Hrm ... sounds like a pretty darn big goal. Some points that might be helpful to ponder.
- Do you plan to let the user "build" his robot, or just have a few basic structures with room for light modification?
- If you allow modification, how do you try to "level" the playing field (since you're not spending real money or real time doing it, something has to limit the robots and make them competitive)?
- Do you plan on having AI bots to play against? Connect to other computers to get other human players in the game?
- I assume this'll be real time. What type of physical simulation do you (realistically) expect?
With that said, I would refer you to
Crystal Space, a very nice graphics/gaming engine that has cross-platform support, 3ds max import abilities, and MS VC++ work files (which all sound like it would be useful with what you have planned ... aslo, they use to have a rudimentary physics support, but I don't know what's happened to it). And you might want to consider
ODE, a pretty serious open source physics engine (for rigid bodies, which robots happen to be).
I'd take a good long hard look at the project, and try to come up with a reasonable set of goals. Remember that successful projects start out small and grow from there. The projects that start out with unrealstic aims tend to fail, as those contributing become disillusioned. Also, some sage advice from my limited experiences --
before you write a single line of code, have a very detailed outline of what the system will be like, to include class hierarchy and interaction, and other such details.
Edit: in regard to Texan's post, which happened to beat me ... I think it's a bit early to begin thinking about specific design issues. Start with a general outline of the project, to inclue non-coding issues too (e.g., a game's nice and all, but nothing without art ... and sound is pretty darn important too, and a whole slew of other issues). Then, work from there. Specific implmentation like what will be a class, etc., will come from the general design, and it's too early to say how you want to design the system. Or that's the way I see it, at least.