Self-Understanding and Self-Extension: A Systems and Representational Approach

Abstract

Abstract-There are many different approaches to building a system that can engage in autonomous mental development. In this paper, we present an approach based on what we term self-understanding, by which we mean the explicit representation of and reasoning about what a system does and does not know, and how that knowledge changes under action. We present an architecture and a set of representations used in two robot systems that exhibit a limited degree of autonomous mental development, which we term self-extension. The contributions include: representations of gaps and uncertainty for specific kinds of knowledge, and a goal management and planning system for setting and achieving learning goals.

Robot task planning and explanation in open and uncertain worlds

“…There are few integrated robot systems that combine domain-independent planning with domain-specific knowledge. Within that community, the need for a highlevel continual planning and execution monitoring subsystem is widely recognized [52,49,25,56]. The domainspecific knowledge in our approach is structured in three layers, as discussed in Section 1.…”

“…The restrictions on that work are that it assumes a prebuilt metric map, segmented into rooms, and does not reason decision-theoretically either about the value of the information, or the sensor unreliability. Hawes et al [15,52] also interleave planning with reasoning about a semantic map, but plan with models of the epistemic effects of actions. The use of a continual replanner allows the use of a closed-world planning domain in what is an open world.…”

“…Each competence maintains its own internal representation (e.g. a metric map), and is itself composed of multiple components that asynchronously update a competence-specific working memory, and transmit abstract information only to the knowledge layer [52]. Components in the competence layer either run continuously or are explicitly called on demand as needed.…”

Home alone: Autonomous extension and correction of spatial representations

“…For a more detailed account of the rest of the Dora architecture, see previous work, e.g. [9], [10].…”

“…We take the multi-layer hybrid approach to map representation described in [10]. Particularly relevant for this paper are the map's place and conceptual layers.…”