Autonomous agricultural robots have experienced rapid development during the last decade. They are capable of automating numerous field operations such as data collection, spraying, weeding, and harvesting. Because of the increasing demand of field work load and the diminishing labor force on the contrary, it is expected that more and more autonomous agricultural robots will be utilized in future farming systems.

The development of a four-wheel-steering (4WS) and four-wheel-driving (4WD) robotic vehicle, AgRover, was carried out at Agricultural Automation and Robotics Lab at Iowa State University. As a 4WS/4WD robotic vehicle, AgRover was able to work under four steering modes, including crabbing, front steering, rear steering, and coordinated steering. These steering modes provided extraordinary flexibilities to cope with off-road path tracking and turning situations. AgRover could be manually controlled by a remote joystick to perform activities under individual PID controller of each motor. Socket based software, written in Visual C#, was developed at both AgRover side and remote PC side to manage bi-directional data communication. Safety redundancy was also considered and implemented during the software development.

One of the prominent challenges in automated navigation control for off-road vehicles is to overcome the inaccuracy of vehicle modeling and the complexity of soil-tire interactions. Further, the robotic vehicle is a multiple-input and multiple-output (MIMO) high-dimensional nonlinear system, which is hard to be controlled or incorporated by conventional linearization methods. To this end, a robust nonlinear navigation controller was developed based on the Sliding Mode Control (SMC) theory and AgRover was used as the test platform to validate the controller performance. Based on the theoretical framework of such robust controller development, a series of field experiments on robust trajectory tracking control were carried out and promising results were achieved.

Another vitally important component in automated agricultural field equipment navigation is automatic headland turning. Until now automated headland turning still remains as a challenging task for most auto-steer agricultural vehicles. This is particularly true after planting where precise alignment between crop row and tractor or tractor-implement is critical when equipment entering the next path. Given the motion constraints originated from nonholonomic agricultural vehicles and allowable headland turning space, to realize automated headland turning, an optimized headland turning trajectory planner is highly desirable. In this dissertation research, an optimization scheme was developed to incorporate vehicle system models, a minimum turning-time objective, and a set of associated motion constraints through a direct collocation nonlinear programming (DCNLP) optimization approach. The optimization algorithms were implemented using Matlab scripts and TOMLAB/SNOPT tool boxes. Various case studies including tractor and tractor-trailer combinations under different headland constraints were conducted. To validate the soundness of the developed optimization algorithm, the planner generated turning trajectory was compared with the hand-calculated trajectory when analytical approach was possible. The overall trajectory planning results clearly demonstrated the great potential of utilizing DCNLP methods for headland turning trajectory optimization for a tractor with or without towed implements.