libgdx / gdx-ai Goto Github PK

In IndexedAStarPathFinder#search there is this line:

// Terminate if we reached the goal node
if (current.node == endNode) return true;

This only works if nodes don't change identity. Depending on how the user implements methods such as getConnections, the nodes might change identity while still remaining logically equal (eg if the type N is the user's custom Coordinate class). I ran into this problem myself, but fixed it by caching my coordinate objects (which I should have been doing anyway), but it's potentially confusing for less experienced developers. If there isn't a compelling reason to to do otherwise, it might be better to replace the == with .equals.

You can reproduce in the tests, you will see that the sprite will turn instead of facing the direction he had when he arrived. I tested with different parameters.

@badlogic is the NOTICE file in this repo really the NOTICE file you want people to redistribute?

Please ensure you have given all the following requested information in your report.

Issue details

Please provide the details of your issue

Reproduction steps/code

Please provide the steps to reproduce this issue, or even better, a SSCCE that reproduces your issue.

Version of gdx-ai and/or relevant dependencies

Please provide the version(s) affected.

Stacktrace

//Please provide the stacktrace if applicable 

It is common case scenario when you want to trigger state change only if current state is different than new one:

if (!mStateMachine.isInState(PlayerState.RUN)) {
    mStateMachine.changeState(PlayerState.RUN);
}

What's about to add some default implementaion of the state machine with this check to be available out of the box? Something like this:

public class StrictStateMachine<E> extends DefaultStateMachine<E> {

    /**
     * @see DefaultStateMachine#DefaultStateMachine(E)
     */
     public StrictStateMachine(E owner) {
        super(owner, null, null);
    }

    /**
     * @see DefaultStateMachine#DefaultStateMachine(E, State)
     */
    public StrictStateMachine(E owner, State<E> initialState) {
        super(owner, initialState, null);
    }

    /**
     * @see DefaultStateMachine#DefaultStateMachine(E, State, State)
     */
    public StrictStateMachine(E owner, State<E> initialState, State<E> globalState) {
        super(owner, initialState, globalState);
    }

    @Override
    public void changeState(State<E> newState) {
        if (!isInState(newState)) {
            super.changeState(newState);
        }
    }
}

MessageDispatcher doesn't work with system resources so making it a Singleton is an anti-pattern, since it heavily limits it's usage.
http://stackoverflow.com/questions/137975/what-is-so-bad-about-singletons

It makes perfect sense, for example, to have separate MessageDispatchers for different parts of the app - one for ui, other one for ai - cause that makes message constants easier to maintain as they aren't clutter together, and it makes it possible to configure time granularity differently for different dispatchers.

So I really like what where the AI extension stands and would love to try some features out. However it seems to me like there is no documentation beyond the tests on how someone would implement something like say a steering behavior. It's nearly impossible for me to pull working code to demo in a new project from the tests. How would I set up say a simple LookWhereYourGoingTest in a new libgdx project? I'm sorry if I am missing something really obvious just need some help.

Like square Otto bus @provide annotation (and Guava event bus), it would be great if we can define Providers that will send a Telegram automatically when listeners registered for a msg :

MessageDispatcher.getInstance().addProviders(this, int ... msgs)

interface Provider {
   Telegraph provides(int msg);
}

When registering msgs, listeners will get Telegraph from providers registered for the given msgs.

In games, entities are living : they are created, destroyed, evolving, changing states, ... When communicating together through Telegrams some newly created entities will not get notified of some important Telegram sent by other entities just before they register. With providers, they will get the last emitted Telegram of the desired type as soon as they register MessageDispatcher.getInstance().addListeners.... REDUCING COUPLING BETWEEN ENTITIES A LOT !

Id like to start discussion about this topic, that im not even sure how to title correctly.
Take for example the pfa tests. We have TiledNodes of certain types. We pathfind on TILE_FLOOR. What if we want have units that can only walk on TILE_WALL or on both?
What would be optimal way of implementing ability to select on what we pathfind? Should something like this even be supported by the library, or should it be left to the user to implement?

