Interview: Atomontage Engine
By Kyle Mann | August 7, 2011 | Interviews | 1 comment | Share
With all the recent buzz over the Unlimited Detail team’s 2011 tech demo, we thought it’d be a great time to check in with Branislav Siles, the man behind Atomontage Engine. Atomontage is a voxel-based middleware solution that if implemented in modern games, promises both realistic physics by atom simulation, as well as greater graphical fidelity. Siles spills the beans on Atomontage Engine’s current state, future goals, and competing solutions such as Unlimited Detail.
Tell us a little bit about Atomontage Engine and its long-term goals.
Atomontage Engine (AE) is basically a physics simulator with the capability to render the simulated environment. It is a hybrid engine – it understands both polygons and atoms. The engine is meant to be fully atom-based storing all relevant data in form of tiny elements, most of which are voxels. My long-term goal is to make AE the first fully volumetric atom-based complex physics-enabled product that will help (or somehow force) the video-game industry to switch to the atom-based paradigm. The ultimate goal is to make AE a distributed massive asynchronous system. In such setup millions of networked computers will be running a massive atom-based world comprising of zillions of atoms. None of the computers will maintain a complete copy of the world. Instead each of the local worlds will be an observation-based subset of the whole thing. The asynchronous observation-driven approach will make it actually possible (for little cost) to maintain a consistent massive physics-based world using hardware with limited resources and networks with limited bandwidth.
How do you see advanced voxel-based engines augmenting gameplay in popular construction-based games like Minecraft?
For me a voxel is a volume element with no complex internal or surface structure. So currently there are not many construction-based voxel-based games out there. The recent success of Minecraft is the result of the clever design of the product as a whole rather than of the voxel-like look of it. I believe that in the coming years construction-based games will become more common and some of them will benefit extremely from their atom-based nature especially if the games are physics-based. You know physics can be a great tool to modify/build the environment with or without the involvement of humans.
What have been the biggest challenges in implementing the team’s vision for Atomontage Engine?
Since 2004 I knew what I wanted to do and somehow also how to do it right. Sadly from 2007 to 2009 I had little time for developing AE which almost resulted in giving up the development of the engine at the end of 2009. Today the biggest challenge for me is to finish the most relevant functionality even without the necessary budget of time and money. My (prospective) colleagues cannot work for me without being paid on a regular basis so my top priority is to release the first executable demo and then get the budget I need once the demo receives the attention it deserves.
Do you have milestones you are hoping to achieve in the coming months?
I do but usually I am not reaching milestones in the planned order. I’m still prototyping some of the most interesting features of AE (chemistry, continuous erosion, etc.) and the most efficient way of implementing them doesn’t seem to be one that could be rigidly planned. The next big milestone is the releasing of the executable demo. But before that I want to demonstrate the 1st person view capabilities of AE – that’s because too many people still believe that voxels will always look badly in close-ups and are therefore useless for 1st person view games. I also want to demonstrate very basic chemistry and possibly some thermodynamics-based effects, too. Animation of voxel-based geometry will be the next milestone but it is of a lower priority as it is not required in the first simple AE-powered game that is being developed here.
What’s your take on the Unlimited Detail engine, which promises limitless “cloud point data” to process billions of points in real-time, and has gotten quite a bit of buzz going recently?
Well UDT seems to be very effective in generating lot of buzz every time they release a demonstration video. That’s a great thing actually. All the buzz helps the other developers of atom-based engines to promote their work, and it’s for free! I know nothing about the real parameters of UDT. Still the numbers they are presenting are easy to understand. The level of realism provided by the atom-based content is not proportional to the number of atoms used. It is proportional to the amount of information present. You can define tons of useless geometry using trillions of voxels compressed down to one kilobyte. But you cannot store a megabyte of useful information of high entropy in the same way. At least not in our Universe. So on the current hardware you can either have quadrillions of boring, almost useless atoms or just billions of interesting , useful ones. The point is in storing only the relevant part of the information in the most efficient way possible while making the data structure fast and versatile enough for the renderer, physics simulator, AI, etc. The only such way I know is using a versatile fully-volumetric (not a SVO) multi-level-of detail voxel-based data model with incomplete data sets (AE uses one such). The repetitive geometry presented in the UDT videos says nothing about any of these key aspects of a high-end realtime atom-based technology.
