends UntypedActor implements ActorMaker {        
        @Override
        void onReceive(Object message) throws Exception {
            if (message instanceof FreeSwitchEvent && message.variableActor) {
                getContext().system().actorSelection(message.variableActor).tell(message, context.self())
            }
        }
    }

    class InboundChannelActor extends BaseActor implements ObjectBinding {

        def akkaService = Holders.applicationContext.getBean("akkaService")
        def freeSwitchActorService = Holders.applicationContext.getBean("freeSwitchActorService")
        def audioActorService = Holders.applicationContext.getBean("audioActorService")
        def speechActorService = Holders.applicationContext.getBean("speechActorService")
        private boolean inputDigitTimerStarted = false
        private inputDigits = []
        
    @Override
        void onReceive(Object message) throws Exception {            
            if (message instanceof KeyPressed) {
                /* put DTMF digit into an array and start a timer. 
        All digits in the inputDigits array will be used when timer expires.*/
                inputDigits << message.digit
                startInputDigitsTimer()
            }
        }
        
        def startInputDigitsTimer(){
                if (!inputDigitTimerStarted){
                    log.info( " INPUT DIGITS TIMER STARTED ")
                    inputDigitTimerStarted = true
                    akkaService.system.scheduler().scheduleOnce(Duration.create(3, TimeUnit.SECONDS),
                        new Runnable() {
                            @Override
                            public void run() {
                                def keysPressed = inputDigits.join("").toString()
                                inputDigits.clear()
                                log.info(" INPUT DIGITS ARE: $keysPressed")
                                if (keysPressed.length()) {
                                    if (keysPressed == "1"){
                                        speechActorService.actorRef.tell(new Speak(channelUuid: channelUuid, words: "Welcome" ), context.self())
                                    } else if (keysPressed == "2") {
                                        speechActorService.actorRef.tell(new Speak(channelUuid: channelUuid, words: "Patching you in to com systems" ), context.self())
                                    } else {
                                        audioActorService.actorRef.tell(new Shout(channelUuid: channelUuid, audioId: keysPressed), context.self())
                                    }
                                    inputDigitTimerStarted = false
                                }
                            }
                        }, akkaService.system.dispatcher()
                    )
                }
            }
    }

So thats the gist of my actor based PBX. Download the code on GitHub for the full picture.

You can test drive this code with WebRTC

Another thing worth noting. I'm using Grails 2.4.3 in this project which includes the latest major release of groovy which implements Traits. In the code above you see my actors implement Traits. Having the ability to do multiple inheritance is a very nice feature. i.e:

trait Channel {
    String channelUuid
    String destination
    String variableActor = null
}

class FreeSwitchEvent implements Serializable, Channel {
    Boolean inbound = false
    Boolean outbound = false
}

class Parked extends FreeSwitchEvent {}

It's like pouring in functionality

trait ObjectBinding {
    def copyObject(source, target) {
        source.properties.each { key, value ->
            if (target.hasProperty(key) && !(key in ['class', 'metaClass', '']))
                target[key] = value
        }
        return target
    }

    def copyMap(Map source, target) {
        source.each{ k, v ->
            try {
                target[k] = v
            } catch(e){log.error(e.message)}
        }
        return target
    }

I hope you enjoyed this solution. You can clone it here on github...

I will follow up on this with a fork directly to Scala. I will show you how case class pattern matching, Some/None; greatly reduce the code base.

1. * Extendable engine that can be used in any scenario.
2. * Typeless rules that can be added or removed at runtime. "convention over configuration"
3. * Decoupled decision that can ask for 1 or more rules engine results.

Design with a soldier's promotion/demotion implementation

Lets start with an interface to define the basic contract for a rules engine.

public interface IRulesEngine {
    def apply(obj)
}

Here is a concrete base implementation of our rules engine. All of our business derived rules engines will extend this. Methods with the word "Rule" at the end will be executed when calling the apply() method.

class RulesEngine implements IRulesEngine {

    def getRules() {
        def rules = []
        this.class.declaredFields.each {
            def field = this."${it.name}"
            if (!it.isSynthetic() && field instanceof Closure && it.name.endsWith("Rule")) {
                rules << it.name
            }
        }
        rules
    }

    def apply(obj) {
        Set responseSet = [] as Set
        rules.each { rule ->
            responseSet << this."$rule"(obj)
        }
        responseSet
    }
}

We can start off with 2 engines. Ones which return promotable and demoteable traits.

class PromotableTraitsRulesEngineService extends RulesEngine {

    // example of a rule that returns many traits
    def heroOnBattlefieldRule = { soldier ->

        def traits = []

        log.info("heroOnBattlefieldRule() ran")

