Fastest biking route

[mathiaskopo]

Hello

I'am trying to calculate the fastest route for biking between two points, the fastest route should take into account that it is faster to bike downhill than uphill. I know that it is possible to do this on raster data with Spatial Analyst Path Distance tool but does anybody know if it is possible to it with a Network Analyst tool with a network?

/Mathias

[jsandhu]

Network Analyst minimizes some cost per edge that is present in an attribute to compute a shortest path. As long as you have a network with the appropriate edge cost, you can achieve your goal. So if you know the uphill/downhill information on each edge, then you can modify the travel time of that edge to make it slower or faster depending on uphill/downhill status of the edge. If this information is not available, then you will have to get a DEM (raster) of that area and then use it to calibrate your network edges based on that. So short answer, yes it is possible, long answer, you will have to work to calibrate your data (or purchase it from a vendor like NAVTEQ's Green Streets product).

Jay Sandhu

[mathiaskopo]

OK, thank you Jay. I only know how to assign a cost to an edge that will be the same going from A to B and B to A, and not like in this case different going from A to B (uphill) and B to A (downhill) but I guess it must be possible, I will try to look into that.

[jsandhu]

Network attributes are bi-directional. When you set up a network dataset and add an attribute, you then use the Evaluators to set up how each direction of the edge will get its value. You can open up the properties of the tutorial datasets and see how the length or time attributes are set up. And

You could add two fields to your street edge attribute table called FT_BikeTime and TF_BikeTime. Then calc these two fields to the appropriate values for each directions of travel. Later when you create a network dataset, you can add an attribute called BikeTime and it will auto-detect these two fields to get the values.

you can read more about attributes here:
http://help.arcgis.com/en/arcgisdesk...000000m000000/
and how to set them up here:
http://help.arcgis.com/en/arcgisdesk...000000n000000/

Jay Sandhu

[mathiaskopo]

Thank you again Jay your answers have been very helpful :-) By the way do you know Thomas Balstrøm from Denmark? He is one of the teachers from University of Aalborg where I’m studying Geoinformatics.

[jsandhu]

Glad to know that my answers have helped. And yes, I do know Prof Thomas Balstrøm very well for almost 25 years. He is an excellent spatial thinker. Glad to know you are studying GIS with him. Give my regards to him.

Jay Sandhu

[mathiaskopo]

Funny how small the world is. I will give your regards to him and thanks again for your help

/Mathias Kofoed Poulsen

[jsandhu]

Well, 25 years ago, the GIS world was very very small!

By the way, it would be good if you post a small summary here on the results of your calibrating your bike network. That is, I am quite interested in this work as the answer of "fastest path" cannot be simply computed by having an uphill/downhill cost. A road edge between two junctions can have many ups/downs and slope changes and it is hard to capture the variability by a single number. I presented a paper on this topic at the 2011 Applied Geography conference. You can see the abstract on page 43 here: http://applied.geog.kent.edu/AGC2011/index.html

Jay Sandhu

[mathiaskopo]

Sorry for the late response but I first saw your post now. Yes it is not an easy task to find the “fastest path” when considering hills. We would also like to estimate the time it takes to bike the path, and the problem here is that the gravitational force is constant, which means that the speed of a bike going down a hill will be a lot faster at the bottom than at the top of the hill. But we will post a summary of our final solution in the beginning of June.

Thanks for the link to your abstract, is your paper “Calibrating a Transportation Network with Terrain Information for Routing Applications” published and is it possible to read it?

[jsandhu]

I do not have a paper but a power point. Send me an email jsandhu at esri
Jay Sandhu

[mathiaskopo]

A Newtonian mechanics approach could be used to account for the speed of a bicycle moving down a slope.

Speed of a bike when only considering the degree of the slope and the initial speed of bike is: v𝑡 =sinƟ a𝑔t +a𝑏t +v0

Where v𝑡 is the speed at the time t, Ɵ the degree of the hill, a𝑔 the gravitational acceleration on the Earth, a𝑏 the acceleration from the bicyclist and v0 the speed at the time t = 0.

The interesting part here is that the speed of the bicycle is dependent of the speed at a previous position on the slope. Therefore will a bicyclist after riding down a hill have a larger speed at the bottom than at the top of the hill, this complicates the estimation of the travel time since the resistance from the wind is dependent of the speed of the bike. This means that it will be necessary to calculate the influence from the hill based on many different initial speeds to get an accurate value. One solution could be just to use an average initial value as speed of the bike.

In either case it is necessary to develop an algorithm based on newton mechanics which then could be used with a DTM model to precalculate hills costs for the different edges in a network which then would be used as all other types of network costs applied to the edges.

An algorithm was not develop in the project we were writing on but it should be possible with an outset in v𝑡 =sinƟ a𝑔t +a𝑏t +v0 and an average speed of a bicyclist in flat terrain to make a estimate for the resistance from the terrain. This will not be 100% perfect but should give a useable approximation.