View Full Version : Multi-agent system for Lightwave Development (Pagent 2.0)

01-20-2013, 03:27 AM
Sorry for my bad English. Please talk in simple English to me. The Software have the Name Pagent. The behaviour of the Agents can manipulate by other Agents or by the Surface (i call this Intelligent Surface). In the Future Pagent have a Cool motionsystem. With this System you can make rules to entry or exit the Motion. The workflow is that you make a Motiontree. Pagent works with a instancesystem. It is possible to render many touthand in less as 20 Seconds per Frame. Don,t call me when it is final. It is when it is done. Here are first very very simple Videos. Please visit my Blog to see more details: http://lpdful.wordpress.com/



01-20-2013, 04:27 AM
yes finally a crowd sim that's in lw! Been waiting for years.
Looks like there's some very visible intersection among characters in the third video.

01-20-2013, 08:19 AM
the collision-system is still not ready. There are so manny Things to do. At first i programmed the Motionsystem. Because the Actor should make different motions and behaviour (walk, rund, stand, look).

01-20-2013, 08:30 AM
looks good, so you have automated behaviors, What's the procedure to get it working IN LW ?

01-20-2013, 09:23 AM
Wow! Very cool. Please keep us updated on your progress.

01-20-2013, 10:37 AM
Most interesting! Here hoping you will keep on developing this (and keep us updated)

What are the agent states based on...is it fuzzy-logic such as Massive?

01-20-2013, 11:54 AM
Nice, will be following this :)

Kev xx

01-20-2013, 12:33 PM
neato! Of course I'm always interested in nice collision calculations. Thanks for sharing. =)

01-21-2013, 08:58 AM
Hello Pille,

When can we see your Lightwave Screen Shots on PAgent2?
Have look at this by: VueScriptsdotcom On YouTube.com

:Tutorial for instances on a terrain .


I would like to know your thoughts on this & how more Complex is/or will be your PAgent Rev X?

Good Luck!

01-21-2013, 10:18 AM
Wow i am very happy about your interest about Pagent. Some Question i will answer:

- Screenshots from Pagent? Hmm. At this Time Pagent is a Script Language but i will codeing a GUI for it. In this gui you can define Situations and the action of the Agent. But some Settings you make it in Lightwave. Give me three Weeks. Then i show it in broken English ;-)
- Have Pagent 2.0 a Collision modell? At this Time No. In the Future YES. I think more then that. In the Future the Agents have a cone-Vsion. You can make rules. A Sample: if i see Agents with color red ==> run away. or, if i see the Face of the Agents with my color then smile to him (smile ist a Motion)
- Do Pagent have Fuzzy logic. At this Time Yes. Fuzzy Logic is very important for natural Behaviour. Without Fuzzy the Agents have a behavior like soldiers. You can define the Skill of Fuzzy for EVERY rool.


Give me a Feedback about my English. Can all People unterstand my Words that i write here.?????????????????????????????????

01-21-2013, 11:15 AM
I understand you just fine. Please keep us in the loop on development. Pagent2 would be a great resource for LW.

Just curious but what was Pagent1?

01-21-2013, 11:37 AM
Pagent 1 was a painful experience. The HD was crashed by streaming. At this Time a Backup was for me a mysterius Thing ;-). But i learn from my Errors.

01-21-2013, 05:07 PM
I think we understand you rather clearly Pille. Of course, it also helps we REALLY want to understand you because your project is so very cool.

Darth Mole
01-21-2013, 05:32 PM
Yep, understand you fine. Very interested in seeing how this progresses. It feels like a natural evolution of LW's flocking and instancing tools. Can you use motion capture files to drive the motions? (For people like me that can't be bothered to rig and animate?)

01-21-2013, 05:51 PM
Looks great. Thanks for doing this!

01-21-2013, 10:07 PM
I understand your english perfectly thanks! I'll keep an eye on this project of yours. =)

01-21-2013, 11:13 PM
Pagent is very promising tool!

01-22-2013, 03:56 AM
nice.. Looking very promising!

01-22-2013, 05:34 AM
To Darth mole,
If you use Vue's great tutorial, see my post URL
Then you will see year ago he used off an offset MMD motion
to create his teams of hill walkers.
So I remember a couple of
I think there are a couple of Japanese houses that sell Mo-Cap in MDD's
off in shelf
if you google it I'm sure they will help you.

I look forward to see great LIghtWave libraries of all kinds of agents

From this- the English is okay Pille ,
onto the Beta keep the posts
Coming for the GUI,
lots of people can advise you on that
in this forum roll on!

01-23-2013, 01:37 PM
Hello Friends of the Lightwave,

The development progresses. In this video you can see the first motion change. The Agent have 4 Motion: 1. Stand 2. Stand to walk 3. walk 4. Walk to stand. The quantity of motions are now unlimited. in the video you see only one kiind of Agent. But You can drive different Motions for different Agents. I am only to lazy to make the Animation. Maybe the Weekend. The next Steps of Development are: 1. Timescale for Motions 2. Senses (ears and eyes) 3. evasion strategie from Agent to Agent. Cool is that all Animation in this video is generate by the userinterface (no change of Code is nessersary.) http://www.youtube.com/watch?v=5Q0opRd9Kwk&feature=youtu.be

01-23-2013, 02:09 PM
Very cool Pille!

Can you go into how your system integrates into Layout? Is it a plugin? Are these rigged characters? What drives their animation? Is it keyframe or baked (mdd) animation?

01-23-2013, 03:34 PM
Very cool Pille!

Can you go into how your system integrates into Layout? Is it a plugin? Are these rigged characters? What drives their animation? Is it keyframe or baked (mdd) animation?

This is definitely looking very cool & yeah I'd like to know a bit more about the whole system ... if possible that is.

01-23-2013, 04:24 PM
Great work, Pille! Don't stop working on this project, please.

I've developed a simple crowd system for LW outside the program using Perl scripting for specific projects. I did it this way for the hurry and because never had a clue how to properly do it inside LW.

01-24-2013, 01:41 PM

this post is in 3 Sections divided. The Pagent architecture, a further Progress, Plans for the futre.

Section 1 The Pagent architecture

Pagent ist a Standalone Application. It is codet in Netlogo ( http://ccl.northwestern.edu/netlogo/ ) Netlogo a very cool Language specifically for Agentsystems. I have in Netlogo write a interpreter who translate all postions, Rotations, Scale for Lightwave. And Pagent have a Dynamic Motionhandling. Pagent use the Instancesystem from Denis Pontonnier (I thing his developent ist very very cool). Pagent use the Pointer from the DPSCAN Files. All Stuff what Pagent generate is Lightwave Scene File format. The Workflow: I Animate in Lightwave some Objects (walkcycle). I make a Scan with the Plugin DPScan. Then i generate a Textfile for every Agents Motion. For Example: Stand 0- 10, Standtowalk 10 - 15, Walk 15 - 45. Then i make a new scene in Lightwave with some Objects. The color of the Objects is Important. When i am ready i render the scene. Now i write the importance of the color in a Textfile. For Example: red color = Startpostion from Agent. Blue color = heading angle 45. Yellow color = random angle. Then i start Pagent. The Agents do his work. then i press the Button "Make Lightwave Scene". Thats it.

Section 2 a further Progress http://www.youtube.com/watch?v=MJ33C-8D7RU
Today i have codet a timescale for the Motions (Pagent Motions that Pagent translate in Keys and dpm-Pointers). This allowing a dynamic speed (for Example differnt walk speed). The actually Features from Pagent at this time is overall:

- Dynamic Motionsystem (Motionentry and Exit, for Example stand, walk)
- Setup and Control the Agents from a color pixel or Image)
- Dynamic Timescale of the Motions

Section 3 Plans for the Future

- a Brain who can think X steps in the Future. It interpolate taking into account his job. For example: Not collide with the other Agent, but follow Agent X.
- the GUI is a Note Based GUI. All Parameters are connectable.
- Import of a Lightwave Rig for Automatic recognize the Loop positions. (When the Motionlibary is bigger I think thas ist very important).
- When no Loop Position is found Pagent shall make a transition with consiteration of Physik. For Example: Not move Feed when Feed is on ground.
- automatic searc in the Motionlibary. The result is a Sugges of Transition from Pagent.

Thanks for all Post. It is a very very good Motivation for me. PS: Check out my Youtube Channel I have code some cool Stuff with Netlogo. For Example Genetic Algorythem drive by neural Network in Self-organising Systems "Digital Evolotion". Good Night to all Lightwavers

01-24-2013, 03:18 PM
Very very interesting. I really hope you continue with development. It sounds like you have a strong grasp of what you're doing so please, be careful when crossing streets. :D

It'd be a shame to see this become vaporware.

01-24-2013, 04:40 PM
You have all our support Pille! and be sure you´ll have lots of testers whenever you need us...we´ll surely follow your progress on this thread.

01-24-2013, 06:05 PM
Yeah.. this is definitely worth keeping an eye on

01-24-2013, 09:36 PM
I'm definitely keeping a close eye out on this too. Anything that makes animating crowds easier is of interest to me.

02-05-2013, 10:03 AM
a little step forward in the development of Pagent 2.0 now the Agents can climb over montains. I used a node for the that, because sticy surface was very slow for that. http://www.youtube.com/watch?v=w2rptk7aU54

02-05-2013, 10:39 AM
Hi pille,
There'you go nice update,
what is your feelings with the improvements in 11.5 Flocking & Instancing?
Will they help you out in PAgent?

02-05-2013, 10:57 AM
can you replace the guys to orcs? and you have the intro of Lord of the rings :D

02-24-2013, 03:18 AM
Beta testing ??

03-02-2013, 08:28 AM
Here are some news from my Multiagent Development:

at first the clip: http://www.youtube.com/watch?v=Dz8bFNphP4s a tip: View this in HD to see more details

- I have fix a Problem when i render Motionblur (at the end of the Loop from every single Agent there was artifacts in the Rendering
- at the first Time you see the new DEFAULT Agent from Pagent. (a woman and Chrildren is in development at this time)

03-02-2013, 12:32 PM
Interesting, but I don´t see any difference or showcase to what is additional or feature wise different from ordinary flocking, the videos doens´t showcase it good anyway, or I have missed something?

make agents clearly different definable with color or something, and so the showcase of it will show behavior better.


03-02-2013, 02:55 PM
when you see the change of the Motions Stand to walk is this the biggest difference at this time. In the next days I code a nodal system to define a Motiontree. When i have a Motion tree, then it is possible to interact with his Environment to drive the different Motions. The environment can be the distance between the Agents. A color of the Surface. you can define Logicterms to (if and then). Pagent is a Standalone Programm. That generate a Lightwave scene File. I hope the endresult can make thinks like sim city (Cars, People, planes). All driven by nodal terms with interaction with his environment. I think in 3 Month i can show the Nodal Workflow. At this time Pagent can understand a script Language. Can you all confirm that you unterstand my English. I am to lazy to use a translater Software

03-02-2013, 03:59 PM
it is perfectly understandable...and interesting!

03-04-2013, 05:39 AM
Its coming through nicely but how different is this from flocking?

03-04-2013, 10:20 AM
Very nice. I hope the execution of the motion tree is not so complex like massive, while powerful, is targetted for specialists not normal artists.
There's more simple crowd sim flowchart in 3dmax. Easy to understand motion tree

The blending of mocap or motion clip would be possible?

03-04-2013, 02:04 PM
- to show the difference to the Flocking System I make another Animation in the next 10 Days. (hopefully you can see the different to the very good Flocking System from Lightwave). The greates differnt is the situational Motion Change. Maybe helps the metaphor of a intelligent Motionmixer. The intelligent is the possibility to Change the Motion not per Tracks but per Situation (for Example Suface Color (the Surfache is unseen by LW Cam, or distances to other Objectsm or Colors on the Surface.)


at the Moment to make a Motiontree is very complex. Because Pagent unterstand at the moment only his own Scriptlanguage. But the Moment i code a Nodesystem. With this System you have different Nodes to describe rules for differnt Situations.


I have see some Tutuorial from Massive and I think it gives many similarities to the intention of this kind of Software.

03-04-2013, 02:10 PM
Do you have a rough timetable when this will be available to people? And is this a commercial product?

03-04-2013, 02:46 PM
I think the first release is not Comercial it is free for all. hmnmm. No timeline no pressure. My Dream is to make a big online Agent Database.

03-05-2013, 06:58 AM
Nice, maybe you answered already, will it be Nodal?
Does it handle instances?
Projet très intéressant et ambitieux!

03-05-2013, 07:02 AM
I have several projects that could benefit from a multi agent system... really interested!

03-05-2013, 09:13 AM
I wish you luck with this Pille..and take your time, the lightwave team itself will probably work on and boost the native system as well, so that will be a compitition, new Ideas from different developers
can help each other and merge in unison if that´s what they decide.

Have you seen miarmy? a plugin for maya, that looks awesome...

Newtek thread here..


03-29-2013, 12:39 PM
In the Video you see the first development Step of the Pagent Nodal Engine.
More Information in the next days.


03-29-2013, 12:49 PM
Very nice!

04-14-2013, 10:54 AM
Any new update ?

