Modelling Human Spatio-Temporal Behaviour: A Challenge for Location-Based Services

Posted: March 31st, 2007 | No Comments »

Mountain, D., Raper, J.F., 2001. Modelling human spatio-temporal behaviour: a challenge for location-based services. In: Proc. of the 6th Internat. Conference on GeoComputation. University of Queensland, Brisbane, Australia, 24-26 September 2001.

With the proliferation of location-aware devices large amount of individuals spatial and temporal record of where they have been and when they were there can be stored. This paper describes an number of algorithms for tailoring and analyze accumulated archives of spatio-temporal data generated by location-aware applications. The aim of this work is to produce automated approaches to summarize individual position histories. This helps to derive high level human behaviror from spatio-temporal data such as enclosing rectangles of typical movement and estimates of transportation mode.

Scientist researching the physics of space and time pioneered in the modelling of phenomena against a spatio-temportal framework. Minkowski for example visualised space and time as a light cone defining a boundary between past and future accessible locations.

Minkowski Cone-1
“All movement must be contained within this region in space-time since nothing can travel faster than the speed of light”. (Rapper 2000) .

Torsten Hägerstrand inspired from Minkowski to theorize “Time Geography“. In this framework, a individual movement can be described as a space-time path whose gradient represents velocity across the 2-dimensional surface. A vertical line represents no movement and increasingly horizontally sloped lines show faster velocities.

Location Trend Extractor
Spatial and temporal views

Summarizing spatio-temporal behaviors starts with breaking down the large history of points into discreet temporal sections, each section representing an episode. The key to define episode is finding breakpoints in the data. The main indicators of breackpoints are:
Temporal/spatial jumps, rapid changes in direction/speed, spatial envelopes, temporal envelopes, map display envelopes. The latest being extremely related to my interest in ways to understand and deliver the granularity of location information:

An application of an envelope is holding the bounding coordinates for map display for the real time delivery of an orientation map for a mobile user. The map scale needs to be context sensitive and this can be achieved by assessing the user’s speed and direction. A user travelling at very slow speed, possibly associated with ‘on foot’ or ‘low speed motor’ transport behaviour, will require a high resolution map, displaying minor roads and pedestrian routes, centred on their current location. A user travelling at high speed will require a coarser resolution map, centred ahead of their current location; either a direct extrapolation from their current direction or a more complex algorithm based on the transportation

Relation to my thesis: People are consumer of time and space. A goal of my current experiment is to develop a new data models and forms of analysis to find spatial and temporal trends and patterns based on explicit, user-generated traces (in contrast to GPS or GSM traces). Understanding how people and crowds explicitly consume space could lead to the emergence of types of granularity of location information, as described in this paper by “map display envelopes”.