Robotics simulation

I’ve recently started planning to create a sort of industrial robot simulation program. I got the idea when i saw a student in my school presenting his thesis, in which he had created a VisualStudio sort of development environment for writing the code for the robot. The problem with this is that you have to do all kinds o steps in order to get the compiled code to the robot without really knowing the end results. So my idea was to create a blender based simulator for this robot, where i would write a python based interpeter for the code. The robot would then perform as it would in real life, but in Blenders 3D environment. Anywho one of my questions is, how do you get the blender to pick something up, for example the robot has a claw that it uses to pick stuff up from the floor. Should I use the physics engine for this or?..

Since your talking a sim, (disclaimer! im not a big game engineguy) I would say, that either you can do some killer programming, or, record some action ipos. How you do that…Im not sure.

eh? you mean like in droidbattles for linux? basically a compiler that uses virtual inputs and outputs that can be assigned to stuff like ‘engine’ and ‘claw’? if so, It’d not be very hard to program… I’d use python, that interprets a python varient in blender’s text editor.

there is a robot sim made in blender, from what i remember you need to use armatures and program it to move with quats (or you could join joint movements together with actions).
here are some links I’ve dug up with google (using “blender robot sim”):

Best way to create an animation of anything with arms/legs is usually through keyframing IPO. If you want the bot to pick up something, just move the object with the model.

For years I have had a dream of driving a blender display (a running game, perhaps) from a database, where the blender game heartbeats and reads a MySql lite or other open source database, and uses the values to do something, like turn a square red when a value gets too high. A visual database monitoring program, if you will. It seems to me that if you modelled the robot and got the GE to read a database that had, for example, commands or positions, and the other software fed the database, you could achieve success.