Looks like the cause is the new GdxFileSystem. Here is the output that I get when trying to compile for GWT:

16:00:58.902 [QUIET] [system.out]    Validating units:
16:00:58.906 [QUIET] [system.out]       [ERROR] Errors in 'jar:file:/Users/baz/.gradle/caches/modules-2/files-2.1/com.badlogicgames.gdx/gdx-ai/1.7.0/fedc63e153ddbd33ea94c1ed379f811c6688269d/gdx-ai-1.7.0-sources.jar!/com/badlogic/gdx/ai/GdxFileSystem.java'
16:00:58.906 [QUIET] [system.out]          [ERROR] Line 41: No source code is available for type com.badlogic.gdx.assets.loaders.resolvers.AbsoluteFileHandleResolver; did you forget to inherit a required module?
16:00:58.906 [QUIET] [system.out]          [ERROR] Line 49: No source code is available for type com.badlogic.gdx.assets.loaders.resolvers.LocalFileHandleResolver; did you forget to inherit a required module?
16:00:58.906 [QUIET] [system.out]       [ERROR] Errors in 'jar:file:/Users/baz/.gradle/caches/modules-2/files-2.1/com.badlogicgames.gdx/gdx-ai/1.7.0/fedc63e153ddbd33ea94c1ed379f811c6688269d/gdx-ai-1.7.0-sources.jar!/com/badlogic/gdx/ai/StandaloneFileSystem.java'
16:00:58.906 [QUIET] [system.out]          [ERROR] Line 69: The constructor FileHandle(String, Files.FileType) is undefined
16:00:58.906 [QUIET] [system.out]          [ERROR] Line 73: The constructor FileHandle(File, Files.FileType) is undefined
16:00:58.906 [QUIET] [system.out]          [ERROR] Line 77: file cannot be resolved
16:00:58.906 [QUIET] [system.out]          [ERROR] Line 77: type cannot be resolved to a variable
16:00:58.907 [QUIET] [system.out]          [ERROR] Line 78: file cannot be resolved to a variable
16:00:58.907 [QUIET] [system.out]          [ERROR] Line 78: type cannot be resolved to a variable
16:00:58.907 [QUIET] [system.out]          [ERROR] Line 82: file cannot be resolved
16:00:58.907 [QUIET] [system.out]          [ERROR] Line 83: file cannot be resolved
16:00:58.907 [QUIET] [system.out]          [ERROR] Line 83: type cannot be resolved to a variable
16:00:58.907 [QUIET] [system.out]          [ERROR] Line 87: file cannot be resolved
16:00:58.907 [QUIET] [system.out]          [ERROR] Line 89: type cannot be resolved to a variable
16:00:58.907 [QUIET] [system.out]          [ERROR] Line 94: type cannot be resolved to a variable
16:00:58.907 [QUIET] [system.out]          [ERROR] Line 98: type cannot be resolved to a variable
16:00:58.908 [QUIET] [system.out]          [ERROR] Line 98: file cannot be resolved
16:00:58.908 [QUIET] [system.out]          [ERROR] Line 99: type cannot be resolved to a variable
16:00:58.908 [QUIET] [system.out]          [ERROR] Line 99: file cannot be resolved
16:00:58.908 [QUIET] [system.out]          [ERROR] Line 100: file cannot be resolved to a variable
16:00:58.908 [QUIET] [system.out]       [ERROR] Errors in 'jar:file:/Users/baz/.gradle/caches/modules-2/files-2.1/com.badlogicgames.gdx/gdx-ai/1.7.0/fedc63e153ddbd33ea94c1ed379f811c6688269d/gdx-ai-1.7.0-sources.jar!/com/badlogic/gdx/ai/btree/utils/DistributionAdapters.java'
16:00:58.909 [QUIET] [system.out]          [ERROR] Line 463: The method countTokens() is undefined for the type StringTokenizer
16:00:58.950 [QUIET] [system.out]    Removing invalidated units
16:00:59.004 [QUIET] [system.out]       Compilation unit 'jar:file:/Users/baz/.gradle/caches/modules-2/files-2.1/com.badlogicgames.gdx/gdx-ai/1.7.0/fedc63e153ddbd33ea94c1ed379f811c6688269d/gdx-ai-1.7.0-sources.jar!/com/badlogic/gdx/ai/btree/utils/BehaviorTreeParser.java' is removed due to invalid reference(s):
16:00:59.004 [QUIET] [system.out]          com.badlogic.gdx.ai.btree.utils.DistributionAdapters
16:00:59.021 [QUIET] [system.out]       Compilation unit 'jar:file:/Users/baz/.gradle/caches/modules-2/files-2.1/com.badlogicgames.gdx/gdx-ai/1.7.0/fedc63e153ddbd33ea94c1ed379f811c6688269d/gdx-ai-1.7.0-sources.jar!/com/badlogic/gdx/ai/btree/utils/BehaviorTreeLibrary.java' is removed due to invalid reference(s):
16:00:59.022 [QUIET] [system.out]          com.badlogic.gdx.ai.btree.utils.BehaviorTreeParser
16:00:59.022 [QUIET] [system.out]       Compilation unit 'jar:file:/Users/baz/.gradle/caches/modules-2/files-2.1/com.badlogicgames.gdx/gdx-ai/1.7.0/fedc63e153ddbd33ea94c1ed379f811c6688269d/gdx-ai-1.7.0-sources.jar!/com/badlogic/gdx/ai/btree/utils/BehaviorTreeLoader.java' is removed due to invalid reference(s):
16:00:59.022 [QUIET] [system.out]          com.badlogic.gdx.ai.btree.utils.BehaviorTreeParser
16:00:59.041 [QUIET] [system.out]       Compilation unit 'jar:file:/Users/baz/.gradle/caches/modules-2/files-2.1/com.badlogicgames.gdx/gdx-ai/1.7.0/fedc63e153ddbd33ea94c1ed379f811c6688269d/gdx-ai-1.7.0-sources.jar!/com/badlogic/gdx/ai/btree/utils/BehaviorTreeLibraryManager.java' is removed due to invalid reference(s):
16:00:59.041 [QUIET] [system.out]          com.badlogic.gdx.ai.btree.utils.BehaviorTreeLibrary
16:00:59.217 [QUIET] [system.out] Wrote 3854 units to persistent cache.
16:01:00.628 [QUIET] [system.out]    Finding entry point classes
16:01:00.629 [QUIET] [system.out]    [ERROR] Aborting compile due to errors in some input files

See libgdx/libgdx#2121

There is a issue with the way gdx ai calculates rotation, it asumes rotation values are continuous, and thats not the case with normalized angles or with the jump from 360º to 0º.(im using degrees instead of radians for ease of explanation)
When the owner is at 350º and wants to face a target that is at 10º, instead of moving 20º degrees it reverses movement, because he doest not know that 10º is the same that 370º and just needs to move 20º, he thinks that just has to reverse the angular velocity to get to the 20º.

i fixed the problem localy using normalized angles in my rotations, and just adding 2Pi to the targetOrientation if his value was lower than owner.getOrientation() int the reachOrientation method.

I have a game I'm working on with procedural generation. To make testing easier I like to use the same random seed on every run.

Instead of using MathUtils.random for doing random stuff in things like the Wander steering behavior, it would be great if it used something like GdxAi.random(), and then allow me to do something like GdxAi.setRandomInstance(myRandom); so that the ai will play out the same way every time with all the other aspects of my game.

edit: also to add, this would be useful for games with save/load functionality. Without being able to set the random instance, players could reload their game over and over to get different outcomes and "cheat" the game.

I was looking through the pathfinding code and found this:

public Array<Connection<N>> getConnections (N fromNode) {
  return nodes.get(fromNode.getIndex()).getConnections();
}

fromNode.getIndex returns the index into the nodes-array. So nodes.get(fromNode.getIndex()) will return the fromNode itself. Basically you could just do the following

public Array<Connection<N>> getConnections (N fromNode) {
  return fromNode.getConnections();
}

Perhaps I overlooked something, but as it seems, the whole nodes-Array of the DefaultIndexedGraph is not really used.

Reference:
http://gamedev.stackexchange.com/questions/1885/target-tracking-when-to-accelerate-and-decelerate-a-rotating-turret

The control equation is vastly superior to the current method of deciding when to stop rotation. With high acceleration, the current method will cause oscillation, whereas the control equation one will halt in few small oscillations.

Would you like a pull request from it?

Issue details

If you have two listeners to a message, and the first one removes itself upon receiving that message, the other listener will not get the message delivered.

I believe it is because of the way the for loop is constructed in the discharge method

    Array<Telegraph> listeners = msgListeners.get(telegram.message);
    if (listeners != null) {
        for (int i = 0; i < listeners.size; i++) {
            if (listeners.get(i).handleMessage(telegram)) {
                 handledCount++;
            }
        }
    }

Reproduction steps/code

Create two listeners to the same message
Have the first remove itself as a listener during the onMessageRecieved call
The second listener will not get the call

Version of gdx-ai and/or relevant dependencies

1.8.0

See discussion and proposed implementation here, #7 (comment)

Ideally, this kind of goal could be specified by an explicit set of nodes or abstract properties.

