Planning optimal paths: A simple assessment of survey spatial knowledge in virtual environments

Abstract

In spatial cognition studies several cognitive factors were analysed in order to identify the aspect that could constitute the basis for the capacity of organising spatial knowledge into survey maps.

This study presents a method for evaluating spatial ability, based on the capacity of obtaining a survey-type spatial knowledge organisation, in a recently explored virtual environment. The ability to plan optimal paths in virtual environments was examined in 40 female adult subjects. Spatial evaluation deriving from navigation of a simple virtual environment was compared with classical spatial survey tasks (wayfinding, pointing and sketch maps) performed after the active exploration of a complex virtual environment.

Results show that there is a relationship between planning optimal paths and other spatial tasks related to survey representation.

These findings highlight how the navigation-supported learning capacity results in a predictive factor for individuals’ assessment of spatial ability.

Introduction

Virtual reality-based environments offer an interesting opportunity for the study of spatial cognition (Morganti, 2003, Péruch and Gaunet, 1998, Tlauka and Wilson, 1996). Flexibility is one of the major virtues of this type of synthetic environment: the layout of the environment can be systematically manipulated and different kinds of interactions can be designed in order to create suitable experimental conditions. Furthermore, virtual environments allow monitoring and recording the behaviours through which an explorer gains spatial knowledge for further evaluation.

Moreover subjective involvement at a personal level in a highly interactive system, such as a virtual environment, allows people to experience the cycles of perception and movement that are the basis for the construction of mental representations of space. In fact, being active has been acknowledged as a key factor for spatial learning in ecological conditions.

Like many studies carried out in natural and virtual environments, it is necessary to control the influence of individual differences in the ability to orientate themselves in space (Lawton et al., 1996, Malinowski and Gillespie, 2001). In particular, there is evidence that variability between subjects in spatial task performance is higher in virtual than in natural spaces (Bliss et al., 1997, Klatzy et al., 1998, Waller et al., 1998, Witmer et al., 1996). By comparing exploration in natural and virtual environments, studies have concluded that most of the abilities involved in learning in a natural world, are also needed for learning in a virtual world, but the latter presents additional demands. Consequently, the need to create an evaluation tool specific for virtual environment application is deeply felt in order to get reliable data (Belingard and Péruch, 2000, Richardson et al., 1999, Waller, 2000, Waller, 2005).

Up to now several different methodological approaches have been used in the assessment of an individual’s navigational abilities and different tools have been developed, such as auto-evaluation questionnaires (Lawton, 1994, Pazzaglia and De Beni, 2001), evaluation of general cognitive level (Juan-Espinosa, Abad, Colom, & Truchaud, 2000), mental rotation tasks (Just and Carpenter, 1985, Casey, 1996) or specifically suited visual-spatial tasks (Colom et al., 2003, De Vega, 1994, Denis, 1996, Poli, 2000, Shah and Miyake, 1996). Given the moderate correlation shown between the results of these kinds of evaluation tools and the navigational ability in a virtual environment, the problem in designing an effective assessment tool is still open (Bailey, 1994, Darken, 1995, Richardson et al., 1999, Waller, 2000).

Even if the specific factors investigated in the previously mentioned studies can be considered very influential in spatial performance, if taken one by one they do not allow for the definition and prediction of global ability to perform complex spatial tasks. In our opinion, a more promising way could be to derive navigational ability from the types of spatial representations that an individual is able to produce in order to adaptively interact with space within a given activity. Our methodological proposal is therefore primarily based on a conceptual framework about spatial representations.

It is a largely shared opinion that spatial knowledge of large-scale environments is organised in two types of mental representations or cognitive maps, route and survey maps (Chown et al., 1995, Golledge, 1990, Golledge, 1999, Kitchin and Freundschuh, 2000). The characterisation of these maps has been debated, but it is generally agreed that in route maps the environment is represented in a viewer-centred, egocentric frame of reference that reflects the person’s navigational experiences, while in survey maps distant places are linked together to form an integrated global overview of the entire environment.

