1/11
Public Transportation in Rural Areas:
The Clustered Dial-a-Ride Problem
Fabian Feitsch November 16th, 2018
Attributions of third party images can be found on slide 12.
2/11
Public Transportation
2/11
Public Transportation
2/11
Public Transportation
2/11
Public Transportation
2/11
Public Transportation
So far, so good?
2/11
Public Transportation
So far, so good?
loooooooomoooooooon walking distance?
2/11
Public Transportation
So far, so good?
loooooooomoooooooon walking distance?
wheather?
2/11
Public Transportation
So far, so good?
loooooooomoooooooon walking distance?
wheather?
waiting time?
2/11
Public Transportation
So far, so good?
loooooooomoooooooon walking distance?
wheather?
waiting time?
Ñ Probably in the city, but not in villages!
2/11
Public Transportation
So far, so good? Ñ Probably in the city, but not in villages!
Ñ Doorstep Service in Rural Areas
3/11
The Dial-a-Ride Problem
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq.
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
start with 0 (driver’s pickup) 0
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
start with 0 (driver’s pickup) 0 enumerate pickups
1
2 3
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
start with 0 (driver’s pickup) 0 enumerate pickups
enumerate dest‘s in same order
1
2 3
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
start with 0 (driver’s pickup) 0 enumerate pickups
4
enumerate dest‘s in same order
1
2 3
5
6
7
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
start with 0 (driver’s pickup) 0 enumerate pickups
4
enumerate dest‘s in same order
1
2 3
5
6
7
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
start with 0 (driver’s pickup) 0 enumerate pickups
4
enumerate dest‘s in same order
1
2 3
5
6
7 S :“number of seats in the vehicle
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
start with 0 (driver’s pickup) 0 enumerate pickups
4
enumerate dest‘s in same order
1
2 3
5
6
7 S :“number of seats in the vehicle
for this presentation: S “ 8
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
start with 0 (driver’s pickup) 0 enumerate pickups
4
enumerate dest‘s in same order
1
2 3
5
6
7 S :“number of seats in the vehicle
Objective: Feasible tour minimizing the sum of total distances.
for this presentation: S “ 8
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
start with 0 (driver’s pickup) 0 enumerate pickups
4
enumerate dest‘s in same order
1
2 3
5
6
7 S :“number of seats in the vehicle
Objective: Feasible tour minimizing the sum of total distances.
for this presentation: S “ 8
3/11
The Dial-a-Ride Problem
A Dial-a-Ride instance is a triple I “ pn, rdi,js, Sq. n :“ number of riders
Number of persons m “ n ` 1 rdi,js :“ distance matrix
start with 0 (driver’s pickup) 0 enumerate pickups
4
enumerate dest‘s in same order
1
2 3
5
6
7 S :“number of seats in the vehicle
Objective: Feasible tour minimizing the sum of total distances.
for this presentation: S “ 8
4/11
An Exact Algorithm
4/11
An Exact Algorithm
There is an exact algorithm by Psaraftis, 1980.
4/11
An Exact Algorithm
There is an exact algorithm by Psaraftis, 1980.
It works similar to the Held-Karp-algorithm.
4/11
An Exact Algorithm
There is an exact algorithm by Psaraftis, 1980.
It works similar to the Held-Karp-algorithm.
Running Time: O˚p3n´1q.
4/11
An Exact Algorithm
There is an exact algorithm by Psaraftis, 1980.
It works similar to the Held-Karp-algorithm.
Running Time: O˚p3n´1q.
Can be generalized to solve partial instances:
4/11
An Exact Algorithm
There is an exact algorithm by Psaraftis, 1980.
It works similar to the Held-Karp-algorithm.
Running Time: O˚p3n´1q.
Can be generalized to solve partial instances:
4/11
An Exact Algorithm
There is an exact algorithm by Psaraftis, 1980.
It works similar to the Held-Karp-algorithm.
Running Time: O˚p3n´1q.
Can be generalized to solve partial instances:
Find best tour such that a) girl is delivered
b) waiting customer is fetched c) boy is still on board.
5/11
Back to Rural Areas . . .
Hogsmeade
Springfield Minas Tirith
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Ñ Locations are inside clusters.
Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Ñ Locations are inside clusters.
Ñ Bypasses do not exist.
Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Ñ Locations are inside clusters.
Ñ Bypasses do not exist.
Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Ñ All riders head in the same direction.
Ñ Locations are inside clusters.
