Designing an optimal robot is a challenge. First, building and evaluating any novel robotic system is expensive and time consuming. Second, today’s robotic systems are not universal and thus improvements can only be assessed in terms of the robotic system’s intended application. We are interested in developing novel computationally assisted processes to design better robotic systems which are deployed and evaluated in real world applications, e.g. agri- and viticulture (such as precision weeding, automated harvesting, plant phenotyping), glacial environments (sampling and mapping) and extreme terrain exploration.