Hi!

In my program i am interrupting the execution from BehaviorTree while they are running. After that i want to reuse the BehaviorTree instance (or pool it).

At the moment im am interating through the nodes of the tree with this method:

    @SuppressWarnings("rawtypes")
    private void reset(Task<?> task) {

        if (!ClassReflection.isInstance(Sequence.class, task) && !ClassReflection.isInstance(Decorator.class, task)
                && !ClassReflection.isInstance(BranchTask.class, task))
            task.childRunning(null, null);

        if (task instanceof Poolable)
            ((Poolable) task).reset();

        for (int i = 0; i < task.getChildCount(); i++) {
            Task child = task.getChild(i);
            reset(child);
        }
    }

At the moment i am abusing the childRunning() method to set the runningTask to null. But i had to add the checks to not get a nullpointer when calling childRunning().

Is there a better way to reset a BehaviorTree or can you provide a method to do this?

Hi!

I wanted to use the ai.utils.CircularBuffer as part of my project, however, I encountered some issues:

Regarding CircularBuffer's store() method ( https://github.com/libgdx/gdx-ai/blob/master/gdx-ai/src/com/badlogic/gdx/ai/utils/CircularBuffer.java ):

    public boolean store (T item) {
        if (isFull()) {
            if (!resizable) return false;

            // Resize this queue
            resize(Math.max(8, (int)(items.length * 1.75f)));
        }
        items[tail++] = item;
        if (tail == items.length) tail = 0;
        return true;
    }

Why does it abort and return false when the circular buffer is full and not resizable? Ring buffers would usually overwrite the oldest entry, but apparently this never happens here.

Also, if the buffer is full and resizable, it will get resized. Thus f (tail == items.length) tail = 0; will probably never get executed.

Another problem:

        CircularBuffer<Float> buffer = new CircularBuffer<Float>(3, false);

        System.out.println(buffer.store(3.2f));
        System.out.println(buffer.store(3.2f));
        System.out.println(buffer.store(3.2f));
        System.out.println(buffer.store(3.2f));

prints true, true, false, false.

So, not only does the "circular" buffer simply don't store new values like a circular buffer should but it also stores one element less than its capacity. This is likely caused by a bug in the isFull() method.

Right now, sending a message using the MessageDispatcher#dispatchMessage() method requires specifying a sender, which is an instance of Telegraph.

A very common case scenario is implementing the interface on the very same class you are dispatching the message from. So, typically, you end up with something like:

MessageManager.getInstance().dispatchMessage(this, 1);

Equally common is having classes that only generate events and do not handle any type of message. Currently, those classes are forced to implement Telegraph in order to broadcast events (or to have access to some other unnecessary Telegraph instance). So, you find yourself writing this

    @Override
    public boolean handleMessage(final Telegram msg) {
        return false;
    }

very frequently, just to dispatch a message. I tend to think the API could benefit from removing the need of having a mandatory sender parameter. If you do not specify one, just dispatch the message to every possible recipient from an anonymous sender.

What do you think?

When using the framework out of libgdx, using it collections can be a "problem" (the same collection isn't shareable between not-libgdx-collections-dependant code and dependant one).

In RadiusProximity and so, in ProximityBase the agent collection used is a 'com.badlogic.gdx.utils.Array' where it could be 'java.lang.Iterable'.

The only usage of the collection in those is to iterate it like:

for (int i = 0; i < agentCount; i++) {
    Steerable<T> currentAgent = agents.get(i);

I would suggest to use the List interface so you could continue using the ".get" method but Array doesn't implement it. However, I think in this cases is it perfectly possible the implicit iteration way:

for (Steerable<T> currentAgent : agents)

As I can see, the default iteration system on 'com.badlogic.gdx.utils.Array' reuses the same iteration object to be gc friendly.

