Theoretic foundations of situating cognitive vision in robots and cognitive systems

A Framework for Hierarchical Perception–Action Learning Utilizing Fuzzy Reasoning

“…It can also be used to model human intentional activity in protocol driven environment. We finally note that the presence of a deductive component within a system that naturally grounds symbols within a perceptionaction hierarchy means that the current system fulfils Sloman's criterion of a symbol tethering system [11].…”

“…The problem is thus one of the combination of abstract deduction with pre-symbolic input. In this way we address the wider cognitive science issues of symbol grounding [7] and symbol tethering (according to Sloman's usage [11]). …”

“…The latter scenario is a top-down process intended to exemplify Sloman's notion of symbol tethering [11], wherein an extensible logic system can build on a partially-grounded set of symbols in order to derive a fully grounded symbol set. Thus, ultimately, the system would be able to acquire e.g., novel sign types, by taking novel perceptual clusters identified by the computer vision system and attaching the appropriate functional semantic label to them from the a priori highwaycode rules.…”

“…Moreover, in applying P-A learning in a first-order logical context, we go beyond the classical problem of symbolic and sub-symbolic integration, extending it to that of symbol tethering [10], [11].…”