Mining Companies Borrow From Gamers' Physics Engines 39
littlekorea writes "Mining companies are developing new systems for automating blasting of iron ore using the same open source physics engines adapted for games such as Grand Theft Auto IV and Red Dead Redemption. The same engine that determines 3D collision detection and soft body/rigid body dynamics in gaming will be applied to building 3D blast movement models — which will predict where blasted materials will land and distinguish between ore and waste. Predictive blast fragmentation models used in the past have typically been either numerical or empirical, [mining engineer Alan Cocker] said. Numerical models such as discrete element method, he noted, are onerous to configure and demanding of resources — both computing and human — and are generally not appropriate for operational use at mines. 'The problem with empirical models, by contrast, is that they tend to operate at a scale too coarse to give results useful for optimizations,' he added, noting typical Kuz-Ram-based fragmentation models (PDF) (widely used to estimate fragmentation from blasting) assume homogeneous geology (the same type of materials) throughout a blast."
Re:I hope they do their math (Score:5, Informative)
Mostly that has to do with incorrect/inappropriate weights being assigned to gameworld objects.
Many world objects weight exactly the same as styrofoam, but somehow have enough kinetic energy when thrown to instagib the bad guy. Others are made to weigh 100x that of lead, but somehow actually get tossed by an explosion instead of simply pushed a little.
If they plug in sane values for mass, center of mass, "bounciness", elasticity, inertia, and gravity, the should get mostly useful simulations. Issues with air pressure (it is an enclosed space, with an explosive charge, after all) might cause problems, but adapting it with another added value as a delta to object vectors (with a fall off radius) would fix it somewhat.
The engine name (Score:4, Informative)
The engine name is Bullet and it's pretty good to work with. I have done some AR stuff with it:
http://www.youtube.com/watch?v=UpBL6eqcJ6A
http://www.youtube.com/watch?v=tnaIVvNKjek
Re:I hope they do their math (Score:4, Informative)
-Developer incompetence: the framework you use may be good, but if you botch it up when integrating or give it insane values like huge boulders weighting a gram, it's not gonna look realistic.
-Real-time limitations: physics is largely simulated iteratively, and thus the larger the step the worse the simulation as errors tend to appear and propagate more when the system is infrequently simulated; if the step is lowered, many problems entirely go away.
-Networking limitations: many games these days are built with the idea of a multiplayer mode, so the engine is geared towards minimizing data transmission and latency at the cost of accuracy.
-Game focus: the physics engine isn't the focus, so it's usually not given a whole lot of processing power, which once again forces approximations; it's also not unusual for things to be tweaked so that they look good, even if that makes them inaccurate, because it's a game.
I'm pretty sure you can get some fairly nice results from the better physics engines if you tweak them to match your needs.