04-14-2013, 11:50 AM
In the Video you see the first development Step of the Pagent Nodal Engine.
More Information in the next days.


Crowd control aside, that's an really nice way to manage nodes!! Really nice.

04-16-2013, 08:30 AM

here are the next Steps in the Pagent Development. The Language in the Video is German. I hope the Pictures in the Movie speaks to you ;-) (in the next time i make a Video in English, (with my expireance it is very funny for the Comunity)). Here is the Video:


OK. Now let´s talk about the Video. The Video shows the first Steps in the Workflow Lightwave to Pagent. Pagent Import automaticly the Motions from the Motionmixer to Pagent. Then you can define a Motioncontainer. In the Motioncontainer you can Place per Drag an Drop your Motions. The Motioncontainer discribe the Exitmotions. When you repeat the same Motion in one Container you increased the probability to Drive this as the next Motion. Pagent Looks in the Container and pick Randomly the Motion.

Now some Words to the Future of Pagent: You habe Nodes like Logic Terms "AND" "OR" "NOT" "NOR" and you have Sensors "EYES" "NOSE" "EARS" and you have Variable like "minmum Distance to next Agent" "Maximum Distance to next Agent" and so on. With this Nodes you can build virtual Prozessors for every Agent. To some Nodes You can Say that they evolving (Mutation of the Prozessor). This is cool for different Behaviour in Same Situations. Maybe I will have Some nodes like neural Interfaces. (You Can Stutdy my Youtube Channel look at the Video "Digitale Evoloution". It Show neural Behaviour in dezentralic Systems.

SSSOOOORRRRRYYYYY for the many Erros in my writeing. I am in hurry. I must Code many things.

Thomas Leitner
04-17-2013, 02:35 AM
Hi Pille,
Dein Video ist sehr interessant.
Zum GUI kann ich noch nicht viel sagen, da ja die wirklich komplizierten Sachen wie Sensoren und dergleichen noch nicht drin sind. Grundsätzlich stellt sich für mich die Frage (ohne genauen Einblick zu haben) warum Pagent nicht als Lightwave Plugin sondern als eigenständiges Programm umgesetzt ist (wie es ausschaut ist es ja eng an Lightwave "gebunden").

Zum Programm selber sind mir zwei Sachen aufgefallen:
Wichtig wäre eine Möglichkeit das Pagent Motions "mixen" kann. Zum Beispiel das man das Kopfdrehen nach rechts (also nur Teile der Geometrie) mit dem Gehen oder Stehen u. s. w. zufällig oder durch Sensoren gesteuert kombinieren kann. So wie das auch mit Morphs oder MDDs möglich ist.
Interessant wäre auch die Möglichkeit die einzelnen Motions als MDDs abzuspeichern und diese dann in Pagent zu verwenden.

Soviel mal ganz schnell.


04-17-2013, 05:29 AM
This is looking great though I couldn't understand a single word in the commentary

04-17-2013, 06:36 AM
Anyone heard the birds chirping in the background? That is funny.

04-25-2013, 05:58 AM

Pageant is looking fantastic! Can't wait to see the next phase of development. Good luck!


04-25-2013, 06:35 AM
Anyone heard the birds chirping in the background? That is funny.

I heard them but I couldn't understand what they where chirping about....Must be German :D.

But nice video. I think I understand what he was talking about but I haven't heard any German in 10 years so obviously I missed some vital info ;)

Darth Mole
08-05-2013, 03:10 PM
Just bumping this back up - I see there's Pagent 2.0 on Pille's blog - any news?

08-05-2013, 08:38 PM
Darth - Thanks for bumping this back up. Somehow I had managed to sleep through and had missed it earlier.

Pilli - Wow! Looks wonderful! And please, do not worry about your english. I have a feeling that the first person who made comment over it, would be quickly beaten (in an fun, semi-polite way) by the rest of us who are SO thankful for your help and hard work.

08-05-2013, 09:19 PM
Good stuff man!

08-05-2013, 10:30 PM
Lightwave group should pick this up or help speed up development asap.

Darth Mole
08-06-2013, 01:09 PM
Well, you'd figure that it slots right in between flocking, instancing and Genoma. With a bit of NevronMotion for good measure.

08-06-2013, 02:57 PM
Yeah I would love to know how things are progressing. I am definitely interested in this.

12-09-2013, 07:13 AM
Any news?

02-07-2014, 11:51 AM
Did this die? :(

02-07-2014, 12:21 PM
hope not:(