Another thing is that realistic rendering is slowly becoming routine while realistic physics isn’t. I’m quite sure that in less then a decade physics simulation will become a thousand times more resource-demanding then rendering. There is no point in developing a perfect renderer without making it a great physics simulator first. The comprehensibility of the problem of developing a commercially successful technology of this kind is on a totally different scale than that of developing a search algorithm.
What are the unique aspects of Atomontage Engine over other voxel-based projects?
I believe that all big players are already developing their voxel-based technologies and I’m quite sure that at least few of the engines share some of the relevant aspects of an efficient atom-based data-model with my engine. Therefore I think that while AE might be the most ready engine out there (theoretically I could release it any day) most of its entrails are not that unique. Still there probably is one key feature I haven’t talked about in the past – it’s the “observation-driven processing” approach I am currently developing (in this context observation is not equal to perception). It is the key component in the yet-to-be-developed massive asynchronous world-model. This particular feature makes it possible to work with extremely large number of elements on a limited hardware in real-time. A fast atom-based data model is a prerequisite to make the feature easy to implement.
How has public response been to the footage and other media the team has released?
Every time I’ve released a video or screenshots I got a number of e-mails from fans and also many have posted their opinions on a number of forums. It has became an almost continuous stream of e-mails after I released the “Playing with LODs” video. It’s really nice to see that people like the project and believe it’s real and capable to what they’ve
seen in the videos. I also got responses from a number of game development companies as well as from a few investors. Obviously there is a real although still a bit silent movement towards an atom-based future of computer games. And with time this shift in attitude is getting more and more massive.
About six months ago, you issued a progress report and a request for financial support to continue the project. Have you reached your funding goals to continue development for the next couple of years?
I got the feedback I wanted but after my communication with a number of companies I’ve decided to release the executable demo first. After the release I will talk with the companies and investors again, if still necessary. The thing is that I cannot waste too much of my time and, if I decided to work on a funded product, it could be particularly hard to dedicate myself to two different tasks – one being the implementation of the product, the other the implementation of the observation-driven approach. There are also a few other ways to make AE a successful project. I could try Notch’s way to success or maybe I could open the project. It’s really hard to say at this moment. The first part of my plan ends with the release of the executable demo. The rest is unknown.
Do you see Atomontage Engine and other voxel solutions as commercially viable in coming years? Will there be a day when virtually all 3D games employ engines like Atomontage?
I believe that the transition to atom-based virtual reality is definitely unavoidable. It is not easy to explain without going into technical details. The point is that working with atoms will be much cheaper then working with polygons. Just think about realistic physics. Better graphics won’t make games more realistic and fun, better physics will. I am sure all big players are clear about this for many years. So the question is not whether but when will all the top engines become atom-based. My guess is that most of them will be upgraded to fully-featured hybrids sometime between the releases of the next two generations of consoles. That could be in 3-8 years from now.
What applications does Atomontage potentially have to offer beyond the gaming industry?
I think that the target market for AE (or some parts of the engine) could be even greater than that of the existing 3D rendering engines and simulators. I also think that many not yet existing applications will emerge soon after atoms become more common in the gaming industry. AE could get integrated into CAD software, educational tools, interactive medical software but also into simulation and logistics software used in some of the heavy industries. Based on the feedback I’ve received during the past 18 months the most promising fields are military stuff as well as heavy industry applications.
In what ways can gamers and developers alike help support the Atomontage project, financially or otherwise?
Currently the best way to support the project are the donations and the word of mouth. People are already talking about the engine and all the buzz has already resulted in a slight increase in the received donations. Also my compile-time generators could make great use of a few more multi-core notebooks. Such a notebook can be partially damaged, just the CPU, memory and LAN card have to be 100% O.K. Currently all new content is being generated on my dual-core dev notebook and that one is already quite outdated and too slow to generate the required amounts of data.
Thank you for your time!
Thank you very much!
Nice interview, Kyle!