Stateless generation

Stateless generation of distributed virtual worlds

JIŘÍ DANIHELKA, LUKÁŠ KENCL, JIŘÍ ŽÁRA

Stateless Generation of Distributed Virtual Worlds

Motivation Complex 3D virtual world shared on multiple

mobile devices

– the devices cannot store the whole world (the world can be potentially infinite)

– the world is generated on-demand based on view frustum

– we used infinite virtual cities as our main use case

– the mobile devices has to show the same geometry

– the devices has to work even when disconnected

Stateless Generation of Distributed Virtual Worlds

Paper idea How the problem can be solved

– mobile devices will share the same pseudo-random generator(share the same seed)

– each device will generate the content of its view frustum

– the content is consistent does not depend on previously generated geometry or states of other clients

– can work even offline

– we call our method “Stateless generation”

Stateless Generation of Distributed Virtual Worlds

Final result

Stateless Generation of Distributed Virtual Worlds

CityEngine CGA grammars

Stateless Generation of Distributed Virtual Worlds

Building modeling

Stateless Generation of Distributed Virtual Worlds

Transforming building to CGA grammer

Stateless Generation of Distributed Virtual Worlds

City generation pipeline

Stateless Generation of Distributed Virtual Worlds

Cars 2

Stateless Generation of Distributed Virtual Worlds

Related work Aliaga, Vanegas, Beneš:

Interactive Example-Based Urban Layout Synthesis

Stateless Generation of Distributed Virtual Worlds

Related work Aliaga, Vanegas, Beneš:

Interactive Example-Based Urban Layout Synthesis

Stateless Generation of Distributed Virtual Worlds

Related work Greuter S., Parker J., Stewart N., Leach G.

Real-time procedural generation of pseudo-infinite cities

Stateless Generation of Distributed Virtual Worlds

Our approach Delaunay triangulation

But can we do this on an infinite grid?

Stateless Generation of Distributed Virtual Worlds

Our approach Delaunay triangulation - proof

Stateless Generation of Distributed Virtual Worlds

Street segment interfaces Generated consistently using a hash function Cannot be overlapping

Stateless Generation of Distributed Virtual Worlds

Stateless generation Creating street nodes

(inspired by ray-tracing sampling)

Stateless Generation of Distributed Virtual Worlds

Stateless generation Creating possible street segments

Stateless Generation of Distributed Virtual Worlds

Stateless generation Minimum weight path

Angles can be transformed to edges using the graph’s dual form called “line graph”

Stateless Generation of Distributed Virtual Worlds

Stateless generation Lot generation

Stateless Generation of Distributed Virtual Worlds

Stateless generation Variations of street layout

Stateless Generation of Distributed Virtual Worlds

Measurements Stateless approach is not too demanding

Stateless Generation of Distributed Virtual Worlds

Measurements Rendering speed on a mobile phone (Lumia 920)

Stateless Generation of Distributed Virtual Worlds

Examples Big city

Stateless Generation of Distributed Virtual Worlds

Examples Antic Greece town

Stateless Generation of Distributed Virtual Worlds

Examples City suburbs

Stateless Generation of Distributed Virtual Worlds

Conclusions Advantages

– infinite worlds generated on-demand– consistent geometry in case of multiple clients– cities look more natural compared to previous work– can be used for scalable tasks (in cloud)

Drawbacks:– only for static content– path adding algorithm is greedy– currently only one level of streets

Thank you for attentionJiří Danihelka

danihjir@fel.cvut.cz

Stateless Generation of Distributed Virtual Worlds