Ñ Bypasses do not exist.
Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Seems to be simpler than the Dial-a-Ride Problem . . . Ñ All riders head in the same direction.
Ñ Locations are inside clusters.
Ñ Bypasses do not exist.
Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Seems to be simpler than the Dial-a-Ride Problem . . . Ñ All riders head in the same direction.
Ñ Locations are inside clusters.
Ñ Bypasses do not exist.
Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Seems to be simpler than the Dial-a-Ride Problem . . . Ñ All riders head in the same direction.
Ñ Locations are inside clusters.
Ñ Bypasses do not exist.
Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Seems to be simpler than the Dial-a-Ride Problem . . . Ñ All riders head in the same direction.
Ñ Locations are inside clusters.
Ñ Bypasses do not exist.
Ñ ÝÑ
T ˚-algorithm Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Seems to be simpler than the Dial-a-Ride Problem . . . Ñ All riders head in the same direction.
Ñ Locations are inside clusters.
Ñ Bypasses do not exist.
Goal:
Ñ ÝÑ
T ˚-algorithm Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Seems to be simpler than the Dial-a-Ride Problem . . . Ñ All riders head in the same direction.
Ñ Locations are inside clusters.
Ñ Bypasses do not exist.
Goal:
Classify instances whose optimal tour is unidirectional.
Ñ ÝÑ
T ˚-algorithm Assumptions:
5/11
Back to Rural Areas . . .
A rural Dial-a-Ride instance typically looks like this:
Hogsmeade
Springfield Minas Tirith
Seems to be simpler than the Dial-a-Ride Problem . . . Ñ All riders head in the same direction.
Ñ Locations are inside clusters.
Ñ Bypasses do not exist.
Goal:
Classify instances whose optimal tour is unidirectional.
(without computing it)
Ñ ÝÑ
T ˚-algorithm Assumptions:
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
Let C1, . . . Cq be the clusters.
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
Let C1, . . . Cq be the clusters.
Let ΥpT , Ci q P R` such that @T : řq
i“1 ΥpT , Ciq “ cpT q
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
Let C1, . . . Cq be the clusters.
Let ΥpT , Ci q P R` such that @T : řq
i“1 ΥpT , Ciq “ cpT q
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
Let C1, . . . Cq be the clusters.
Let ΥpT , Ci q P R` such that @T : řq
i“1 ΥpT , Ciq “ cpT q Let ΦpCi q be a lower bound on ΥpT ˚, Ciq.
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
Let C1, . . . Cq be the clusters.
Let ΥpT , Ci q P R` such that @T : řq
i“1 ΥpT , Ciq “ cpT q Let ΦpCi q be a lower bound on ΥpT ˚, Ciq.
Theorem (= Classifier):
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
Let C1, . . . Cq be the clusters.
Let ΥpT , Ci q P R` such that @T : řq
i“1 ΥpT , Ciq “ cpT q Let ΦpCi q be a lower bound on ΥpT ˚, Ciq.
Theorem (= Classifier): @Ci : ΦpCiq “ ΥpÝÑ
T ˚, Ciq ñ T ˚ “ ÝÑ T ˚
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
Let C1, . . . Cq be the clusters.
Let ΥpT , Ci q P R` such that @T : řq
i“1 ΥpT , Ciq “ cpT q Let ΦpCi q be a lower bound on ΥpT ˚, Ciq.
Theorem (= Classifier): @Ci : ΦpCiq “ ΥpÝÑ
T ˚, Ciq ñ T ˚ “ ÝÑ T ˚ Proof. Via exchange argument.
6/11
A Classifier
Hogsmeade
Springfield Minas Tirith
Idea: Distribute the costs of a tour to the clusters.
Let C1, . . . Cq be the clusters.
Let ΥpT , Ci q P R` such that @T : řq
i“1 ΥpT , Ciq “ cpT q Let ΦpCi q be a lower bound on ΥpT ˚, Ciq.
Theorem (= Classifier): @Ci : ΦpCiq “ ΥpÝÑ
T ˚, Ciq ñ T ˚ “ ÝÑ T ˚ Proof. Via exchange argument.
TODO!
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
7
2 8
3 4 9
10 5
m “ 6
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted.
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted.
weight: 2
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted.
weight: 3
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted.
weight: 3
Ñ Count atomic journeys!
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i.
pickup cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i.
