Realistic Autofarming Closed-Loop Tractor Control over Irregular Paths Using Kinematic Thomas Bell, Michael O’Connor, V. K. Jones, Andrew Rekow, Gabriel Elkaim and Bradford Parkinson (Stanford University, CA, USA) High-precision ‘ autofarming ’ makes possible farming techniques previously impractical using metre-level Differential -based control systems : techniques such as tape irrigation, the elimination of guess rows, and precise contour farming. A Carrier-Phase Differential positioning and attitude system with centimetre-level and accuracy was installed in a large farm tractor. Four types of trajectories (lines, arcs, spirals, and curves) were identified as basic building blocks necessary to generate a ‘ global ’ trajectory for a realistic autofarming path. Information about each trajectory type was translated into reference state specifications that a linear controller used to control the tractor over velocities between and ms to within approximately cm ( σ) without implement and cm ( σ) with implement on sloped terrain using a previously developed tractor model. These results are a significant step towards a realistic autofarming system because they not only demonstrate accurate control over various realistic operating speeds but over different types of trajectories necessary for a commercial system. . . At Stanford University, research into centimetre-level automatic tractor control is an outgrowth of previous positioning research involving aircraft. – Because a farm typically has good satellite visibility, agriculture was chosen to research practical land-based applications of Carrier- Phase Differential (). At Stanford, we have coined the term ‘ autofarming ’ to include not only precise positioning and control of agricultural vehicles, but also other technologies which could be integrated into such a precise control system. There are three components to high-precision autofarming : measuring the tractor location to within centimetres ; determining the desired tractor trajectory satisfying some user-defined requirement (such as spraying a field with minimum overlap) ; and actually controlling the tractor to move along the reference trajectory. The first component has been met through . The second component involves robot motion planning. This paper focuses on the third component : a control system that steers the tractor along basic trajectory types including lines, arcs, spirals, and curves. The majority of realistic trajectories would incorporate a combination of these four ‘ building blocks ’. Arc and spiral paths are often encountered in circular irrigation patterns. Curved trajectories could come from field boundaries (such as a stream) or from contour farming.