Behind any robot task or interaction is a representation that should (a) enable sufficient contextualization; (b) support any existing predefined, learned, and/or reusable skills onboard the robot; (c) be verifiable at design time and behave consistently at run-time; and (d) can be tested, executed, and modified for reuse on a variety of different robot morphologies. Capturing the end user’s desired task or interaction as a computational artifact (i.e., a representation) has long played a pivotal role in robotics.

Many robotic subfields have traditionally employed a variety of different representational techniques, often borrowing from artificial intelligence (AI), to achieve their respective objectives. Examples include variants of LTL, logic, automated planning languages, languages for Belief-Desire-Intention, custom variations of these, and many more than we can list. The problem is that there is a lack of awareness of these representations in the research community, or how these representations can/should be applied.