Abstract
In this paper, we study the optional preference model for the facility location game with two heterogeneous facilities on a line interval [0, 1], by further enforcing the requirement of a minimum distance \(0\le d\le 1\) between the two facilities. Each agent has one of three favorable preferences towards the two facilities, i.e., facility 1, facility 2, or optional preference. Here, we consider two variants of the optional preference model: Min (caring for the closer one) and Max (caring for the further one). In both variants, each agent wishes to get close to his preferred facilities, and thus his cost is his distance to his preferred facility. In this game, we consider agents’ locations as public information and agents’ preferences as private information which needs to be reported by agents. The objective is to design a mechanism for the two facilities’ locations such as to minimize the maximum cost of agents (MinMax) and achieve truthful report of agents’ preferences. Given any value of d, for both variants, we propose a strategyproof mechanism with an approximation ratio of 2. We also establish lower bounds of any deterministic strategyproof mechanism for both variants and show that the gaps between the lower bounds and the upper bounds are relatively small.
Similar content being viewed by others
Data availability
Enquiries about data availability should be directed to the authors.
References
Anastasiadis E, Deligkas A (2018) Heterogeneous facility location games. In: Proceedings of the 17th international conference on autonomous agents and multiagent systems (AAMAS), pp 623–631. https://doi.org/10.5555/3237383.3237476
Ben-Porat O, Goren G, Rosenberg I, et al (2019) From recommendation systems to facility location games. In: Proceedings of the AAAI conference on artificial intelligence (AAAI), pp 1772–1779. https://doi.org/10.1609/aaai.v33i01.33011772
Cai Q, Filos-Ratsikas A, Tang P (2016) Facility location with minimax envy. In: Proceedings of the 25th international joint conference on artificial intelligence (IJCAI), pp 137–143. https://doi.org/10.5555/3060621.3060641
Chen X, Fang Q, Liu W, et al (2020) Strategyproof mechanisms for 2-facility location games with minimax envy. In: International conference on algorithmic applications in management, Springer, pp 260–272. https://doi.org/10.1007/s10878-021-00711-7
Cheng Y, Yu W, Zhang G (2013) Strategy-proof approximation mechanisms for an obnoxious facility game on networks. Theor Comput Sci 497:154–163. https://doi.org/10.1016/j.tcs.2011.11.041
Duan L, Li B, Li M, et al (2019) Heterogeneous two-facility location games with minimum distance requirement. In: Proceedings of the 18th international conference on autonomous agents and multiagent systems (AAMAS), pp 1461–1469. https://doi.org/10.5555/3306127.3331858
Feigenbaum I, Sethuraman J (2015) Strategy proof mechanisms for one-dimensional hybrid and obnoxious facility location models. In: Workshop on incentive and trust in e-communities at the 29th AAAI conference on artificial intelligence (AAAI), pp 8–13
Fong KC, Li M, Lu P, et al (2018) Facility location games with fractional preferences. In: Proceedings of the 32th AAAI conference on artificial intelligence (AAAI), pp 1039–1046. https://ojs.aaai.org/index.php/AAAI/article/view/11458
Keijzer B, Wojtczak D (2018) Facility reallocation on the line. In: Proceedings of the 27th international joint conference on artificial intelligence (IJCAI), pp 188–194. https://doi.org/10.5555/3304415.3304443
Li M, Lu P, Yao Y, et al (2020) Strategyproof mechanism for two heterogeneous facilities with constant approximation ratio. In: Proceedings of the twenty-ninth international joint conference on artificial intelligence (IJCAI), pp 238–245
Lu P, Wang Y, Zhou Y (2009) Tighter bounds for facility games. In: Proceedings of the 5th international workshop on internet and network economics (WINE), pp 137–148. https://doi.org/10.1007/978-3-642-10841-9_14
Meir R (2019) Strategyproof facility location for three agents on a circle. In: Proceedings of the 18th international conference on autonomous agents and multiagent systems (AAMAS), pp 2120–2122. https://doi.org/10.1007/978-3-030-30473-7_2
Procaccia AD, Tennenholtz M (2009) Approximate mechanism design without money. In: Proceedings of the 10th ACM conference on Electronic commerce (ACM-EC), pp 177–186. https://doi.org/10.1145/2542174.2542175
Procaccia A, Wajc D, Zhang H (2018) Approximation-variance tradeoffs in facility location games. In: Proceedings of the AAAI conference on artificial intelligence (AAAI), pp 1–8. https://ojs.aaai.org/index.php/AAAI/article/view/11451
Serafino P, Ventre C (2014) Heterogeneous facility location without money on the line. In: Proceedings of the 21st European conference on artificial intelligence (ECAI), pp 807–812. https://doi.org/10.3233/978-1-61499-419-0-807
Sonoda A, Todo T, Yokoo M (2016) False-name-proof locations of two facilities: economic and algorithmic approaches. In: Proceedings of the 30th AAAI conference on artificial intelligence (AAAI), pp 615–621. https://doi.org/10.5555/3015812.3015904
Sui X, Boutilier C, Sandholm TW (2013) Analysis and optimization of multi-dimensional percentile mechanisms. In: Proceedings of the 27th AAAI conference on artificial intelligence (AAAI), pp 367–374
Tang P, Yu D, Zhao S (2020) Characterization of group-strategyproof mechanisms for facility location in strictly convex space. In: Proceedings of the 21st ACM conference on economics and computation (ACM-EC), pp 133–157. https://doi.org/10.1145/3391403.3399471
Todo T, Iwasaki A, Yokoo M (2011) False-name-proof mechanism design without money. In: Proceedings of the 10th International conference on autonomous agents and multiagent systems, vol 2 (AAMAS), pp 651–658
Wada Y, Ono T, Todo T, et al (2018) Facility location with variable and dynamic populations. In: Proceedings of the 17th international conference on autonomous agents and multiagent systems (AAMAS), pp 336–344. https://doi.org/10.5555/3237383.3237437
Xu X, Li B, Li M et al (2021) Two-facility location games with minimum distance requirement. J Artif Intell Res 70:719–756. https://doi.org/10.1613/jair.1.12319
Xu X, Li M, Duan L (2020) Strategyproof mechanisms for activity scheduling. In: Proceedings of the 19th international conference on autonomous agents and multiagent systems (AAMAS), pp 1539–1547. https://doi.org/10.5555/3398761.3398938
Yuan H, Wang K, Fong KC, et al (2016) Facility location games with optional preference. In: Proceedings of the 22nd European conference on artificial intelligence (ECAI), pp 1520–1527. https://doi.org/10.3233/978-1-61499-672-9-1520
Zhang Q, Li M (2014) Strategyproof mechanism design for facility location games with weighted agents on a line. J Comb Optim 28(4):756–773. https://doi.org/10.1007/s10878-013-9598-8
Zou S, Li M (2015) Facility location games with dual preference. In: Proceedings of the 14th international conference on autonomous agents and multiagent systems (AAMAS), pp 615–623. https://doi.org/10.5555/2772879.2773234
Acknowledgements
This work was fully supported by Project No. CityU 11200518 from Research Grants Council of HKSAR.