α “ 3
pickup cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i. β :“ #leftbound persons with dr ě i.
pickup cluster of r
dropoff cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i. β :“ #leftbound persons with dr ě i.
β “ 1
pickup cluster of r
dropoff cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i. β :“ #leftbound persons with dr ě i. γ :“ #left-entering persons with pr ě i.
pickup cluster of r
dropoff cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i. β :“ #leftbound persons with dr ě i. γ :“ #left-entering persons with pr ě i. δ :“ #right-entering persons with dr ď i.
pickup cluster of r
dropoff cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i. β :“ #leftbound persons with dr ě i. γ :“ #left-entering persons with pr ě i. δ :“ #right-entering persons with dr ď i.
ΥpT , Ciq “ inpCiq ` αCiCi`1 ` βCiCi´1 ` γCi´1Ci ` δCi`1Ci
See thesis
for proof of cpTq “ ř
ΥpT, Ci q.
pickup cluster of r
dropoff cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i. β :“ #leftbound persons with dr ě i. γ :“ #left-entering persons with pr ě i. δ :“ #right-entering persons with dr ď i.
ΥpT , Ciq “ inpCiq ` αCiCi`1 ` βCiCi´1 ` γCi´1Ci ` δCi`1Ci
See thesis
for proof of cpTq “ ř
ΥpT, Ci q.
pickup cluster of r
dropoff cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i. β :“ #leftbound persons with dr ě i. γ :“ #left-entering persons with pr ě i. δ :“ #right-entering persons with dr ď i.
ΥpT , Ciq “ inpCiq ` αCiCi`1 ` βCiCi´1 ` γCi´1Ci ` δCi`1Ci
See thesis
for proof of cpTq “ ř
ΥpT, Ci q.
pickup cluster of r
dropoff cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i. β :“ #leftbound persons with dr ě i. γ :“ #left-entering persons with pr ě i. δ :“ #right-entering persons with dr ď i.
ΥpT , Ciq “ inpCiq ` αCiCi`1 ` βCiCi´1 ` γCi´1Ci ` δCi`1Ci
See thesis
for proof of cpTq “ ř
ΥpT, Ci q.
pickup cluster of r
dropoff cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i. β :“ #leftbound persons with dr ě i. γ :“ #left-entering persons with pr ě i. δ :“ #right-entering persons with dr ď i.
ΥpT , Ciq “ inpCiq ` αCiCi`1 ` βCiCi´1 ` γCi´1Ci ` δCi`1Ci
See thesis
for proof of cpTq “ ř
ΥpT, Ci q.
pickup cluster of r
dropoff cluster of r
7/11
Distribute Costs to Clusters
Hogsmeade
Springfield Minas Tirith
Assign the parts of a tour to clusters.
7
2 8
3 4 9
10 5
m “ 6
Obs.: Edges of a tour are weighted. Ñ Count atomic journeys!
Every cluster Ci has four counters:
α :“ #rightbound persons with pr ď i. β :“ #leftbound persons with dr ě i. γ :“ #left-entering persons with pr ě i. δ :“ #right-entering persons with dr ď i.
ΥpT , Ciq “ inpCiq ` αCiCi`1 ` βCiCi´1 ` γCi´1Ci ` δCi`1Ci
See thesis
for proof of cpTq “ ř
ΥpT, Ci q.
Todo: ΦpCiq ď ΥpT ˚, Ciq
pickup cluster of r
dropoff cluster of r
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6 3
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6 3
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6 3 “
t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities?
3 “
t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities? S “ rt4u, t8us S “ rt4, 8us
3 “
t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities? S “ rt4u, t8us S “ rt4, 8us Additionally: List of Portals P.
3 “
t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities? S “ rt4u, t8us S “ rt4, 8us Additionally: List of Portals P.
“rl, ls, rl, rs‰
3 “
t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities? S “ rt4u, t8us S “ rt4, 8us Additionally: List of Portals P.
“rl, ls, rl, rs‰
Given S and P the lower bound can be estimated.
3 “
t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities? S “ rt4u, t8us S “ rt4, 8us Additionally: List of Portals P.
“rl, ls, rl, rs‰
Given S and P the lower bound can be estimated.
Solve internal tours.
3 “
t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities? S “ rt4u, t8us S “ rt4, 8us Additionally: List of Portals P.
“rl, ls, rl, rs‰
Given S and P the lower bound can be estimated.
Solve internal tours.
3
Compute lower bounds for α, β, γ and δ.
