Precision Navigation and Control Techniques for Autonomous Ground Vehicles
Joana Moll, A. Evers, Nicholas A. Baine, K.S. Rattan · Proceedings of the ION 2013 Pacific PNT Meeting · 2013
This paper reviews the design and implementation for a 7-State Kalman Filter that fuses IMU and GPS sensor data for use on an autonomous mowing robot that is also used for Simultaneous Localization and Mapping (SLAM). The Autonomous Team at Wright State University has developed a Robotic Lawnmower for the AFRL's Institute of Navigation (ION) Autonomous Lawnmower Competition. This project has won the competition several times, including first place for the year 2012. Although the competition was discontinued, further work was done in enhancing the operation of the platform. Past teams relied too heavily on using differential GPS with a Real Time Kinematic (RTK) solution (GNSS). The purpose of this research was to make the system less reliant on DGPS for situations when the satellite view is degraded. This was accomplished by performing multisensor fusion with IMU and GPS using an Extended Kalman Filter (EKF) using the Flat Earth assumption as well as providing mapping through techniques known as SLAM. The Flat Earth assumption necessitates only the folllwing 7 states: Position (North), Position (East), Velocity (North), Velocity (East), roll, pitch, and heading (yaw). SLAM is accomplished through LIDAR sensing and storage on an embedded computer system. This paper provides the theoretical foundation, design, and validation for this implementation.