Techniques for Modeling Simulation Environments for Modular Robotics

In modular robotics, complex structures can be formed from basic modules to solve tasks which would be difficult for a single robot. The development of techniques for adaptation and evolution of multi-robot organisms is the subject of Symbrion project Levi and Kernbach (2010). In the project, the bio-inspired evolutionary algorithms are massively simulated prior to run them on a real hardware. It is crucial to evolve behaviors of the robots in a simulation, that is close to a real world. Hence, accurate and efficient representation of an environment in the simulation is needed. The robots learn simple motion primitives or complex movement patterns during many runs of the evolution. The learned skills are then used during experiments with a real hardware. In this paper, we present methods for building 3D model of a real arena using a laser rangefinder. The resulting 3D models consist of triangles. They can be constructed in various level of details using state-of-the-art methods for 3D reconstruction. We will show, how the size of the models influences the speed of the simulation.