Generative Robot Design Using Text2Image Models

Master's Thesis

Ongoing Master's Thesis


Explain the importance of robots today and in the future. Show that robots can be applied in a vast variety of scenarios. Every scenario comes with specific requirements for functionality, but also for the design. Here we focus on the design, since currently robots are mostly designed manually. With the growing number of robots, manual tasks should be minimized. State the importance of the design for robots. For example, robots for human interaction, need to be accepted and appealing to increase usage and trust towards the robot. On the other hand industry robots are mainly rated on their functionality, but might also benefit from a modern design for marketing and a good design can support the perception of quality. The domain specific requirements are part of the research and need to be considered when designing.

To address these challenges, a systematic approach for creating robot designs should be implemented. This will make the design process a more structured and scientific and rely less on a good manually imagined design idea. The starting point is using a text2image Model where requirements for the robot design can be used as an input. This can iteratively be refined by further specifying the input and selection of promising outputs of a set of images. This approach is based on the implementation by Robinson [Rob23]. This implementation will be the initial starting point for this Master Thesis.

NAME: Arne Bartenbach

Main Examiner


To the top of the page