Simulation Environment for Testing Guidance Algorithms with Realistic GPS Noise Model

Abstract

This paper presents a simulation framework and realization for a tractor-trailer system with an active joint in an agricultural field setting. In addition, all real system measurements are also provided for hardware-in-loop testing. In the simulator, the kinematics and dynamics of the tractor and the trailer are modelled, as well as noises related to each measurement. The noise statistics and typical amount of wheel slip were identified from field tests. For hardware-in-loop testing the simulator provides the measurements in CAN-bus (ISO 11783 standard) and the commands to the steering valve are also received via the bus. CAN-bus is the only link between the guidance system under test and the simulation environment. In addition, the simulation environment contains a graphical front end, where the trajectories of the vehicle can be observed and analyzed.The most challenging noise related to the simulation of an environment is related to the GPS and its inaccuracies, as the noise properties are far from Gaussian white noise. In this paper an error model for GPS noise is presented as a position measurement and noise range statistics to the measurement. The simulation environment is tested with nonlinear model predictive control algorithms.