[sotm10] geovelo, a route planner for bicycle
TRANSCRIPT
Geovelo, a route planner for bicycle
Geovelo, a route planner for bicycle
G. Sauvanet, E. Neron, H. Baptiste
Laboratoire d’InformatiqueUniversite Francois Rabelais Tours
Polytech’Tours - Departement Informatique64, Avenue Jean Portalis
37200 ToursFRANCE
11 juillet 2010
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 1
Geovelo, a route planner for bicycle
Outline
1 Presentation
2 The Bi-Objective Shortest Path problemModeling a road networkMono-objective problemBi-objective problem
3 Conclusion
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 2
Geovelo, a route planner for bicycle
Presentation
Context
PhD CIFRE LI/Association � Autour du Train � (March 2008)
� Autour du Train � : promoting alternative modes of travellike bicycle
Today : no route planner really adapted to the bicycle inFrance
Geovelo
available on 3 cities (Paris, Nantes, Tours)
Database : Postgresql/Postgis, OSM Data, Osmosis
Website : Cloudmade maps, Openlayers
Multi-objective routing engine
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 3
Geovelo, a route planner for bicycle
Presentation
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 4
Geovelo, a route planner for bicycle
Presentation
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 4
Geovelo, a route planner for bicycle
Outline
1 Presentation
2 The Bi-Objective Shortest Path problemModeling a road networkMono-objective problemBi-objective problem
3 Conclusion
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 5
Geovelo, a route planner for bicycle
Modeling a road network
Let G = (V ,A) be agraph with :
V the set ofnodes,
A the set of arcs,
the cost functiondistance : A→R+
insecurity : A→R+
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 6
Geovelo, a route planner for bicycle
Modeling a road network
Let G = (V ,A) be agraph with :
V the set ofnodes,
A the set of arcs,
the cost functiondistance : A→R+
insecurity : A→R+
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 6
Geovelo, a route planner for bicycle
Modeling a road network
Let G = (V ,A) be agraph with :
V the set ofnodes,
A the set of arcs,
the cost functiondistance : A→R+
insecurity : A→R+
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 6
Geovelo, a route planner for bicycle
Modeling a road network
Let G = (V ,A) be agraph with :
V the set ofnodes,
A the set of arcs,
the cost functiondistance : A→R+
insecurity : A→R+
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 6
Geovelo, a route planner for bicycle
Modeling a road network
Let G = (V ,A) be agraph with :
V the set ofnodes,
A the set of arcs,
the cost functiondistance : A→R+
insecurity : A→R+
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 6
Geovelo, a route planner for bicycle
Presentation of the mono-objective problem
Mono-objective problem
Let us define :
a graph G ,
a start node s and a target node t.
Goal :
compute shortest path p linking s to t iemin
∑a∈p distance(a)
Experiments on a graph with 136 199 nodes and 345 267 arcs :
Dijkstra algorithm
average of 0.1 seconde on 100 routes.
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 7
Geovelo, a route planner for bicycle
Presentation of the mono-objective problem
Mono-objective problem
Let us define :
a graph G ,
a start node s and a target node t.
Goal :
compute shortest path p linking s to t iemin
∑a∈p distance(a)
Experiments on a graph with 136 199 nodes and 345 267 arcs :
Dijkstra algorithm
average of 0.1 seconde on 100 routes.
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 7
Geovelo, a route planner for bicycle
Presentation of the mono-objective problem
Mono-objective problem
Let us define :
a graph G ,
a start node s and a target node t.
Goal :
compute shortest path p linking s to t iemin
∑a∈p distance(a)
Experiments on a graph with 136 199 nodes and 345 267 arcs :
Dijkstra algorithm
average of 0.1 seconde on 100 routes.
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 7
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
Bi-objective problem
2 objectives : minimize distance and insecurity
2 conflicting objectives⇒ shortest route is often adapted for the car, so it isdangerous for bicycle
Insecurity of a path = sum of the insecurity of each arc of thepath
Insecurity = distance . insecurity coefficient (inspired fromdangerous material transportation)
insecurity coefficient depends on the nature of an arc : bicyclepath, bike lane, no facilities, etc.
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 8
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
Problem
Solving bi-objective shortest path problem is hard
Not an unique path, but a set of efficient paths
Example on Paris : > 500 efficient paths and computationtime > 1 minute
Basic solution
Transform bi-objective problem in mono-objective problem
Linear combination of distance and insecurity : new cost(a) =α.distance(a) + (1− α).insecurity(a)
⇒ fast and used by many (all ?) route planner, but not really good
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 9
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 10
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 10
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 11
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 12
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 13
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 14
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 15
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 16
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
new cost(a) = α1.distance(a) + (1− α1).insecurity(a)
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 16
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
new cost(a) = α1.distance(a) + (1− α1).insecurity(a)
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 16
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
new cost(a) = α2.distance(a) + (1− α2).insecurity(a)
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 16
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
new cost(a) = α2.distance(a) + (1− α2).insecurity(a)
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 16
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
new cost(a) = α3.distance(a) + (1− α3).insecurity(a)
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 16
Geovelo, a route planner for bicycle
Presentation of the bi-objective problem
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 16
Geovelo, a route planner for bicycle
Research
Our research
1 Compute all efficient paths : improvements of labelingalgorithms for bi-objective problem [ROADEF 2009, MOPGP2010]
2 Compute the best compromise path : [ISCO 2010, JMMA2010]
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 17
Geovelo, a route planner for bicycle
Outline
1 Presentation
2 The Bi-Objective Shortest Path problemModeling a road networkMono-objective problemBi-objective problem
3 Conclusion
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 18
Geovelo, a route planner for bicycle
Conclusion
Conclusion :
Prototype works (http://www.geovelo.fr).
Further work :
user feedback on OSM data
Test with more objectives (effort, tourist Interest...)
Mobile version
Let user set preferences of the routing engine
G. Sauvanet, E. Neron, H. Baptiste SotM 2010 - Girona 19