        War*.campaigns*.battles.collect { battle -> battle.heros in soldier}.each { battle ->
            traits << """heroOnBattlefield - ${battle.campaign.war} ${battle.campaign}"""
        }

        return traits
    }
}

class DemotableTraitsRulesEngineService extends RulesEngine {

    // example of a rule that returns many traits
    def insubordinationOnLeaveRule = { soldier ->

        def traits = []

        log.info("insubordinationOnLeaveRule() ran")

        soldier.leaves.findAll{it.insubordination == true}.each { leave ->
            traits << """insubordinationOnLeave - ${leave.country} ${leave.city} ${leave.place}"""
        }

        return traits
    }

    // example of a rule that returns 1 traits
    def insubordinationDuringBattleRule = { soldier ->

        def traits = []

        log.info("insubordinationDuringBattleRule() ran")

        def hasInsubordinationDuringBattleRule = War*.campaigns*.battles*.insubordinates.find { insubordinate ->
            insubordinate.soldier == soldier
        }

        if (hasInsubordinationDuringBattleRule)
            traits << "insubordinationDuringBattle"

        return traits

    }
}

Now we can create a service that aggregates and applies these traits.

class PromotionRulesEngineService implements IRulesEngine {

    def promotableTraitsRulesEngineService
    def demotableTraitsRulesEngineService

    def apply( soldier ){
        
        def traits = []
        
        traits << promotableTraitsRulesEngineService.apply( soldier )
        traits << demotableTraitsRulesEngineService.apply( soldier )
        
        return traits
    }
}

Lets get on to our decision already...

class PromotionService {

    enum Decision {PROMOTE, DEMOTE, NONE, DISHONORABLE_DISCHARGE}

    def promotionRulesEngineService

    def decide(soldier) {

        def traits = promotionRulesEngineService.apply(soldier)

        if (traits.collect{it == "insubordinationDuringBattle"}.count())
            return Decision.DISHONORABLE_DISCHARGE

        if (traits.collect{it.contains "insubordinationOnLeave"}.count() < 10  
        && traits.collect{it.contains "heroOnBattlefield"}.count() > 0 )
            return Decision.PROMOTE

        if (traits.collect{it.contains "insubordinationOnLeave"}.count() > 3  
        && traits.collect{it.contains "heroOnBattlefield"}.count() == 0 )
            return Decision.DEMOTE

        return Decision.NONE
    }

}

	case class Heal(item:String)
	 
	trait Healing {
	    val handleHealingMessages = {
	        case Heal(item) => println("I healed with  " + item)
	    }
	}

	case class Eat(item:String)

	trait Eating {
	    val handleEatingMessages = {
	        case Eat(item) => println("I ate " + item)
	    }
	}
	 
	case class Attack(item:String)
	 
	class Soldier extends Actor with Healing with Eating {

	    def handleMessages = {
	        case Attack(item) => println("I'm attacking with my " + item)
	    }
	 
	    def act() = {
	        loop {
	            react {			 
		         /* order matters: if more than one trait handles the same message the first one in
			 the orElse chain wins.  I want my solder to choose "attack" over healing or eating */
	                handleMessages orElse handleHealingMessages orElse handleEatingMessages             
	            }
	        }
	    }
	}
	 
	val soldier = new Soldier
	soldier.start
	soldier ! Attack("battle riffle")
	soldier ! Attack("pistol")
	soldier ! Attack("knife")
	soldier ! Heal("xstat sponges")
	soldier ! Eat("field ration")
	

Click here to get your own drone

def list = ['a','b','a','c','a','a','d','e','d','d','d','d','d']

JAVA

Map map = new HashMap<>();
for (T t : list) {
    Integer val = map.get(t);
    map.put(t, val == null ? 1 : val + 1);
}
Entry max = null;
for (Entry e : map.entrySet()) {
if (max == null || e.getValue() > max.getValue())
    max = e;
}
assert 'd' == max.getKey();

JAVAish...

def m = [:]
list.each {m[it] = m.get(it,0)+1}
assert 'd' == m.max{ it.value }.key

GROOVY

assert 'd' == list.max{Collections.frequency(list, it)}

The FPV live video feed to your phone uses 2.4ghz Wifi. Connected to your remote is a small range extender which connects to a camera on the Phantom 2 Vision/+. You can increase your FPV range by upgrading the antennas on your WiFi repeater and/or Vision camera. It's important to note that these upgrades can be done in several phases. Each incrementally increasing your range.

Upgrading the "pilot side" Wifi repeater

• Electric Drill, Drill Press or a Dremel
• U.FL Mini PCI to RP-SMA Pigtail Antenna WiFi Cable - available at amazon
LHCP 2.4 GHz antenna. I use the LHCP 2.4 GHz Half Sphere Helix 9dBi available at fpvlr.com
• Industrial strength Velcro

Now your FPV should be doubled by upgrading "pilot side" internal antenna's.

Upgrading the "phantom side" Wifi

Upgrading Complete!!!

You might have noticed I have a pilot side remote antenna upgrade. I will update this post with the 5.8ghz remote "pilot/driver" upgrades soon...

Happy Flying!

Click here to get your own drone