Multi-Autonomous Ground-robotic International Challenge (MAGIC) 2010

Abstract

Abstract : This final report describes multiple Unmanned Ground Vehicles (UGVs) and their autonomous navigation techniques developed by the US-Asian Team that can safely expedite reconnaissance in cluttered urban environments. The developed multi-UGV system consists of four high-performance UGVs with efficient and intelligent autonomous navigation capabilities. They can detect, locate, classify, recognize and track suspicious Objects of Interest (OOI) and then neutralize any confirmed threats while mapping the entire indoor and outdoor environments. As a result of the development and implementation of the hybrid feature-based/scan-matching Simultaneous Localization and Mapping (SLAM) technique, the multi-UGV system can localize OOI and build a map within the targeted accuracy of 25 cm. The (1) implementation of the model-predictive control and the hybrid SLAM and (2) the Occupancy Grid Mapping (OGM) on the Graphic Processing Unit (GPU) after formulating them within a single recursive Bayesian estimation framework, this enables autonomous navigation at speeds of at least 8 km/h. Performance tests in two buildings and one outdoor area have shown the efficacy of the developed multi-UGV system.