Minecraft builds Manhattan on a 1: 1 scale



    Christopher Mitchell, a doctoral student at New York University, painstakingly works on a science project every day, collecting from Minecraft blocks an exact model of the Manhattan area , down to copies of individual buildings. The area of ​​the island is 58.8 km 2 , the building is very dense, so there is a lot of work. During the year, Christopher laid 71 billion blocks, and for the model had to use part of the university cluster with 300 processor cores and 200 GB of RAM.

    In many modern games, the action takes place on a map of New York, including Crysis 2, Crysis 3, Spider Man 2, Grand Theft Auto IV and Godfather II. Each has a map, but they all suffer from shortcomings: incomplete coverage and lack of detail, excessive styling and compression, writesArs Technica.

    Using the significant computing power of the university data center and various algorithms, Mitchell intends to make a fully-fledged scalable map, which other game developers will be able to use later. Minecraft players will also be able to play on a real city map.

    The scientist takes information from various sources, including Google Earth. He had to reverse engineer the Google Earth protocol to extract house models, and Mitchell worries that he might violate the license to use the service.



    1: 1 Manhattan modeling is part of an even larger project (!) Called Sparseworld. It combines the geodetic data of the USGS EROS service, three-dimensional building models from Google 3D Warehouse to create models. In principle, the Sparseworld virtual space can be expanded beyond Manhattan, there would be computing power.



    Sparseworld is based on the TopoMC project, the engine of which creates a digital model of the real world on a scale of 1: 6, and Mitchell rewrote TopoMC to simulate 1: 1. In the future, he plans to switch from Python to another platform and create an algorithm for importing tree information from USGS EROS to automatically place them on the map.

    Also popular now: