Action Representation for Intelligent Agents Using Memory Nets

Memory Nets (Eggert et al.: Memory Nets: Knowledge Representation for Intelligent Agent Operations in Real World, 2019) are a knowledge representation schema targeted at autonomous Intelligent Agents (IAs) operating in real world. Memory Nets are targeted at leveraging the large body of openly available semantic information, and incrementally accumulating additional knowledge from situated interaction. Here we extend the Memory Net concepts by action representation. In the first part of this paper, we recap the basic domain independent features of Memory Nets and the relation to measurements and actuator capabilities as available by autonomous entities. In the second part we show how the action representation can be created using the concepts of Memory Nets and relate actions that are executable by an IA with tools, objects and the actor itself. Further, we show how action specific information can be extracted and inferred from the created graph. The combination of the two main parts provide an important step towards a knowledge base framework for researching how to create IAs that continuously expand their knowledge about the world.