Compiling module com.supercookie.game.GdxDefinition
   Validating units:
      [ERROR] Errors in 'jar:file:/C:/Users/Cookie/.gradle/caches/modules-2/files-2.1/com.badlogicgames.gdx/gdx-ai/1.3.2-SNAPSHOT/8263c0e657fd93f1c5f88216efd14cd2e9548f5e/gdx-ai-1.3.2-SNAPSHOT-sources.jar!/com/badlogic/gdx/ai/btree/parser/TreeLineProcessor.java'
         [ERROR] Line 59: The method newInstance() is undefined for the type Class
         [ERROR] Line 63: No source code is available for type java.lang.InstantiationException; did you forget to inherit a required module?
   [ERROR] Aborting compile due to errors in some input files
:html:compileGwt FAILED

Am I missing something here since the move? I am working against 1.3.2-SNAPSHOT

Issue details

The guard in a selector is not handed correctly. The guard is checked and if it fails the run method is called anyway. In SingleRunningChildBranch line 88 is an else missing. The run method is called later on the child that succeeded. This child is called twice.
I could only see it in the debugger, when executing step by step. I used to set a breakpoint in LeafTask:42, this gets called twice for the last child.

Reproduction steps/code

root
  selector
    (care urgentProb:0) walk
    bark # bark will be called two times

Version of gdx-ai and/or relevant dependencies

gdx-ai1.8.0 and below

Solution

I solved it myself with the following code:
SingleRunningChildBranch: Line 82

                } else {
                    runningChild = children.get(currentChildIndex);
                }
                runningChild.setControl(this);
                runningChild.start();
                if (runningChild.checkGuard(this))
                    run();
                else
                    runningChild.fail();
            } else {
                // Should never happen; this case must be handled by subclasses in childXXX methods
            }
        }
    }

instead of

                } else {
                    runningChild = children.get(currentChildIndex);
                }
                runningChild.setControl(this);
                runningChild.start();
                if (!runningChild.checkGuard(this))
                    runningChild.fail();
                run();
            } else {
                // Should never happen; this case must be handled by subclasses in childXXX methods
            }
        }
    }

Following this thread:
http://www.badlogicgames.com/forum/viewtopic.php?f=17&t=16550#p70988

Two things would be needed:

1. Make previousState in MessageDispatcher protected, so a method for setInitialState can be made (or better, add it so there's no need to extend it :D)
2. MessageDispatcher needs a method that returns an array of "Savable" objects, with the information of the pending messages, so they can be saved (serializing them or whatever).

Thank you! :)

Hey everyone,

Would it be possible to get a Gitter or some other form of Chat system for GDX-Ai?

Just curious,
HeadClot

For the moment the pathfinder finds connection between nodes by calling getConnections() on the IndexedNode.

My problem is that there is no context to decide if a connection should exist or not or to provide a different cost for the connection.

Simple example, you could imagine that some units could go over water cells while other cannot. Or that some agent have a different cost for a specific connection. Problem is, there is no way to tell for which entity (i.e. context) the pathfinding is running.

I've opened this to discuss behavior trees API enhancements.

@implicit-invocation
Resuming discussion #4 (comment) ...
Not sure why you don't like the XML format, but I think that it has some advantages:

• comments are supported which can be really useful
• it's more powerful than json or any inline tab-based format.
• being it a standard format, external tools like editors (written in any language) can easily read/write a behavior tree.

So I'm playing with the XML format just to see what can be done.
Currently I can successfully load a behavior tree from this file:

<BehaviorTree>
  <Import task="com.badlogic.gdx.ai.tests.btree.dogtasks.BarkTask" as="Bark"/>
  <Import task="com.badlogic.gdx.ai.tests.btree.dogtasks.CareTask" as="Care"/>
  <Import task="com.badlogic.gdx.ai.tests.btree.dogtasks.MarkTask" as="Mark"/>
  <Import task="com.badlogic.gdx.ai.tests.btree.dogtasks.RestTask" as="Rest"/>
  <Import task="com.badlogic.gdx.ai.tests.btree.dogtasks.WalkTask" as="Walk"/>
  <Root>
    <Selector>
      <Parallel>
        <com.badlogic.gdx.ai.tests.btree.dogtasks.CareTask urgentProb="0.8"/>
        <AlwaysFail>
          <com.badlogic.gdx.ai.tests.btree.dogtasks.RestTask/>
        </AlwaysFail>
      </Parallel>
      <Sequence>
        <Bark times="3"/>
        <Walk/>
        <Bark/> <!-- times defaults to 1, see BarkTask source code -->
        <com.badlogic.gdx.ai.tests.btree.dogtasks.MarkTask/>
      </Sequence>
    </Selector>
  </Root>
</BehaviorTree>

I added the "Import" tag to improve readability. It allows you to use the given alias in place of the fully qualified class name of the task. Actually Selector, Parallel, etc.. are predefined imports. Also, the "as" attribute is optional, meaning that the simple class name is used as the alias, i.e.

