Archive for June, 2013

Latest developments in the PaaS universe

June 28, 2013

Hi folks,

I’ve been considering the feasibility of maybe setting up a small PaaS on RightScale using Amazon EC2.  There are a few bits and pieces that might facilitate this.  However, most recently, I came across this interesting post, which pointed me to Dokku, a project that is only a few days old, but looks quite promising and very interesting.

An alternative strategy for PaaS cloud hosting might be to write a RightScript or three and get Sandbox installed subsequent to Docker on a virtual box on EC2, via this tutorial (a key difference being that RightScripts would likely be taking the place of Vagrant in setting up the system, if using RightScale).  However Docker itself only supports Python (as yet – of course, the project is still young!), and does not yet (it would seem) support databases, as required with Django apps.  Dokku, however, I believe supports Django and a number of other techs that are not even Python based – and it has some quite clever chaps from MIT, I believe, driving it.

Dokku builds not only on docker, but makes use of a couple of other useful github projects in order to work its magic.

Essentially it should allow the admin to deploy multiple data-driven websites / cloud services (eg, data monitoring / data feed input, then email / sms notify relevant subscribed users, say by google cloud messaging) / applications on the same server, but, according to the project description, it does not yet support multitenancy – which I assume would mean multiple admins / admin user logon (controlled by master admin), not does it yet support multihosts (ie, scaling to multiple virtual boxes, or failover between same).

More here.

The jaws that bite, the claws that catch

June 23, 2013

I recently went to a local park, and spotted my first lyrebird.

There were actually two of the critters.  For some reason their movement reminded me a bit of Velociraptors in Jurassic Park.

Guards vs Skeletons

June 4, 2013

Hi folks,

So I’ve done a little more work on my project.  This time I’ve sandboxed the progress in a new scene, with a road and some buildings.  The main aspects of experimentation have been Unity 4.1’s new MecAnim animation system, patrol pathing, and basic AI for recognition, attack, retarget, knockout and respawn.

TL;DR – here is the video:

As you can see the combat behaviour is not totally optimised but the general idea is there.

In terms of working materials, I have used a couple of assets from the asset store (one purchased – the guard and associated mecanim animations, the other free – the skeleton model and animations), and also some AI scripts following a bread trail starting  here.  Working with the AI scripts was quite interesting.  Amongst the key learnings I gathered (or rather, the key bugs), I discovered that quite a few were simply due to the fact that I had transforms in the wrong location.  Making extensive use of Debug.Log statements to check that the flow of the script was working as I expected was very useful and helped guide me to avoid various pitfalls.  I also found that simply defining obvious booleans (eg, Engaging, Attacking, Friendly) was helpful for debugging.

Basically what I wanted to implement was the following – a situation where there are three types of creature: players, guards, and mobs (enemies).  Enemies view players and guards as viable targets, players and guards view enemies as viable targets.  If there are none of the other present, guards (or monsters) will path on a preset sequence of waypoints.  If a hostile rel them passes into their line of sight (I found that checking for ray collision was largely empty unless I raised the ray emanating from the creature a nominal height; this was due to the nature of capsule colliders I am using for creatures), or is sufficiently close (their “noise” > given threshold value) then they move to engage that target.

Once they are adjacent to the target they move to attack it (I found that I needed to reset the transform periodically for creatures when they were attacking, otherwise it bugged out, ie they did not stay fixed in place.  Even when I did this I became confused since it was still bugging out, but then I realised I need to fix the rotation as well.  So prior to attack I calculated the initial position and rotation information and then fed these as temp vars into the attack function).  If the target moves out of range they follow and re-engage.  If the target is reduced to < 0 hp it is destroyed; the associated spawn point for that target is then empty.  The spawn point then detects this and a cooldown timer is initiated.  Once this cooldown timer reaches 0 a new creature is popped at the spawn point and starts to patrol, etc.  As part of the spawn process the list of targets for the creature which struck the knockout blow is refreshed, and a new goal (hostile creature) assigned if there are any in range.  It then proceeds to attack the new hostile creature.

