# Simple strategies versus optimal schedules in multi-agent patrolling

Akitoshi Kawamura, Makoto Soejima

Research output: Chapter in Book/Report/Conference proceedingConference contribution

5 Citations (Scopus)

### Abstract

Suppose that we want to patrol a fence (line segment) using k mobile agents with given speeds v1,..., vk so that every point on the fence is visited by an agent at least once in every unit time period. A simple strategy where the ith agent moves back and forth in a segment of length vi/2 patrols the length (v1 +· · ·+vk)/2, but it has been shown recently that this is not always optimal. Thus a natural question is to determine the smallest c such that a fence of length c(v1 + · · · + vk)/2 cannot be patrolled. We give an example showing c ≥ 4/3 (and conjecture that this is the best possible). We also consider a variant of this problem where we want to patrol a circle and the agents can move only clockwise. We can patrol a circle of perimeter rvr by a simple strategy where the r fastest agents move at the same speed, but it has been shown recently that this is not always optimal. We conjecture that this is not even a constant-approximation strategy. To tackle this conjecture, we relate it to what we call constant gap families. Using this relation, we give another example where the simple strategy is not optimal. We propose another variant where we want to patrol a single point under the constraint that for each i = 1,..., k, the time between two consecutive visits of agent i should be ai or longer. This problem can be reduced to the discretized version where the ai are integers and the goal is to visit the point at every integer time. It is easy to see that this discretized patrolling is impossible if 1/a1+· · ·+1/ak < 1, and that there is a simple strategy if 1/a1+· · ·+1/ak ≥ 2. Thus we are interested in the smallest c such that patrolling is always possible if 1/a1 + · · · + 1/ak ≥ c. We prove that α ≤ c < 1.546, where α = 1.264... (we conjecture that c = α). We also discuss the computational complexity of related problems.

Original language English Algorithms and Complexity - 9th International Conference, CIAC 2015, Proceedings Peter Widmayer, Vangelis Th. Paschos Springer Verlag 261-273 13 9783319181721 https://doi.org/10.1007/978-3-319-18173-8_19 Published - Jan 1 2015 9th International Conference on Algorithms and Complexity, CIAC 2015 - Paris, FranceDuration: May 20 2015 → May 22 2015

### Publication series

Name Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 9079 0302-9743 1611-3349

### Other

Other 9th International Conference on Algorithms and Complexity, CIAC 2015 France Paris 5/20/15 → 5/22/15

### Fingerprint

Fences
Optimal Strategy
Schedule
Mobile agents
Circle
Computational complexity
Clockwise
Integer
Mobile Agent
Perimeter
Line segment
Consecutive
Computational Complexity
Strategy
Unit
Approximation

### All Science Journal Classification (ASJC) codes

• Theoretical Computer Science
• Computer Science(all)

### Cite this

Kawamura, A., & Soejima, M. (2015). Simple strategies versus optimal schedules in multi-agent patrolling. In P. Widmayer, & V. T. Paschos (Eds.), Algorithms and Complexity - 9th International Conference, CIAC 2015, Proceedings (pp. 261-273). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9079). Springer Verlag. https://doi.org/10.1007/978-3-319-18173-8_19
Algorithms and Complexity - 9th International Conference, CIAC 2015, Proceedings. ed. / Peter Widmayer; Vangelis Th. Paschos. Springer Verlag, 2015. p. 261-273 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9079).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Kawamura, A & Soejima, M 2015, Simple strategies versus optimal schedules in multi-agent patrolling. in P Widmayer & VT Paschos (eds), Algorithms and Complexity - 9th International Conference, CIAC 2015, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9079, Springer Verlag, pp. 261-273, 9th International Conference on Algorithms and Complexity, CIAC 2015, Paris, France, 5/20/15. https://doi.org/10.1007/978-3-319-18173-8_19
Kawamura A, Soejima M. Simple strategies versus optimal schedules in multi-agent patrolling. In Widmayer P, Paschos VT, editors, Algorithms and Complexity - 9th International Conference, CIAC 2015, Proceedings. Springer Verlag. 2015. p. 261-273. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-18173-8_19
Kawamura, Akitoshi ; Soejima, Makoto. / Simple strategies versus optimal schedules in multi-agent patrolling. Algorithms and Complexity - 9th International Conference, CIAC 2015, Proceedings. editor / Peter Widmayer ; Vangelis Th. Paschos. Springer Verlag, 2015. pp. 261-273 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{3ca400462b3a4c06bdb164f0ea1fb46f,
title = "Simple strategies versus optimal schedules in multi-agent patrolling",
abstract = "Suppose that we want to patrol a fence (line segment) using k mobile agents with given speeds v1,..., vk so that every point on the fence is visited by an agent at least once in every unit time period. A simple strategy where the ith agent moves back and forth in a segment of length vi/2 patrols the length (v1 +· · ·+vk)/2, but it has been shown recently that this is not always optimal. Thus a natural question is to determine the smallest c such that a fence of length c(v1 + · · · + vk)/2 cannot be patrolled. We give an example showing c ≥ 4/3 (and conjecture that this is the best possible). We also consider a variant of this problem where we want to patrol a circle and the agents can move only clockwise. We can patrol a circle of perimeter rvr by a simple strategy where the r fastest agents move at the same speed, but it has been shown recently that this is not always optimal. We conjecture that this is not even a constant-approximation strategy. To tackle this conjecture, we relate it to what we call constant gap families. Using this relation, we give another example where the simple strategy is not optimal. We propose another variant where we want to patrol a single point under the constraint that for each i = 1,..., k, the time between two consecutive visits of agent i should be ai or longer. This problem can be reduced to the discretized version where the ai are integers and the goal is to visit the point at every integer time. It is easy to see that this discretized patrolling is impossible if 1/a1+· · ·+1/ak < 1, and that there is a simple strategy if 1/a1+· · ·+1/ak ≥ 2. Thus we are interested in the smallest c such that patrolling is always possible if 1/a1 + · · · + 1/ak ≥ c. We prove that α ≤ c < 1.546, where α = 1.264... (we conjecture that c = α). We also discuss the computational complexity of related problems.",
author = "Akitoshi Kawamura and Makoto Soejima",
year = "2015",
month = "1",
day = "1",
doi = "10.1007/978-3-319-18173-8_19",
language = "English",
isbn = "9783319181721",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "261--273",
editor = "Peter Widmayer and Paschos, {Vangelis Th.}",
booktitle = "Algorithms and Complexity - 9th International Conference, CIAC 2015, Proceedings",

