System Identification of a Farm Vehicle Using Carrier-Phase Differential GPS

The automatic operation of farm vehicles can have great benefits both in farm productivity and hazardous or impossible operations. Automatic control offers many potential improvements over human control; however, previous efforts have failed largely due to sensor limitations. Carrier Phase Differential GPS (CDGPS) is an enabling technology that provides a high-bandwidth, lownoise measurement of multiple vehicle states. System identification techniques can then be used to generate a mathematical model for automatic control system design and implementation. In this work, previous controls research on a large tractor test bed is extended to demonstrate two different methods of system identification. Using a priori knowledge of the tractor dynamics, an extended Kalman filter is implemented and demonstrates model parameter identification. A Linear Quadratic Regulator (LQR) controller, based on these parameters, performs closed loop line tracking with a demonstrated error of better than 1.8 cm standard deviation. The same data is used with the Observer/Kalman Filter Identification (OKID) method, which assumes no a priori information about the system dynamics. It is shown that the estimator/controller designed with this system demonstrates equivalent experimental performance. The OKID methodology differs from the extended Kalman filter by utilizing solely the input and output streams to determine the structure and order of the plant model. INTRODUCTION Autonomous guidance of ground vehicles is not a novel idea. Previous attempts have largely failed due to sensor limitations. Some experimental systems require cumbersome auxiliary guidance mechanisms in or around the field [1,2]. Others rely on vision systems that require clear daylight, good weather, or field markers that require deciphering by pattern recognition [3,4]. Since the advent of modern GPS receivers, a single, low-cost sensor is now available for measuring multiple vehicle states. GPSSystem Identification of a Farm Vehicle Using Carrier-Phase Differential GPS Gabriel Elkaim, Michael O’Connor, Thomas Bell, and Dr. Bradford Parkinson, Stanford University