Modeling, simulation and control of an agricultural mobile robot

This thesis is focussed on the modeling of an agricultural mobile robot for the development of an intermediate velocity controller in between the high-level planner and the low-level motor controller. This robot is currently operated by humans and it is currently developed towards becoming autonomous. A bond-graph model for the robot is created and parametrised in order to offer an object for control design in simulation. In the modeling, a special emphasis on agricultural conditions, i.e. low speeds and variable terrain structures, is given. Parameters are identified by various means, which include motor identification, weighing, curve fitting and linear regression. For this, the model is separated into simpler submodels to reduce the number of unknown parameters to be identified in a step. A PI controller is parametrised and tested in simulation, giving good results. However, its performance on the actual system is yet to be proven. Three different advanced control approaches, adaptive control, feedback linearisation and model predictive control, are theoretically outlined.