Infinite Fun Space

Large scale virtual worlds can require vast amounts of data both to describe their shape and appearance and to fill them with interesting content. This data must typically be authored by artists and programmers, placing great demands on time and space to create and store it. At runtime, the data must be loaded, sorted through and filtered to present the user with a compelling experience that does not overwhelm the computing and rendering resources available. Distributed systems, such as massively multi-player on-line games, have the additional bottleneck of needing to fetch the data from remote servers.

Infinite Fun Space explores a number of techniques for enabling large scale virtual worlds including on-the-fly procedural generation of the world and its content, hierarchical level-of-detail management of the world data, and plausible emulation of object behaviors through event-based simulation.

• minimize per-frame tests through predictive, deferred event-based object simulation
• use plausible emulation of low priority behaviors to eliminate events
• compute n-body interaction between objects of widely varying sizes using the hierarchy of nested offset grids
• on-the-fly, context sensitive procedural terrain generation
• efficient terrain tessellation