Cognitive maps are useful for a wide variety of purposes, but fundamentally for wayfinding. In this activity, the representations serve to aid navigation within the mapped environment in order to reach a target. In contrast with route maps, survey maps are more flexible and effective, as they offer the choice of alternative paths to connect distant places, for example in the creation of shortcuts.

In order to predict how an individual will be able to perform a spatial task, it is therefore interesting to discover if, when needed, they are able to organise information in a survey representation.

From an experimental point of view it is possible to investigate survey knowledge by using typical spatial tasks, such as sketch maps, pointing and wayfinding tasks.

Sketch map tasks are considered effective for externalising survey maps, as they require the production of an external representation based on a birds-eye perspective (Billinghurst & Weghorst, 1994). Some difficulties arise in that sketching maps requires drawing abilities and it is difficult to interpret the results. Another drawback of sketch map tests depends on how difficult it is to quantitatively evaluate the drawings. A distinction has to be made between distortions that result from limited knowledge and ones that depend on difficulties in producing externalisation (Foreman & Gillet, 1997). To avoid these problems, another distinctive feature of survey maps, hierarchical organisation, can be used. Different studies have shown that knowledge about large-scale environments is represented in terms of macro-regions, defined by anchor points (Golledge, 1987), or “centroids”, reciprocally linked by spatial relationships, containing in turn other connected micro-regions, according to a part-whole relation. Thus, the primary elements of the representations are portions of space limited by visual barriers and gateways (Chown et al., 1995); examples include walls and doors of buildings or hills and paths in a valley. In a building, a significant part of the hierarchy is a cluster of rooms connected with corridors that give access to the cluster itself. Aggregates of clusters constitute the layout of the building. Accordingly we have decided to consider hierarchies as a main factor to be reflected in sketch maps.

Pointing is another task extensively used in spatial knowledge evaluation. Participants are generally asked to indicate, by lines in the air, the position of an unseen distant target point, usually from different vantage positions. According to several authors only external pointing trials are able to highlight survey knowledge (Carassa et al., 2002, Gaunet et al., 2001). This kind of pointing requires an explorer moving along a route pointing towards a target, that is not located along the travelled route. Pointing performance appears to become easier through repeated exploration sessions; corroborating the hypothesis that performance depends on the gradual creation of survey knowledge (Ruddle, Payne, & Jones, 1997). Even though pointing in virtual reality with a restricted visual angle might be considered a sound survey task, the latter could greatly jeopardise the ability of assessing directions (see for example Montello, Richardson, Hegarty, & Provenza, 1999).

Finally, a wayfinding task can be used to investigate the capacity of organising knowledge in a survey map, under specific conditions. For example, if the individual is asked to find the shortest route to a target point, not previously travelled, a wayfinding task constitutes the most ecologically valid spatial task. Not all wayfinding tasks necessarily require a survey ability; the easier they are the more likely it is for them to be correctly executed by simply resorting to a type of route representation. In wayfinding behaviour surveys representation allows two fundamental spatial activities; on one hand subjects can check their position within the entire environment during navigation (Chen and Stanney, 1999, Darken, 1995, Heth et al., 1997, Thorndyke and Goldin, 1983); on the other hand it allows planning in advance new paths, (shortcuts included) connecting distant landmarks especially if they are not visible at the same time (Carassa & Geminiani, 2002). Several studies showed that a fairly good knowledge of the environment is necessary in order to get a good survey representation which is normally obtained after a long stage of familiarisation or, within the experimental field, through an intensive exploration of the whole environment.

In conclusion, all these standard survey representation tests appear to be quite difficult to achieve at experimental level, even in virtual reality environments.

For these reasons we chose to introduce a new task, to specifically investigate the individual capacity of organising spatial information in a survey map through quick exploration of a virtual environment. In particular, we have designed a task able to investigate survey representation: planning in advance. This method evaluates the capacity to plan in advance new shortcuts connecting distant landmarks (planning optimal paths). The experimental hypothesis is that the capacity of planning in advance optimal paths is related to classical survey tasks performance, i.e., wayfinding, pointing and sketch map tasks.

If this proves to be sound from the experimental viewpoint, the planning in advance task will thus become a quicker and more effective test to experimentally evaluate survey maps.