  <Import task="com.badlogic.gdx.ai.tests.btree.dogtasks.BarkTask"/>

creates the task alias "BarkTask".
Also, I added task parameters, see urgentProb in CareTask and times in BarkTask.
The attribute value is parsed according to the type of the corresponding field of the task class. For example, urgentProb is a float and times is an int. Supported types are: int, Integer, float, Float, boolean, Boolean, long, Long, double, Double, short, Short, char, Character, byte, Byte, and String.

Of course, we can maintain both formalisms as long as they have the same structural features. I mean, unlike task parameters, imports are just a syntactic sugar so they are not mandatory for the inline tab-based formalism.

I think we can use a "btree" branch in order to experiment with BT improvements while keeping the master branch clean.

I have seen your implementation of the hierarchical pathfinding and mine is slightly different. I just have a tilemap and each tile has a node with connections. Now I want to divide my map in "sectors" and build nodes on the edges where there is a entrance.

• Is there some implementation already available that does this for me? Since this would be pretty much the same for each basic tile map I would expect something like this is already in place. Something like this:

  (forgot a view connections).

If so, I would like to know :)

If not:

• So i lookup all the borders and add level 2 nodes where there is an entrance. I (try) to connect all those within a "sector" and the bordering neighbors.
• Is this sufficient, does the pathfinder find paths in between the level 2 nodes once it finds a level 2 path? Or do I need to store the actual shortest path in the connections? I can imagine it is the former since looking up A -> B -> C -> ... -> Z is much faster then looking up A -> Z where the open list will grow out of control.

I see you store all the nodes in a single array and use an offset for higher levels. However I want to make a dynamic map. And once a node changes it will update it surrounding nodes and if level 2 changes as well it does the same. But that means I need something more flexible then a ArrayList or (like I do on the 1st level) create all possible nodes and alter there connections.
I would appreciate some pointers in the right direction.

Thanks in advance,

Hi, because of the way radius is calculated in CirclePattern:

float radius = memberRadius / (float)Math.sin(Math.PI / numberOfSlots);

it get divided by 0 if there is only 1 slot. We should be able to generalize formation even for 1 slot only.

The face behavior is now unusable with linear velocity behaviors and since linear velocity behaviors zero the angular component and face behavior zeroes the linear component. Furthermore, it is impossible to extend ReachOrientation behavior, because it uses the linear component for temp calculations.

In my use case, when player moves a ship, both face and arrive behavior are used simultaneously. I feed these values to Box2d that then does the physics for me.

Currently I can work around this by using calling calculate steering twice: once for movement, and once for rotation. This solution is not bad, I am just not sure if it is intended, since it seems obvious to use Face behavior with seek and evade behaviors.

When using the same behavior tree in many objects is a waste of resources to have the tree instantiated for all of them. Instead, it could be great to have them updated like:

public void start(T object);
public Status execute(T object);

where the object is the blackboard object being updated like:

tree.step(object);

This way, the same tree could be used for as many objects as they are needed. In lot of games it is normal to use the same ai tree for lot of entities instead of having a different one for every of them.

This is an easy enhancement that I think is very useful.

Here is the gist: https://gist.github.com/scooterman/b03dfe1560cfc3b977a3

This is how I've implemented the loop (as a branch, heh) that suffices my needs. I'm asking for guidance (if there is a desire to use it on the main lib) on how to integrate it. The first thing should be able pass the leaf as a parameter instead of the first child, something like:

exhaust leaf: "entitiesInRange"

I'm working on 2D top down game and I'm trying to make heavy enemy which slowly builds up his speed and crash into the wall. His body can be affected by external forces (push back on hit, attract by magnet etc.) so he has pretty big linear damping.

I'm trying to achieve described behavior by using Arrive Steering Behavior but I can't make him accelerate slowly.

Is there any way to make Arrive SB work in such situation?

