How do cyclists make their way? - A GPS-based revealed preference study in Copenhagen

It is the objective of the study to determine the extent to which human navigation is affected by perceptions of our immediate surroundings or by already established knowledge in terms of a cognitive map. The motivation is to contribute to the knowledge about human navigation and to inform planning with estimates of bicyclists' route preferences and willingness-to-pay' (in terms of transport distance vs. utility/disutility of route characteristics). The core method is choice modelling of observed route data. Thousand two hundred and sixty seven trips performed by 183 cyclists in Copenhagen (Denmark) were recorded by GPS. The trips were map-matched to a digital road and path network, which enabled the generation of choice sets: one for navigation as influenced by perception of immediate surroundings, comprising edges connected to network-nodes (hereafter called the edge dataset), and one for navigation based on a priori knowledge, comprising the trip itself and a number of alternative routes generated by a labelling algorithm (hereafter called the route dataset). The results document that choices based on characteristics of both the route and the edge data can be estimated and provide reasonable and significant parameter estimates regarding cyclists' preferences. Length was significant and negative, illustrating that cyclists - everything else kept equal - prefer to bike shorter distances. Preferences regarding characteristics of bike path, presence of traffic lights and road types show similar results to the two types of data; most importantly that routes with facilities, such as curbed tracks and segregated bikeways, were significantly preferred. The study concludes that cyclists' wayfinding can be modelled as choices based on both an edge dataset and a route dataset and, thus, may be influenced by both perceived information and a priori knowledge. We suggest that future analyses of movement and route preferences take both modes into account as actual movement may be based on a combination of the two and because assessment of the influence of the immediate, perceivable surroundings can provide information not to be considered in a wayfinding approach. In our case differences in preferences along the route is for example found, but this can be expanded in future studies to also include dynamic aspects such as weather and crowding.