NTRODUCTION

On a university campus, we can expect that learning will take place in both indoor spaces and outdoor spaces. The first set of learning environments consists of enclosed spaces, such as classrooms, lecture theatres, laboratories, libraries, corridors, etc. The second set of environments that can support learning consists of open spaces, such as quadrangles, courtyards, plazas, pedestrian pathways, etc. As universities develop learning places for students, attention must be focussed on both indorr and outdoor places; design guides are plentiful for the design of indoor learning spaces but few are available for outdoor spaces (Jamieson 2003).

In contrast to indoor spaces, learning in outdoor spaces is a more casual activity, within which countless patterns of learning activities might emerge. For example, students may work in a group around a garden table, read a assigned texts in print or a mobile device while lying on the lawn or have a conversation about a lecture subject while sharing a bench. These learning activities may be interspersed among non-learning activities, such as revisiting a sports match. Therefore it becomes a challenge to differentiate those activities that are related to learning from the ones that are mainly recreational, especially as we cannot infer the activity through a nature of the space as we might for a classroom.. Consequently, it is challenging to recognise specific activities correlated to learning in outdoor spaces and to identify the combination of parameters that support this.

This paper presents preliminary results of research to develop an algorithmic model to represent how campus users interact with one another, as well with the surrounding environment, in order to examine the dynamics of learning activities in open spaces and thus develop design guidance for such spaces. The research hypothesis is that learning activities in outdoor spaces contribute to the life of the academic purpose of the campus and that this can then inform design decisions for such spaces to better support outdoor informal learning. The investigation process seeks to answer the following research questions:

1. Can we interpret a learning activity through patterns of activity observed by remote sensing?
2. Is it possible to make a correlation between learning activity and behaviour in outdoor spaces?
3. How do environmental parameters and spatial arrangements influence the dynamics of learning activities in outdoor spaces?

Comprehension about the importance of learning in outdoor spaces can lead to infrastructure that can assist the campus manager during decision-making, helping to predict what kind of change will occur in the campus as a whole. Such a model can then be used to simulate activity in new designs for spaces and postulate user activity in these spaces. The contribution to knowledge in this research do not come from the tools used or developed over the research process, but the strategy used to attempt to comprehend learning phenomena in outdoor space.

METHODOLOGICAL APPROACH

Three sites around the school of design were defined as locations for the research: the outdoor spaces located adjacent to the north, south and east façades of the building. These spaces have benches arranged in a variety of spatial configurations and a grass lawn that allows users to sit alone, or in small or large groups. These three sites allow the emergence of a wide variety of interaction patterns between users.

Researchers gathered, processed and analysed data from these places according to two different scales.  On one hand, we sought to comprehend focused encounters, namely, the ways in which people arrange themselves in relation to each other and space (Kendon, 1990), what kind of activity they are doing (Lawson, 2001) and what kind of object they are holding (Gibson, 1986). On the other hand, the research looked for information about the research field dynamic on a broader scale, seeking to discover how people flow and encounter one another, and how this leads to different types of activity.

The research has been developed in three main phases: data collection, analyses and development of remote sensing prototypes. In the first phase, the following methods were used to gather data:

Interviews: the researcher interviewed sixty students around the faculty building to identify if the studied sites were places where students were developing learning activities, what kind of objects they were using during the task, and to comprehend which factors led users to choose these spaces to conduct the activity;

Time-lapse video: a high definition camera with fisheye lens was set up to collect images in intervals of 10 seconds, from 08hs00 to 17hs00, for 10 days over a period of three weeks. The aim was to collect images and then process them both manually, looking for patterns, and then to develop an capability to process these using computer vision techniques.

RESULTS

Data collected during the fieldwork was processed. Interview answers were used to make a correlation between learning and the objects hold by students. Time-lapse video was analysed seeking to comprehend the places’ dynamics: intensity of people’s movement (such as walking) and the quantity of users using sitting spots over time. Images were also used as sources to collect sample images of people developing different activities, providing data for the last phase. These analyses led to the following preliminary conclusions:

1. From the interviews, we understand that students usually use these spaces for convenience: before classes, for a short period, not only to wait, but also to develop tasks that can be correlated to learning before moving to a lecture (e.g., reading a paper, writing some notes, or using the computer). Post analysis of the 27000 images collected in the time-lapse mode confirmed this dynamic of use;
2. Interviews revealed that 90% of the students that were reading documents in A4 format (papers or notebooks) and 70% of students using computers were considered to be engaged in a learning activity (these activities were the ones that had higher correlation with learning);
3. Unexpectedly, only 20% of students in groups were considered to be developing a learning activity.

These preliminary conclusions led to the development of three applications, using computer vision methods. The objective was to automate manual analysis that had been done frame-by-frame. Each tool had, respectively, the following purposes: (1) to capture intensity of movement, (2) to identify preferred sitting spots in the studied site and (3) to track one activity related to learning. As will be seen in the paper, the activity tracked was reading, because, as was previously explained, interviews indicated that students doing this activity in the studied site have higher probability of been in a learning task in the studied site.

DISCUSSION

From this analysis we posit that it is possible to establish a correlation between learning activities and patterns of behaviour in outdoor spaces. We note that the interviews were essential in this interpretation;  different places and contexts can lead to different behaviour or perception of what constitutes a learning activity. After accomplishing this “calibration”, it is possible to sense remotely activities that have greater correlation to learning according to the results obtained from interviews.  Nevertheless, objects play an important role in the process because the presence of an object assists in identifying activities correlated to learning in outdoor spaces. For instance, students reading A4 paper are more likely engaged in an activity in which the object is explicitly correlated to learning; as a result, the probability of tracking learning is higher.

Identifying where learning takes place in outdoor spaces depends on the comprehension of the dynamic of the site studied and in which context it is inserted. Before conducting any remote sensing data collection it is important to identify where learning activities happen, how they happen, ascertain if it is an explicit process or not and define a pattern of occurrence.  Remote sensing can then be used to collect data to be analysed, affording the comprehension of the place usage for a long-term period.

The research does not seek to discuss what is learning or and what is not, because this process can manifest in different ways, sometimes in a more formal and other times in a more informal activity. Future work will seek to add environmental parameters such as temperature, humidity, wind speed and luminosity to comprehend how activities correlated to learning fluctuate according to microclimate changes.