8 ways level of detail could improve digital twins

Level of detail (or LOD) is an age-old graphics optimization technique widely used in the game industry. It has traditionally been referred to as a method for structuring visual information to prioritize detail and the associated complexity for nearby or essential objects. This allows game engines to generate a richer experience with less computational overhead. 

Over the years, similar techniques have been applied to other domains, such as structuring satellite data to allow you to smoothly zoom in on Google Maps without first having to download the entire Earth. The game development community has expressed an interest in adding LOD support to emerging digital twins standards such as USD.

Different approaches for implementing LOD in digital twins and metaverse use cases could also improve communication, planning, analytics, data management, user experience and artistic expression.

Here is a walkthrough of how these different approaches to level of detail work across different realms:

Better graphics

James Clark introduced the notion of the level of detail in computer graphics in a 1976 article in the Communications of the ACM. He suggested an approach for “structuring the environment” to improve rendering by avoiding unnecessary computation. Clark went on to co-found Silicon Graphics in 1982 and then Netscape in 1994. Over the years, pioneers perfected techniques for a continuous level of detail that allow players to smoothly increase resolution as players move through virtual worlds or zoom into far away objects. Foveated rendering takes advantage of the extreme difference between central vision (the fovea) and peripheral vision to shave details in less noticeable places. The latest game engines like Epic Unreal 5 are getting better at intelligently selecting which LODs to render for massive 3D scenes.  

Cartographic generalization

Singapore’s country-scale digital twin uses level-of-detail features built into CityGML to characterize the resolution of data sets describing buildings, roads, and other physical features. In this context, LOD depicts the complexity of structures, amount of data and semantic aspects of data. For example, CityGML has five categories ranging from LOD-0 for representing regional landscapes to LOD 4 for characterizing architectural, structural and functional elements in rich detail. This approach also helps aggregate multiple objects that may not be relevant to a particular query to a lower level. 

Planning generalization

The architectural, engineering, and construction industry uses a related concept called Level of Development in Building Information Modeling (BIM) to characterize changes in technical design depth across a project’s development process. It describes the level to which planning teams have fleshed out the specifications, geometry and attached information. In the early stages, planning groups may just want to quickly estimate the overall cost and complexity of a project before proceeding. Later, domain experts such as electricians, plumbers and structural engineers can plan out exact gauges of wire and pipe in richer depth. These later levels of development can help plan orders and schedule the construction sequence so that teams do not interfere with each other. 

User experience

In good experience design, it is often helpful to guide a user’s attention to a particular detail. For example, it might be more beneficial to highlight the exact screws a repair technician needs to remove rather than render a scene in complete detail using an augmented reality overlay. Researchers believe that using LOD for glanceable interfaces could clarify complicated repairs and procedures. In musical concerts, visual augmentation with LOD could enhance the audience experience. 

Analytical exploration

Data exploration and analysis can often benefit by analyzing the relationships of things across different granularities of time, space or domain. For example, a business analyst may be more interested in how various business departments or regions performed across days, months, or years. Sometimes knowing how sales went compared to the same season last year is more interesting than last month. Tableau pioneered the concept of Level of Detail Expressions to help quickly recategorize raw data to help find valuable insights. Similar techniques could help surface new insights buried within digital twins by exploring the underlying data from different viewpoints. 

Conserving data

The data sets powering some larger digital twins can grow into the petabytes and beyond. New specifications like 3D Tiles Next can stream the appropriate level of detail to keep pace with interactive exploration for these massive datasets. The 3D Tiles spec supports hierarchical level of detail (HLOD) so only visible tiles are streamed, saving networking bandwidth. The spec is already powering various digital twins for streaming game-like geospatial experiences like Maxar’s and Blackshark’s earth-scale digital twins and Cesium’s 3D geospatial engine for Unreal. 

Artistic creativity

Bokeh is a popular photographic technique that allows creators to blur the background to highlight a subject artistically. Innovations in the 3D creatorverse could extend these techniques to the metaverse and digital twins to enable new kinds of artistic expression. Newer light field displays might use these techniques for new types of game and entertainment experiences. 

Communicating clearly

Good writers think through different levels of abstraction when explaining concepts to diverse audiences. A description for a fellow geek may elaborate in-depth on the benefits of specific file formats or technical specs. A more general audience may be satisfied to know which approach is safer, faster or cheaper. Similar approaches would allow us to use a single digital twin or model to communicate to different audiences. 

In an industrial context, mechanical engineers, electrical engineers, business managers and technicians will be interested in different details. Similarly, medical researchers, doctors and patients would likely see more benefit from varying levels of detail from a medical digital twin. Natural language processing tools summarize books today. Similar 3D modeling summaries could clarify important details in digital twins to broader audiences.