[BibTeX] [RIS]
Spatial Directionals for Robot Navigation
Type of publication: Inproceedings
Citation: TenbrinkWinterboer05
Booktitle: Thematic Session on Motion Encoding, 21st Scandinavian Conference of Linguistics, June 1-4, 2005
Year: 2005
Abstract: Previous research on spatial projective terms such as left and front typically focuses on static (locative) usages. In these approaches it is often assumed that dynamic (directional) applications, i.e., those expressing motion into a direction specified by an expression such as to the left or forward, can be (more or less) directly derived from insights gained on the interpretation of the locative expressions (e.g., Herskovits 1986, Eschenbach 2004). This assumption goes back to a proposal by Miller & Johnson-Laird (1976) who state that dynamic usages are closely interrelated to static ones, as reflected by the fact that the same basic expressions can often be used in both kinds of contexts. Without doubt, there is a high degree of overlap between these two kinds of applications of spatial terms. Therefore, the present approach starts from an overview on insights gained in the literature with respect to the application of locative terms. Crucially, their interpretation depends on the identification of underlying reference systems, which may be external or internal; relative or intrinsic; visual-field based, group-based or landmark-based; they may use the speaker's, the hearer's, or a third person's perspective; and so on. Altogether, a range of aspects contribute to enabling the identification of a spatial relationship in the real world. The automatic recognition and application of these insights in a robotic system is one of the aims of the major research project SFB/TR8 on Spatial Cognition, which this contribution is a part of. Due to their close relationship to locatives, the interpretation of directionals potentially involves similar complexities. For example, in the sentence "Put the cup behind the plate", the origin of the underlying relative reference system (necessary because plates do not possess intrinsic sides) still needs to be identified, while in "Move the chair behind yourself", the reference system is intrinsic and the relatum (the addressee) is identical to the origin. But often, the relatum is not mentioned explicitly, as when an entity is moving in a direction specified by a directional, e.g., in "turn left". Along with the usage of projective terms statically for goal-based instructions, our research has shown that speakers also frequently use the terms dynamically, indicating directions in which a robot might move, starting from the robot's current position. At first sight, the interpretation of such directionals seems trivial; indeed, it seems to be a common assumption that robots can be more easily instructed by indicating incremental movements rather than directly specifying the goal position. Unfortunately, the insights on locatives cannot directly be applied to directionals because of a number of aspects in which the two application contexts differ. For example, there is a difference in applicability between forward and to the front that deserves special attention. The latter expression, if uttered in a dynamic context, depends on the internal regions of a background object such as a train, while the former is restricted to the forward motion of the entity involved. Moreover, a brief instruction (to a robot) like right may be interpreted either as a rotation or a movement instruction; in both cases, the quantity of the movement needs to be determined. Thus, the analysis of the applicability conditions and the interpretational range of directional terms is an important research field in its own right. The aim of the presentation is twofold. First, we introduce the main aspects involved in the interpretation of projective terms, starting from basic insights on locative usages and then pointing out the relationship to directionals. Here, interesting cases such as potential ambiguities in interpretation will be addressed. Second, we discuss our robotic system that is capable of interpreting directionals uttered spontaneously by human users uninformed about the robot's abilities. This system was developed in several iterations on the basis of the results of experimentation. In this field, we discuss problem areas encountered during the development process, specifically with respect to their linguistic and cognitive implications, and discuss the solutions found in the current implementation. References Herskovits, Annette. 1986. Language and spatial cognition. Cambridge University Press. Eschenbach, Carola 2004. Contextual, Functional, and Geometric Components in the Semantics of Projective Terms. In L. Carlson & E. van der Zee (eds.), Functional features in language and space: Insights from perception, categorization and development. Oxford: Oxford University Press. Miller, G. and P.N. Johnson-Laird. 1976. Language and Perception. Cambridge University Press: Cambridge.
Userfields: project={I1-OntoSpace}, status={Other},
Keywords:
Authors Tenbrink, Thora
Winterboer, Andi
Attachments
    Notes
      Topics