Long-Term Simulation of Dynamic, Interactive Worlds with MORSE

  • Marc Hanheide
Schlagworte: DDC: 004 (Data processing, computer science, computer systems)

Abstract

In this talk I present recent work in the context of the European Integrated Project STRANDS (Spatio-temporal representations for Cognitive Control in Long-term Activities, http://strands-project.eu) that employ the MORSE simulator. In particular, I present three application domains to facilitate long-term simulation of dynamic worlds within that project. First, MORSE is employed in a continuous integration and testing framework helping to ensure high code quality, not only on compilation levels, but also on a level of system integration and deployment. MORSE is being used to run system level tests defined for the Jenkins continuous integration platform, enabling STRANDS to maintain a high-level of code consistency required to successfully participate in events such as the Robot Marathon. Secondly, I outline the use of MORSE, and particularly its flexible build scripts, to automatically generate worlds from Qualitative Spatial Relations (QSR). These allow to define a world qualitatively on the level of builder scripts with defined probability distribution of metric object positions and orientations. In STRANDS, this ability is exploited to generated randomised world in a controlled way, both to study the formation of QSRs and to generate randomised worlds for the before-mentioned testing framework. Finally, I present our work on Human-Robot Spatial Interaction and our preliminary efforts to simulate crowds in a robotic simulator. This final contribution is mostly work in progress with implementation in MORSE still pending. But initial results have been obtained to simulate thousands of agents in a simulated airport environment by using a hierarchical representation of vector maps to generate individual trajectories for agents. This work will lead to a more realistic simulation of human-inhabited environment and is use in STRANDS' research on human-robot spatial interaction.
Veröffentlicht
2014-12-31