Scalability of Sensor Simulation in Ros-gazebo Platform with and without Using Gpu

Abstract

Autonomous mobile robots depend heavily on sensors for interpreting the external environment and planning their movement. Given the complexity of even simple environments, the use of simulation is necessary to test sensory perception and movement strategies efficiently. Accurate and efficient sensor simulation is key to the successful development and fielding of autonomous mobile robots. In this paper, we investigate the scalability of LIDAR-type sensor simulation in ROS-supported Gazebo, an open-source multi-robot simulator. Specifically, the performance and scalability of the simulator are assessed with and without support from GPU in two virtual worlds, one highly-complex in terms of the number of collision targets and one not-complex. In both cases, we increase the number of sensors and evaluate system performance. The results show that the use of GPU helps improve performance, providing better speedups in the complex scene. However, the number of sensors that can be used is limited when using GPU.