Natural Feature Tracking on the OPERA Maestro platform

This paper will present the results of porting the Extended Kalman Filter (EKF) Simultaneous Localization and Mapping (SLAM) Natural Feature Tracking (NFT) algorithm using the Automatically Tunable Linear Algebra System (ATLAS) for use in Tilera´s Tile64 or OPERA´s Radiation-Hardened By Design (RHBD) Maestro chip. ¹²This implementation of EKF SLAM was previously analyzed for performance on a RAD750 but performance limitations imposed by the platform raised concerns over the ability to operate at high rates while providing acceptable accuracies as target complexity increased in scale. The Maestro chip is being developed by the U.S. Government under the On-board Processing Expandable Reconfigurable Architecture (OPERA) program. OPERA is based on the Tilera´s commercial Tile64 chip with the addition of a floating point unit to each core.