I decided that skeletons should have 50hp and guards 100hp.  They both do a similar amount of damage.  In the video you can see the cycle of destroy-respawn-destroy-respawn, an unending tide of guards and skeletons leaping to rejoin the fray to replace their fallen comrades.

One more thing is worth mentioning.  I also kept track of variables to make sure that the animation playing is appropriate – this is the MecAnim part of the picture.  So the guards and skeletons have state machines for their animations, as well as for their AI behaviour.


I am reasonably satisfied with the outcome, but I know that in theory the implementation could be a lot cleaner.  My code is not totally unreadable but still is, for some scripts, way too long and should be written in a more sensible, concise format.  I’m also certain that in a few places I’ve written out the same functionality more than once; a bit more focus on reusability certainly would not go amiss.  And also redundancy – I’m fairly certain that the code is not as tight as it probably should be, in fact, I know that it isn’t.  This applies more generally to the project at large.  There are many places in my work where I’ve coded the same functionality for different things in different contexts.  Or even functions that have not been used.

So a little work needs to be done at some point on said matters.

In particular, todos that spring foremost to mind with this particular line of work (many of which, such as style, readability, best practice etc. I likely will not implement since this is primarily a “finding my coding feet” ongoing exercise):

  • make it clearer if a creature is in combat (ie, none of this running around in circles spontaneously business) – it could be that artifacts such as running around in circles occur because noise isn’t refreshed immediately for the goal creature when the original target is destroyed.  This could be fairly easily redressed.  Or maybe I might like to simply have a fixed (small) radius wherein if the goal is within that circle the creature automatically moves to attack (this is currently not the case, but it should be, and that is what noise/”sound” detection (generalisation of said idea) was supposed to do)
  • possibly aim to fix creatures in place during combat – try to make things more static
  • morale stat for creatures – chance to run away if allies nearby down / no support (possibly hard) or hp low; then move randomly away from attacker, or seek “wall” tagged object to attempt to throw off attacker?
  • knockout animation & possible option (if player dealt k.o.) for ability to loot the creature; then despawn timer prior to respawn
  • refactor the code to make some of the classes less lengthy and so that they have a more logical structure
  • subclass creature.cs (my main AI script) as part of a SFS_creature.cs script so that it interacts well with smartfoxserver
  • NPC_targeting.cs (my main AI targeting / listing script) – make this dependent more on “Broadcast messages”, possibly local messages.  There is a messaging system I am currently using for health bars – this however is more general and can be extended.  If the world becomes quite large and hence contains 100s of monsters, guards, players etc I do not want every single monster / guard / player to have a list of length 400 encoding and tracking the position of everything else in the world – that would be untenable.  But that is precisely how the program works currently.  Evidently this will have to be fixed, so that only enemies / guards in noise / detection / aggro radius are detected.  Or “targeting radius” which is larger than aggro radius.
  • Incorporate the idea of “threat” levels depending on which creature is damaging which – ie first calculated purely as a function of total damage dealt, then maybe factoring in threat generation abilities (eg taunt – sets to top threat) or threat reduction abilities (feint), and threat reset effects (moving beyond targeting radius – as per point above).  Evidently want the creature with highest threat to be the creature targeted.

The next part of the project will focus on hooking up a mysql database with the smartfoxserver software I am using as described here.  Basically I’d like to aim at storing 8 fields: username, password, spawn location (x,y,z) and last seen location (x,y,z).  Essentially I would like last location to be updated every time the player logs out, and then persisted in the database, so that when they log back in (if it is not a first time logon), they appear at that location.

If it is a first time logon they appear at the default spawn location, and last location is null until populated.

The architecture of the goal application will essentially be

client unity application <-> server SFS2X application <-> mysql database

with <-> denoting two way communication, the client application written in C#, the server application written in Java, as is currently the case; the new component will be the database.  Naturally a database can consist of more than one table, and more than 8 fields.  In a full blown, sophisticated application, there is no reason that all data about the players should be able to be stored in an appropriate database schema.  So certainly there is considerable room for extending this aspect of my learning, once I’ve managed to get this proof-of-principle up and running.

Consequent to this I might look at readapting this patrol example for SFS2X-deployed unity games, or something completely different.