Welcome on the simulator webpage of the Doh!Bots. You can find here the newest informations, screenshots and movies from the developement of our simulator.
Hope, you enjoy it!
If you have any questions, please don't hesitate to contact me: robert.schmidt[at]uni-dortmund[dot]de

After few years usage of the SimRobot simulator, we decided this year to give Microsoft Robotics Studio a try. It seemed to be a interessting plattform due to the service-oriented-architecture with the ability to distribute the services across the network. Furthermore, the easy usage of accelerated graphics and Ageia PhysX was also an attractive aspect - and one of the reasons for choosing this plattform.

For the simulated robot model (or DHB-1 SE, if you like ), I took the parts-models provided by Lynxmotion, which are also used in our real robot. The great torso-model was made by Adalbert. For assemblying the single parts to complex ones (e.g. bracket + servo + horn) I used Blender. More details on assembling the robot will follow. By now, here are some screenshots:

The first milestone of the simulator is finished. By now, we can connect to our robotcode through a TCP/IP connection and receive the joint angles from our walking engine. The simulated robot is now able to execute all kinds of preanimated moves (e.g. kick, stand up). I'm really positively surprised about the fact, that all moves - designed for the real robot - are also running almost without any adaptations in the simulator, and moreover - they are looking quite similar on both! Here are some movies:

• Walk
• Walk (other perspective)
• Special Actions (push ups, stand up, kicks)

Unfortunately, we had some problems with the simulation of the omni-vision cam. It's quite problematic since almost everything in our framework is based on the percepts that are computed from the image. It means: without camera - no autonomous robot. Therefore, a little workaround has been made: we directly push to the robotcode the absolute robot- and ball-positions from the simulator, so it hasn't to compute it explicitly.
The values are of course idealized, but for the development of first behaviour-scetches it's fine.

Yes, we already tested a simulation with three robots. The simulation ran on a P4 2.4GHz HT, 1024MB, Radeon 1600XT with about 10-15 fps. The simualted robots were controlled by robotcodes running on other machines across the network.

• simulated omni-vision camera
• simulated sensors (e.g. gyroscope, inclination)