“t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities? S “ rt4u, t8us S “ rt4, 8us Additionally: List of Portals P.
“rl, ls, rl, rs‰
Given S and P the lower bound can be estimated.
Solve internal tours.
3
Compute lower bounds for α, β, γ and δ.
Add costs up and obtain lower bound ΦS,PpCiq.
“t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities? S “ rt4u, t8us S “ rt4, 8us Additionally: List of Portals P.
“rl, ls, rl, rs‰
Given S and P the lower bound can be estimated.
Solve internal tours.
3
Compute lower bounds for α, β, γ and δ.
Add costs up and obtain lower bound ΦS,PpCiq. ñ min ΦpCi qS,P “ ΦpCiq ď ΥpT ˚, Ciq
“t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities? S “ rt4u, t8us S “ rt4, 8us Additionally: List of Portals P.
“rl, ls, rl, rs‰
Given S and P the lower bound can be estimated.
Solve internal tours.
3
Compute lower bounds for α, β, γ and δ.
Add costs up and obtain lower bound ΦS,PpCiq. ñ min ΦpCi qS,P “ ΦpCiq ď ΥpT ˚, Ciq
|Ci | “ 6:
5 227 236 choices
kk kik hk
kk kj
“t8u, t4u‰
8/11
Lower Bound on Υ p T
˚, C
iq (Sketch)
Idea: Any T induces an ordered partition on every cluster.
Hogsmeade
Springfield Minas Tirith
7
2 8
4 9
10 5
m “ 6
Other Possibilities? S “ rt4u, t8us S “ rt4, 8us Additionally: List of Portals P.
“rl, ls, rl, rs‰
Given S and P the lower bound can be estimated.
Solve internal tours.
3
Compute lower bounds for α, β, γ and δ.
Add costs up and obtain lower bound ΦS,PpCiq. ñ min ΦpCi qS,P “ ΦpCiq ď ΥpT ˚, Ciq
|Ci | “ 6:
5 227 236 choices
kk kik hk
kk kj
Practical
Limit!
“t8u, t4u‰
9/11
Evaluation
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
Exact: 120 s ÝÑ
T ˚-Algorithm: 3 ms Classifier: 4 s for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
Exact: 120 s ÝÑ
T ˚-Algorithm: 3 ms Classifier: 4 s Classifier’s Accuracy:
for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
Exact: 120 s ÝÑ
T ˚-Algorithm: 3 ms Classifier: 4 s
Ratio T ˚ “ ÝÑ T ˚ Classifier’s Accuracy:
for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
Exact: 120 s ÝÑ
T ˚-Algorithm: 3 ms Classifier: 4 s
Ratio T ˚ “ ÝÑ T ˚ Clusters close together („ 6km):
Classifier’s Accuracy:
59 % for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
Exact: 120 s ÝÑ
T ˚-Algorithm: 3 ms Classifier: 4 s
Ratio T ˚ “ ÝÑ T ˚ Clusters close together („ 6km):
far apart (ě 16 km):
Classifier’s Accuracy:
59 % 100 % for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
Exact: 120 s ÝÑ
T ˚-Algorithm: 3 ms Classifier: 4 s
Ratio T ˚ “ ÝÑ T ˚ Clusters close together („ 6km):
far apart (ě 16 km):
Recall Classifier’s Accuracy:
59 % 100 % for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
Exact: 120 s ÝÑ
T ˚-Algorithm: 3 ms Classifier: 4 s
Ratio T ˚ “ ÝÑ T ˚ Clusters close together („ 6km):
far apart (ě 16 km):
Recall Classifier’s Accuracy:
59 % 0.4 100 % for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
Exact: 120 s ÝÑ
T ˚-Algorithm: 3 ms Classifier: 4 s
Ratio T ˚ “ ÝÑ T ˚ Clusters close together („ 6km):
far apart (ě 16 km):
Recall Classifier’s Accuracy:
0.4 0.9 59 %
100 % for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
Exact: 120 s ÝÑ
T ˚-Algorithm: 3 ms Classifier: 4 s
Ratio T ˚ “ ÝÑ T ˚ Clusters close together („ 6km):
far apart (ě 16 km):
Recall Classifier’s Accuracy:
0.4 0.9 ÝÑ
T ˚-Algorithm as Heuristic:
59 % 100 % for n “ 12
9/11
Evaluation
Ñ First artificial instances, then realistic instances.