}

TY - GEN

T1 - Simple strategies versus optimal schedules in multi-agent patrolling

AU - Kawamura, Akitoshi

AU - Soejima, Makoto

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Suppose that we want to patrol a fence (line segment) using k mobile agents with given speeds v1,..., vk so that every point on the fence is visited by an agent at least once in every unit time period. A simple strategy where the ith agent moves back and forth in a segment of length vi/2 patrols the length (v1 +· · ·+vk)/2, but it has been shown recently that this is not always optimal. Thus a natural question is to determine the smallest c such that a fence of length c(v1 + · · · + vk)/2 cannot be patrolled. We give an example showing c ≥ 4/3 (and conjecture that this is the best possible). We also consider a variant of this problem where we want to patrol a circle and the agents can move only clockwise. We can patrol a circle of perimeter rvr by a simple strategy where the r fastest agents move at the same speed, but it has been shown recently that this is not always optimal. We conjecture that this is not even a constant-approximation strategy. To tackle this conjecture, we relate it to what we call constant gap families. Using this relation, we give another example where the simple strategy is not optimal. We propose another variant where we want to patrol a single point under the constraint that for each i = 1,..., k, the time between two consecutive visits of agent i should be ai or longer. This problem can be reduced to the discretized version where the ai are integers and the goal is to visit the point at every integer time. It is easy to see that this discretized patrolling is impossible if 1/a1+· · ·+1/ak < 1, and that there is a simple strategy if 1/a1+· · ·+1/ak ≥ 2. Thus we are interested in the smallest c such that patrolling is always possible if 1/a1 + · · · + 1/ak ≥ c. We prove that α ≤ c < 1.546, where α = 1.264... (we conjecture that c = α). We also discuss the computational complexity of related problems.

AB - Suppose that we want to patrol a fence (line segment) using k mobile agents with given speeds v1,..., vk so that every point on the fence is visited by an agent at least once in every unit time period. A simple strategy where the ith agent moves back and forth in a segment of length vi/2 patrols the length (v1 +· · ·+vk)/2, but it has been shown recently that this is not always optimal. Thus a natural question is to determine the smallest c such that a fence of length c(v1 + · · · + vk)/2 cannot be patrolled. We give an example showing c ≥ 4/3 (and conjecture that this is the best possible). We also consider a variant of this problem where we want to patrol a circle and the agents can move only clockwise. We can patrol a circle of perimeter rvr by a simple strategy where the r fastest agents move at the same speed, but it has been shown recently that this is not always optimal. We conjecture that this is not even a constant-approximation strategy. To tackle this conjecture, we relate it to what we call constant gap families. Using this relation, we give another example where the simple strategy is not optimal. We propose another variant where we want to patrol a single point under the constraint that for each i = 1,..., k, the time between two consecutive visits of agent i should be ai or longer. This problem can be reduced to the discretized version where the ai are integers and the goal is to visit the point at every integer time. It is easy to see that this discretized patrolling is impossible if 1/a1+· · ·+1/ak < 1, and that there is a simple strategy if 1/a1+· · ·+1/ak ≥ 2. Thus we are interested in the smallest c such that patrolling is always possible if 1/a1 + · · · + 1/ak ≥ c. We prove that α ≤ c < 1.546, where α = 1.264... (we conjecture that c = α). We also discuss the computational complexity of related problems.

UR - http://www.scopus.com/inward/record.url?scp=84944725820&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84944725820&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-18173-8_19

DO - 10.1007/978-3-319-18173-8_19

M3 - Conference contribution

AN - SCOPUS:84944725820

SN - 9783319181721

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 261

EP - 273

BT - Algorithms and Complexity - 9th International Conference, CIAC 2015, Proceedings

A2 - Widmayer, Peter

A2 - Paschos, Vangelis Th.

PB - Springer Verlag

ER -