Compressing ROS Sensor and Geometry Messages with Draco

Abstract

Current development in sensor technology allows to capture highly detailed geometric 3D environment models fast and online with laser scanners or RGB-D cameras. However, in most rescue and service scenarios, the available bandwidth between operator and controlled robots is limited. In this paper, we integrate Google's geometry compression library Draco into ROS to reduce the required bandwidth when sending geometry data from a remote-controlled robot to a RViz instance running on the operator's computer. Besides a prototypical implementation, we present an in-depth analysis of the achievable compression rates, accuracy and resulting latency when using Draco for this purpose.