Runtimes:
Exact: 120 s ÝÑ
T ˚-Algorithm: 3 ms Classifier: 4 s
Ratio T ˚ “ ÝÑ T ˚ Clusters close together („ 6km):
far apart (ě 16 km):
Recall Classifier’s Accuracy:
0.4 0.9 ÝÑ
T ˚-Algorithm as Heuristic:
Approximation Quality (empiric): ď 1.1
59 % 100 % for n “ 12
10/11
Topology of Street Networks
10/11
Topology of Street Networks
Street Networks often do not meet the assumptions.
10/11
Topology of Street Networks
Street Networks often do not meet the assumptions.
Example #1:
Rural Instance
10/11
Topology of Street Networks
Street Networks often do not meet the assumptions.
Example #1:
Rural Instance
10/11
Topology of Street Networks
Street Networks often do not meet the assumptions.
Example #1:
Rural Instance
T ˚ bypasses a cluster!
10/11
Topology of Street Networks
Street Networks often do not meet the assumptions.
Example #1:
Rural Instance
T ˚ bypasses a cluster!
Yet, no false positive.
10/11
Topology of Street Networks
Street Networks often do not meet the assumptions.
Example #1:
Rural Instance
T ˚ bypasses a cluster!
Yet, no false positive.
ñ Classifier is robust to some extent.
10/11
Topology of Street Networks
Street Networks often do not meet the assumptions.
Example #2:
Regional Instance
10/11
Topology of Street Networks
Street Networks often do not meet the assumptions.
Example #2:
Regional Instance
Really hard scenario . . .
10/11
Topology of Street Networks
Street Networks often do not meet the assumptions.
Example #2:
Regional Instance
Really hard scenario . . .
10/11
Topology of Street Networks
Street Networks often do not meet the assumptions.
Example #2:
Regional Instance
Really hard scenario . . . False positives are to be expected in this case.
11/11
Conclusion
11/11
Conclusion
The Exact Algorithm considers unsensible tours.
11/11
Conclusion
The Exact Algorithm considers unsensible tours.
11/11
Conclusion
The Exact Algorithm considers unsensible tours.
11/11
Conclusion
The Exact Algorithm considers unsensible tours.
Intuition yields the ÝÑ
T ˚-algorithm.
11/11
Conclusion
The Exact Algorithm considers unsensible tours.
Intuition yields the ÝÑ
T ˚-algorithm.
A classifier decides if the ÝÑ
T ˚-algorithm can be used.
11/11
Conclusion
The Exact Algorithm considers unsensible tours.
Intuition yields the ÝÑ
T ˚-algorithm.
A classifier decides if the ÝÑ
T ˚-algorithm can be used.
If yes, only a fraction of time is needed to get T ˚.
11/11
Conclusion
The Exact Algorithm considers unsensible tours.
Intuition yields the ÝÑ
T ˚-algorithm.
If no, virtually no time is wasted.
A classifier decides if the ÝÑ
T ˚-algorithm can be used.
If yes, only a fraction of time is needed to get T ˚.
11/11
Conclusion
The Exact Algorithm considers unsensible tours.
Intuition yields the ÝÑ
T ˚-algorithm.
If no, virtually no time is wasted.
A classifier decides if the ÝÑ
T ˚-algorithm can be used.
If yes, only a fraction of time is needed to get T ˚.
No false-positives: Optimal route is guaranteed.
11/11
Conclusion
The Exact Algorithm considers unsensible tours.
Intuition yields the ÝÑ
T ˚-algorithm.
If no, virtually no time is wasted.
A classifier decides if the ÝÑ
T ˚-algorithm can be used.
If yes, only a fraction of time is needed to get T ˚.
No false-positives: Optimal route is guaranteed.
11/11
Conclusion
The Exact Algorithm considers unsensible tours.
Intuition yields the ÝÑ
T ˚-algorithm.
If no, virtually no time is wasted.
A classifier decides if the ÝÑ
T ˚-algorithm can be used.
If yes, only a fraction of time is needed to get T ˚.
No false-positives: Optimal route is guaranteed.
12/11
Attributions
The above icons are made by Freepik from flaticon.com
Ð CC 3.0 BY by SimpleIcon from flaticon.com (c) Map Images from OpenStreetMap (osm.org)
13/11 The following slides were abandoned at some point
and not officially shown at the presentation. They may contain errors or are incomplete. Maybe they help you nonetheless.