I've made Box2dArriveTest with my parameters: Box2dCustomArriveTest.zip

character = createSteeringEntity(world, container.greenFish, false);
character.setMaxLinearSpeed(8);
character.setMaxLinearAcceleration(40);

MassData massData = new MassData();
massData.center.set(character.getBody().getMassData().center);
massData.I = character.getBody().getMassData().I;
massData.mass = 1f;

...

final Arrive<Vector2> arriveSB = new Arrive<Vector2>(character, target) //
    .setTimeToTarget(0.1f) //
    .setArrivalTolerance(0.001f) //
    .setDecelerationRadius(0.5f);

if (steeringBehavior != null) {
        steeringBehavior.calculateSteering(steeringOutput);
        if (this.body != null && !steeringOutput.linear.isZero()) {
            System.out.println(steeringOutput.linear);
            this.body.applyForceToCenter(steeringOutput.linear, true);
        }
        // applySteering(steeringOutput, deltaTime);
    }

I'm using gdx-ai 1.5.0

calculateSteering is returning a NaN,NaN vector.
I am using the example Box2dRadiusProximity with the CollisionAvoidance steering behavior. No errors are thrown, my box2d body just disappears once i approach the radius distance, very weird behavior. I don't understand where the issue is spawning from

The docs in Parallel states:

parallel task will succeed if all the children succeed, fail if one of the children fail.

but that doesn't seem to be the case. I have this:

Parallel<Brain> parallel = new Parallel<>();
parallel.addChild(new TaskA());
parallel.addChild(new TaskB());     

I would expect parallel to stop if TaskA fails, but that doesn't happen. Is it a bug or am i misunderstanding something?
Full example

Related:
libgdx/libgdx#2793
http://www.reddit.com/r/gamedev/comments/2u8ytc/gdxai_150_released/co6b3hx

To ease using this library in other projects without bringing the whole graphic library, it'd be nice if it was split into a shared module and the core library,

It would be great if we could work with Vector2 and Vector3 being interfaces so we could share code between different platforms without having to refactor things or having to duplicate vectors.

It can be a good practice to use local attributes, like velocity or positions but in some cases, for memory optimization purposes, or even different usage/programming styles (ie: having the position shared between the steering agent and the visual object so it automatically changes the value) it is comfortable to have the same object.

Using, for example the JME3 engine we have a Vector3f (which is the equivalent to the Vector3). Well, if I want to use the Vector3f I have to make the conversions from them or make a thin wrapper/decorator:
"MyVector3 implements Vector".

This is just fine but the same problem can be found: the code must be done or with Vector3 or with MyVector3 so if I want to put/share that code between engines it is not so straight-forward. However, if Vector3 were an interface (MyVector3 implements Vector3), I could do my code and share it between engines without being aware about refactoring it in so much places and without wasting "so much" resources.

EDIT by davebaol

• Pathfinding TODO List
• Pathfinding wiki page (to be finished)

Hello, now that you have your own extension, maybe you can start considering an A* implementation to go together with that wonderful steering, right? :)

On initial observation, seems like the default constructor puts the BehaviorTree into a state where the rootTask field cannot be set.

Hi,

I was looking the examples on steering behavior, and I found ans issue. I'm not 100% sure I got it right, but let me explain. On the Scene2D example, the steeringOutput is calculated, and scaled with time and added to the velocity.

private void applySteering (SteeringAcceleration<Vector2> steering, float time) {
        // Update position and linear velocity. Velocity is trimmed to maximum speed
        position.mulAdd(linearVelocity, time);
        linearVelocity.mulAdd(steering.linear, time).limit(getMaxLinearSpeed());
}

So the steering linear is treated as a force. On the other hand, on the example with box2D, the steeringOutup is calculated, then scaled to deltaTime, and finaly applyed as a force to the body:

protected void applySteering (SteeringAcceleration<Vector2> steering, float deltaTime) {
        boolean anyAccelerations = false;

        // Update position and linear velocity.
        if (!steeringOutput.linear.isZero()) {
            Vector2 force = steeringOutput.linear.scl(deltaTime);
            body.applyForceToCenter(force, true);
            anyAccelerations = true;
}

In this case the steeringOutput.linear is scaled to delta before is used as a force. Then the box2D will scale again with dalta before applying to velocity. The result is a much weaker force.

As I sad, I am not sure if I've wrong understood it, or it's a mistake on the examples. And if so, which of the to cases should we use in our code?

Thanks,
StefanW

I discovered that parsing of behavior tree data fails when BehaviorTreeParser debug is set to anything but BehaviorTreeParser.DEBUG_NONE.

I traced the problem down to line 327 in BehaviorTreeReader.java:

if (debug) System.out.println("endLine: indent: " + indent + " taskName: " + statementName + " data[" + p + "]" + data[p]);

Which throws an ArrayIndexOutOfBoundsException.

