Simulation in the LAAS Architecture1 (Extended abstract)

While software for robotics cannot avoid being tested for real, it is often difficult and time-consuming to do so. In multi-robot developments, one may be unable to fully experiment its software since putting an experiment into place for a few aerial and terrestrial vehicules can quickly become difficult to do. In these cases, being able to extensively test in simulation is a very important thing. We present here a simulation environment adapted to the LAAS Architecture1. This architecture is designed for the control of autonomous robots. The simulation system has been designed to address the major issues in this kind of problems: • being able to use the same code in both simulated and real environments. • being able to reproduce the behavior in non-simulated systems: sensor accuracy, execution latencies in communication (inter-robot and hardware-to-software). • scalability in terms of robots and world size for multi-robot simulations. We present first a quick overview of the LAAS architecture, then the simulation system, and how we did achieve to address the three points outlined above. I. THE LAAS ARCHITECTURE The LAAS Architecture[1] for Autonomous Systems is a now classical 3-layer architecture. • the low-level layer is a functional layer based on the GenoM tool[2]. In GenoM modules, there are two means o