Development of a Geospatial Data-Sharing Method for Unmanned Vehicles Based on the Joint Architecture for Unmanned Systems (JAUS)

摘要

A task performed almost effortlessly by humans, perception is perhaps one of the most difficult tasks for autonomous vehicles. While substantial research has been done to develop these technologies, few studies have examined ways for multiple heterogeneous unmanned systems to cooperate in their perception tasks. Our study examined ways to model both perceived and a priori geospatial information, and formatting these data so that they can be used by the growing unmanned systems community. We introduce a perception system model, consisting of distributed smart sensors. This system of sensors was developed for the Team CIMAR entry into the inaugural DARPA Grand Challenge autonomous vehicle competition held in March 2004. The Smart Sensor Architecture proved to be a power method of distributing the possessing of sensor data to systems developed by engineers who best knew a particular sensor modality. By standardizing the logical, transport, and electrical interfaces, the smart sensor architecture developed into a powerful world modeling method. We also investigated current geospatial data-modeling methods used in the unmanned systems and geodetic information systems (GIS) communities. Our study determined the commonalities among current methods and resulted in a first- generation geospatial data-sharing standard for unmanned systems compliant with the Joint Architecture for Unmanned Systems (JAUS).