Title: Image generating device and image generating method
Abstract: An image generating device for forming a behind object prepared with polygon data as a part of the background of a virtual three-dimensional space and generating images of the movement of a character object prepared with polygon data in this virtual three-dimensional space. A visual field seen from a viewpoint is operated, and behind objects OTa, OTb, and OTc positioned within this visual field and observable from the viewpoint are specified. Moreover, judged is whether it is the specific condition wherein the character object OTch is within the visual field and positioned in the space closer to the viewpoint side than the specified behind object. The character object OTch is incorporated as the object of display only when it is judged as being this specific condition. (end of abstract)
This is very interesting... I'll quote a rundown by 'Staplerfahrer Klaus' from Shenmue dojo:
Did anyone actually read it? It has a quasi-relevance to Shenmue in describing some similar elements, although I highly doubt it's related to plans for that particular series. It may indicate that Sega have plans for something vaguely akin to Shenmue.
A good part of it discusses a presumably novel occlusion culling technique in near-broken English. I found it difficult to understand, but it's good material for speculation. On modern machines (including all current-gen consoles AFAIK) the optimal way of doing graphics is to unload as much work as possible to the GPU, including pretty much all culling, yet they repeatedly describe their method as running on the CPU. So they're probably talking about fairly dated hardware. Perhaps the PSP or DS?
Another of their major points is a system whereby there are two progression modes, auto and manual, whose proportion can be specified at runtime, or determined at runtime for any specific event. Basically "auto" events are performed entirely by the machine with no player intervetntion, and "manual" events are executed in real-time with the player in control. Apparently the idea is that it allows greater customization of difficulty levels for individual players, and allows the developers to immediately change the game to make it more ideal for the market condition at any time (I don't like the sound of the latter). It is mentioned that some events can be invariantly auto, so I guess cutscenes can exist. They also state that their system could enable players to play "game elements" they excel at, and let the computer handle the parts they're poor at. So presumably Sega have something with quite diverse gameplay in mind. Finally, something bizarre: they mention "business use" of games built on this system. The only explanation I can think of is that Sega are investigating the potential of serious games.
Anyway, now the Shenmue-esque parts. They make a direct reference to a game built on their system having both RPG and action elements. Time of day changes in real-time (time acceleration isn't mentioned, presumably that depends on the specific game). Dynamic weather is briefly discussed, including "rain, sleet, and snow". They state that weather conditions can be different at different parts of the world, which may suggest that their system has the capability for large worlds, but I'm not sure. In the real world you can see differences in rain densities just looking down one short dead-end street. Weather states are "set in advance", but it's not clear how far; I imagine they're just completely pre-fabricated at dev time as in Shemue, especially considering that weather could be important to the storyline. Game-world items described are invariably modern ("telephone pole", "TV", "speakers") or of an ambiguous time but applicable to the modern world ("house", "warehouse", "remains of a building"). Note that the entire patent description appears to be related to a game engine, so they're probably just examples with no relevance to a particular game Sega is making.
Finally, procedural generation of indoor environments is discussed at reasonable length, indicating that it's a pretty important part of the system. It can be executed at dev time or runtime. I won't go into detail as their description is clear, but it sounds pretty versatile, creating elements of indoor environments from the walls all the way to small items, based on the building's use and other parameters (such as "tidy, untidy, deserted"). The stated aims of this part of the system are to reduce dev time and minimize game media reads. To speculate, the latter aim indicates the game engine discussed is for systems with either slow or severely capacity-restrained media. Alternatively, it could be that the unstated purpose is to enable fairly large worlds with a high proportion of explorable indoor environments. Really, there's no other reason they'd bother implementing the procedural generation system; it would be cheaper to just create the environments than tune and optimize the relevant algorithms. Further, even the Nintendo DS, the modern system with the smallest media capacity, could easily store nearly two million object positions/rotations using just 1/4 of the 256 MB cart. It and all other modern systems are hardly constrained enough by media->memory bandwidth.
I suggest everyone with an interest in Sega reads the document. Even though there's nothing of direct relevance to Shenmue, it does describe something a fuck of alot more interesting than anything else Sega has done in the last few years. To anyone taking the plunge, note that when they say "behind" they mean "background".
EDIT- I guess I should post a quick summary of the similarities. To anyone only bothering to read this, Sega are talking about a game engine, not a game. It's exclusively technical so it's really impossible to say whether there's any real relevance, but described are:
-Full 3D RPG/action gameplay
-Time of day changes in real-time
-Weather including rain and snow
-Plurality of indoor environments