A forward Markov model for predicting bicycle speed
- PDF / 1,701,508 Bytes
- 23 Pages / 439.37 x 666.142 pts Page_size
- 87 Downloads / 171 Views
A forward Markov model for predicting bicycle speed Petter Arnesen1 · Olav Kåre Malmin1 · Erlend Dahl1
© Springer Science+Business Media, LLC, part of Springer Nature 2019
Abstract Speed prediction of different transport modes is important in applications such as route planning, transport modelling and energy calculations. In this paper we model bicycle speed as a function of slope and horizontal curvature. We developed two models, one with dependence between subsequent observations (a forward Markov model) and one without such a dependence (a generalised linear model). We show through prediction on out-ofsample data that the model including dependence between observations outperforms the model without. To estimate and evaluate our models we use a data set collected using a smart phone application. The data collected includes different sources of error, and therefore we introduce various filtering methods to make the data more appropriate for statistical analysis and model estimation. Keywords Bicycle speed modelling · GLM · GPS data · Markov model
Introduction When deciding how to move within the transport system, travelling time from your origin to destination is often a critical parameter to consider. Different (combinations) of transport modes would produce different travel times, and is naturally taken into consideration when deciding how and when to travel. To calculate and understand travel times within the transport network, modelling speed of different transport modes under exogenous variables such as road geometry, traffic, vehicle power and so on, is a much used approach and is of great importance in many applications, such as transport modelling and traffic planning. Modelling the speed of cars is a much studied problem, see for instance Hassan and Sarhan (2011) and many of the references therein. However, the increased focus on pedestrians and bicyclists requires more realistic models for these modes too (Romanillos et al. 2016). Bicycling is particularly sensitive to the geometry of the road. For instance, the speed going downhill will typically be much higher than going uphill, and the travelling time (and energy consumption) will change correspondingly. A notable exception to this are the electric bicycles (Xu et al. 2015) where one would expect the speed to be less dependent on such parameters, particularly the vertical slope. * Petter Arnesen [email protected] 1
Department of Mobility and Economics, SINTEF, Trondheim, Norway
13
Vol.:(0123456789)
Transportation
Time and energy usage are critical parameters to bicyclists. Unlike car usage where energy usage is a simple question of fuel cost, bicycle energy usage is a question of effort, stamina, health and indirectly motivation to complete a bicycle trip. This makes realistic modelling of speed important when evaluating possible new bicycle routes or working with the route choice problem in transport modelling (Cascetta and Papola 2001; Hawas 2004; Hood et al. 2011; Quattrone and Vitetta 2011; Vitetta 2016 and many of the refe
Data Loading...
