Agents, objects and events. A computational approach to knowledge, observation and communication (Q2711199)

This thesis is concerned with the question how to model the information of an intelligent agent about a changing external reality in a computational way. It is assumed that the information of an agent can be represented in some dynamic computational structure which will be called a knowledge state: (a) private and subjective, i.e. relating an agent's knowledge not to the world as it is, but to the world as it appears to him, (b) realistic and useful, (c) dynamic in the way that the knowledge state of the agent can change through inference and belief-revision processes. The main subjects are representation forms of this kind of information, its computational realisation, and the relationship between this information and the environment of the agent. As to changes in external reality, the effects of external events to knowledge states will be explained. NEWLINENEWLINENEWLINEThe main approach can be summarized as follows: (a) A direct computational approach to the modelling of knowledge, using a framework in which concepts figure as first-class citizens, and are emphasized at the expense of individual objects' or 'entities'. Conceptual knowledge is related to an external world through an explicit account of observation and recognition. (b) To understand time and change, an ontology is developed that reflects the relationship between events and state changes of objects. (c) Communication is modeled in terms of the dynamics of the knowledge states of speaker and listener. NEWLINENEWLINENEWLINEContents: 1. Introduction; 2. Type theory; 3. Knowledge and observation; 4. Modeling dynamic objects: A model of a changing universe which is formalized in predicate logic; 5. Time, types and memory: Translation of the formalization to type theory; 6. Type theory and DRT: Knowledge states can be related to Discourse Representation Theory (DRT) which is associated to natural language interpretation (DRS: Semantic representations of DRT); 7. Inter-agent communication: Investigation of how an information transfer between knowledge states can be achieved; 8. Conclusions; 9. Appendix: A robot book (which is a `book' in type theory, and uses the observational data following from observations in 5.3).