Not sure if it makes any difference, but I fed a string of tree data to BehaviorTreeParser instead of a file. The string loads fine when debug is off.

On snapshot, PathfinderRequest#statusChanged is set to true by PathfinderRequest#changeStatus but never reset.

This causes each call to IndexedAStarPathFinder#search(request, timeToRun) to restart the search, so a route that takes longer to calculate than timeToRun will never finish

Working around it for now by setting stateChanged by hand after the first search, but assume this is something the finder is supposed to do.

request = new PathFinderRequest<>(..);
request.changeStatus(PathFinderRequest.SEARCH_INITIALIZED);

finder.search(request, TimeUtils.millisToNanos(25))

I've been looking around but It doesn't seem to be possible to use variables when defining trees by text foramt. What I'm really trying is to do is, for example (taken from the "dog example"):

subtree name:"actOnYourOwn"
  selector
    (random success:0.1) piss
    parallel policy:"selector"
      wait seconds:"uniform,3,6"
      randomSelector
        wander gait:$(gaitType)
        wander gait:"walk"
        lieDown
        sit

Where I could import the "actOnYourOwn" tree in a manner that $(gaitType) would be replaced but a given value, allowing to create general tasks to be used by many different entity personalities.

Imagine our characters is moving on the surface of a sphere. So there will be 2 ways to calculate distance between them. But all of gdx-ai behaviors use owner.getPosition() to calculate distance.

I think Steerable interface should have getDistance(point:Vector2):float getDistanceSq(point:Vector2):float method and behaviors will use them to decide how to seek, evade, flee...

Hi,

i have found the following problem with the Arrive Steering Behavior:

When the steerable is approaching the target, the steeringOutput begins to "jump around". You can see this, when you look at the steeringOutput angle:

steeringOutput= [0.09993597:0.0035781164], len= 0.10000001, angle:2.0505471
steeringOutput= [0.099935964:0.003578156], len= 0.1, angle:2.05057
steeringOutput= [0.09993597:0.0035780247], len= 0.1, angle:2.0504947
steeringOutput= [0.09993595:0.0035785697], len= 0.1, angle:2.050807
steeringOutput= [0.09993608:0.0035749506], len= 0.10000001, angle:2.0487323
steeringOutput= [0.063587725:0.0023101782], len= 0.06362968, angle:2.0806732
steeringOutput= [0.09993937:0.0034819124], len= 0.10000001, angle:1.9953921
steeringOutput= [-0.07271707:-0.0016047317], len= 0.07273477, angle:181.26422
steeringOutput= [0.099956475:0.0029501538], len= 0.1, angle:1.6905588
steeringOutput= [-0.09558454:0.002485239], len= 0.09561685, angle:178.51062
steeringOutput= [0.09999966:2.6154998E-4], len= 0.1, angle:0.14985727
steeringOutput= [-0.096632786:0.02573139], len= 0.1, angle:165.08925

After some debugging, the problems seems to be in this line of code in Arrive.class:

// Acceleration tries to get to the target velocity without exceeding max acceleration
        // Notice that steering.linear and targetVelocity are the same vector
        targetVelocity.sub(owner.getLinearVelocity()).scl(1f / timeToTarget).limit(actualLimiter.getMaxLinearAcceleration());

When timeToTarget has the default value of .1f and the velocity is very small, for example [0.011008702:0.1401149], the resulting vector is to long [0.11008702:1.401149] and it gets limited to the MaxLinearAcceleration.

When i commenting out this code

.scl(1f / timeToTarget)

or setting timeToTarget to 1, everything works fine.

Is this behavior the way it should work? For what do i need the timeToTarget?

According to the documentation, the MessageDispatcher can delay the message. The delay is always given in seconds.

However, some games could use another piece of time. For example, on turn-based game, time can be expressed in turns, not in seconds. So we could send a message in 3 turns for example.

Is there any reason to force the delay in seconds ? Could gdx-ai provide a more "generic way" to delay a message ?

This is clearly annoying to have to override method if it wont be used.

The limit decorator, stated on ai's wiki is not present anymore. It has changed to something else?

Some small enhancements to the SteeringBehavioursTest suite that could be useful :-)

Would it be possible to add a test for fleeing to the SteeringBehavioursTest?

Also would it be possible to add a toggle to the behaviour of the AI actors to be able to stay within the display area (i.e. not wrap around on hitting the world bounds) and behave appropriately?)

Keep up the good work :-D