Agricultural robotics is a discipline necessary for the future health and welfare of humanity, yet it is harder than suspected by those outside the field. Many challenges remain, economic as well as the technological ones of performing in unstructured, harsh environments. A general methodology for the design and development of agricultural robots is proposed using the thesis’ central work, the Vegebot lettuce-picking robot, as an example. An embodied approach to design is recommended, with development based on rapid iterations of prototypes in the field, together with deep involvement of the customer. Three key building blocks are described for automating the harvesting of the iceberg lettuce under challenging and uncertain field conditions. First, the lettuces are localised and classified using a data-driven method; this process is referred to here as Detection. Second, the end effector is moved into position under conditions of high environmental noise and uncertainty; this process is referred to here as Approach. Third, the lettuces are harvested with a custom designed end effector that incorporates a camera, pneumatics, a belt drive and a soft gripper. The general name given to this process is Manipulation. The Vegebot project is described in detail and the derived methodology is outlined.