simplifying crowd automation in the virtual laboratory of archaeology

20
SIMPLIFYING CROWD AUTOMATION IN THE VIRTUAL LABORATORY OF ARCHAEOLOGY Luís Miguel Sequeira, Leonel Caseiro Morgado, Eduardo Solteiro Pires November 22

Upload: luis-miguel-sequeira

Post on 06-Jul-2015

226 views

Category:

Education


3 download

DESCRIPTION

Preview of my presentation tomorrow at SLACTIONS 2013. Be there or be square — it's free to attend (I think) and you don't even need to move from your chair, just log in via Second Life — http://www.slactions.org/2013/

TRANSCRIPT

Page 1: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

SIMPLIFYING CROWD AUTOMATION IN THE

VIRTUAL LABORATORY OF ARCHAEOLOGY

Luís Miguel Sequeira,Leonel Caseiro Morgado, Eduardo Solteiro Pires

November 22

Page 2: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Virtual Archaeology

Using computer models for visualising cultural artifacts and heritage sites

Recent work also include crowd simulation (no more empty sites!) Example: Uruk in SL

Recent: Formulate hypothesis and test them on a virtual lab of archaeology

Page 3: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Lisbon 1755Project started informally in 2005, formally around 2007

Goal: depict the city of Lisbon that was destroyed by an earthquake on Nov 1st, 1755

Visitors will be able to attend special events and interact with inhabitants (AI-driven intelligent agents)

http://lisbon-pre-1755-earthquake.org/

Page 4: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology
Page 5: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Historians FTW!

Early attempts at virtual archaeology: showing off technology, historians/archaeologists only as consultants... at the end of the project

Today: Methodologies like the London Charter put the teams under supervision of a historian

Result: scientific accuracy

Page 6: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Virtual worlds

No need for separate modelling/rendering pipeline (when using prims)

No need to develop the engine

Historians can also formulate hypothesis and move objects around without the need to talk to the technical team!

Page 7: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Make it simple!

But what about crowd simulation?

Crowd simulation is not new...

Virtual worlds are not new...

Intelligent agents are not new...

... all, however, require programmers to implement simulation parameters!

How can historians do that?

Page 8: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Typical simulations

How did people use the spaces?

Can we recreate events in a convincing way using AI?

Can we see emergent behaviour (peasants avoiding to bump into nobles, or seeking priests for a blessing?)

Agents interacting with human visitors

Page 9: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

The concept (1)

Classical model with perception/selection/behaviour modules

Movement is towards attractors/getting away from repellers (e.g. other agents) — typical from swarms

Agents pick up their own rules for better achieving their goals using genetic algorithms (GA)

Page 10: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

The concept (2)

For the GA, the fitness function depends on the length of the longest, unimpeded line of sight (as calculated by the ray-casting functions in LSL) and the nearness of the current goal

All parts of the simulation engine use “standard” algorithms; further research could explore alternatives

Page 11: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology
Page 12: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

What’s new?

It’s cruel to expect historians to learn about programming AIs...

Instead, we borrow ideas from games: agents have goals which are represented by cubes on the terrain

So to “programme” a goal, a historian just drops a cube on the ground. Simple!

Page 13: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology
Page 14: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Defining goalsEnergy level: Moving around reduces energy; goal is to find inn to eat/home to sleep

Money level: Agents earn a daily wage (motivation to go to job area); working reduces energy level

Happiness level: Working is boring, so agents will look for entertainment areas

Page 15: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology
Page 16: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Interaction moduleNPC belong to a ‘class‘ — e.g. nobleman, priest, peasant, merchant, foreigner, beggar...

Matrix of interaction between each class (peasant bows before nobleman)

Very simple: friend/foe/neutral but this may be changed later

Friend/foe act as potential attractors/repellers too

Page 17: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Summarising...

To set goals, historians place cubes around the location, and adjust them in real-time

To define NPC/NPC interactions, historians assign NPC to classes on a web page (no programming)

Page 18: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Status & LimitationsPrototype is not yet functional

Suitable only for small crowds (too many cubes otherwise!)

Difficulties making it work across SL and OpenSim (original requirement)

User validation was not performed yet

Overall concept fits requirements as per interviews, but might not be suitable

Page 19: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Future workCubes attracting just one NPC (which sits on it)

Combine cubes (sitting on one activates another)

Interaction matrix: per NPC and dynamically adjusted

Autonomously-moving cubes?

Games? Other platforms?

Page 20: Simplifying Crowd Automation in the Virtual Laboratory of Archaeology

Thank you for listening!

I welcome your feedback and comments!

[email protected]