14/11
The Objective Function
0
m 1
2 3
5
6
7
14/11
The Objective Function
A tour T is a permutation of r0, 2m ´ 1s.
0
m 1
2 3
5
6
7
14/11
The Objective Function
A tour T is a permutation of r0, 2m ´ 1s.
0
m 1
2 3
5
6
7
T “ r0, 3, 1, 5, 7, 2, 6, 4s
14/11
The Objective Function
A tour T is a permutation of r0, 2m ´ 1s.
0
m 1
2 3
5
6
7 T feasible ô
T “ r0, 3, 1, 5, 7, 2, 6, 4s
14/11
The Objective Function
A tour T is a permutation of r0, 2m ´ 1s.
0
m 1
2 3
5
6
7 T feasible ô T r1s “ 0 & T r2ms “ m
T “ r0, 3, 1, 5, 7, 2, 6, 4s
14/11
The Objective Function
A tour T is a permutation of r0, 2m ´ 1s.
0
m 1
2 3
5
6
7 T feasible ô T r1s “ 0 & T r2ms “ m
& precedences obeyed
T “ r0, 3, 1, 5, 7, 2, 6, 4s
14/11
The Objective Function
A tour T is a permutation of r0, 2m ´ 1s.
0
m 1
2 3
5
6
7 T feasible ô T r1s “ 0 & T r2ms “ m
& precedences obeyed
& S not violated
T “ r0, 3, 1, 5, 7, 2, 6, 4s
14/11
The Objective Function
A tour T is a permutation of r0, 2m ´ 1s.
0
m 1
2 3
5
6
7 T feasible ô T r1s “ 0 & T r2ms “ m
& precedences obeyed
& S not violated
S ě 3 T “ r0, 3, 1, 5, 7, 2, 6, 4s
14/11
The Objective Function
A tour T is a permutation of r0, 2m ´ 1s.
0
m 1
2 3
5
6
7 T feasible ô T r1s “ 0 & T r2ms “ m
& precedences obeyed
& S not violated
S ě 3
T minfeasible
ř2m
i“2 kpi ´ 1q ¨ d
”
T ri ´ 1s, T ris ı
Objective: T “ r0, 3, 1, 5, 7, 2, 6, 4s
14/11
The Objective Function
A tour T is a permutation of r0, 2m ´ 1s.
0
m 1
2 3
5
6
7 T feasible ô T r1s “ 0 & T r2ms “ m
& precedences obeyed
& S not violated
S ě 3
kpjq is the number of persons after step j of T .
T minfeasible
ř2m
i“2 kpi ´ 1q ¨ d
”
T ri ´ 1s, T ris ı
Objective: T “ r0, 3, 1, 5, 7, 2, 6, 4s
15/11
An Exact Algorithm
15/11
An Exact Algorithm
1
2 3
4
5
15/11
An Exact Algorithm
1
2 3
4 Find a tour with 6 steps: 5
15/11
An Exact Algorithm
1
2 3
4 Find a tour with 6 steps: 5
1 2 3 4 5
15/11
An Exact Algorithm
1
2 3
4 Find a tour with 6 steps: 5
1 2 3 4 5
0
cost: 0
15/11
An Exact Algorithm
1
2 3
4 Find a tour with 6 steps: 5
1 2 3 4 5
0
cost: 0
1
cost: 9
2
cost: 20
15/11
An Exact Algorithm
1
2 3
4 Find a tour with 6 steps: 5
1 2 3 4 5
0
cost: 0
1
cost: 9
2
cost: 20
4
cost: 41
2
cost: 49
15/11
An Exact Algorithm
1
2 3
4 Find a tour with 6 steps: 5
1 2 3 4 5
0
cost: 0
1
cost: 9
2
cost: 20
4
cost: 41
2
cost: 49
1
cost: 60
5
cost: 49
15/11
An Exact Algorithm
1
2 3
4 Find a tour with 6 steps: 5
1 2 3 4 5
0
cost: 0
1
cost: 9
2
cost: 20
4
cost: 41
2
cost: 49
1
cost: 60
5
cost: 49
2
cost: 57
15/11
An Exact Algorithm
1
2 3
4 Find a tour with 6 steps: 5
1 2 3 4 5
0
cost: 0
1
cost: 9
2
cost: 20
4
cost: 41
2
cost: 49
1
cost: 60
5
cost: 49
2
cost: 57
4
cost: 98
5
cost: 92