Funding
The authors have not disclosed any funding.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have not disclosed any competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix A: Proof of Theorem 1
Proof
Let \({\mathcal {M}}\) be Mechanism 1. Suppose that agent i misreports his preference from \(p_i\) to \(p_i'\). Denote preference set \({{\textbf {p}}}'=\{p_i',{{\textbf {p}}}_{-i}\}.\) Denote \((y_1,y_2)=\mathcal M({{\textbf {x}}}, {{\textbf {p}}})\in D_d\) and \((y_1',y_2')={\mathcal {M}}({\textbf {x}}, {{\textbf {p}}}')\in D_d.\) Then we have five cases.
Case 1: \(d\le x_n-x_1.\) In this case, by Mechanism 1, without loss of generality, we assume that
and thus \((y_1,y_2)=(x_1,x_n).\) We have three subcases for \(p_i.\)
-
1.
\(p_i=\{F_1\}.\) In this subcase, by Mechanism 1, agent i with preference \(\{F_1\}\) has his cost \(c_i((x_1,x_n)|(x_i,\{F_1\}))=x_i-x_1.\) After agent i misreports his preference to \(p_i'=\{F_2\}\) or \(p_i'=\{F_1,F_2\},\) if \((y_1',y_2')\) remains unchanged, agent i ’s cost is unchanged. Otherwise, we have \((y_1',y_2')=(x_n,x_1).\) Agent i’s cost changes to \(c_i((x_n,x_1)|(x_i,\{F_1\}))=x_n-x_i.\) Further, it holds that
$$\begin{aligned}&MC((x_1,x_n)|{{\textbf {x}}}, {{\textbf {p}}}')> MC((x_n,x_1)|{\textbf {x}}, {{\textbf {p}}}'). \end{aligned}$$(A2)Denote \(\alpha _1=MC((x_1,x_n)|{{\textbf {x}}}_{-i}, {{\textbf {p}}}_{-i})\) and \(\alpha _2=MC\) \(((x_n,x_1)|{{\textbf {x}}}_{-i}, {{\textbf {p}}}_{-i}).\) We have two sub-subcases for \(p_i'.\)
-
(a)
\(p_i'=\{F_2\}.\) We can simplify (A1) and (A2) as
$$\begin{aligned}&\max \{\alpha _1,x_i-x_1\}\le \max \{\alpha _2, x_n-x_i\}, \end{aligned}$$(A3)$$\begin{aligned}&\max \{\alpha _1,x_n-x_i\}> \max \{\alpha _2, x_i-x_1\}, \end{aligned}$$(A4)respectively. Assume by contradiction that \(x_i-x_1>x_n-x_i.\) By (A3), we have
$$\begin{aligned} \max \{\alpha _1,x_n-x_i\}\le \max \{\alpha _1,x_i-x_1\} \le&\max \{\alpha _2, x_n-x_i\}\le \max \{\alpha _2, x_i-x_1\}, \end{aligned}$$which contradicts (A4). Hence, \(x_i-x_1\le x_n-x_i,\) which implies that agent i cannot decrease his cost.
-
(b)
\(p_i'=\{F_1, F_2\}.\) We can simplify (A2) as
$$\begin{aligned}&\max \{\alpha _1,\min \{x_n-x_i, x_i-x_1\}\} >\max \{\alpha _2, \min \{x_n-x_i, x_i-x_1\}\}. \end{aligned}$$(A5)Assume by contradiction that \(x_i-x_1>x_n-x_i.\) By (A5),
$$\begin{aligned} \max \{\alpha _1, x_i-x_1\}\ge&\max \{\alpha _1,x_n-x_i\} =\max \{\alpha _1,\min \{x_n-x_i, x_i-x_1\}\} \nonumber \\ >&\max \{\alpha _2, \min \{x_n-x_i, x_i-x_1\}\} =\max \{\alpha _2, x_n-x_i\}, \end{aligned}$$which contradicts (A1) and (A3). Hence, \(x_i-x_1\le x_n-x_i,\) which implies that agent i cannot decrease his cost.
-
(a)
-
2.
\(p_i=\{F_2\}.\) Due to symmetry, this subcase is similar to Subcase 1.
-
3.
\(p_i=\{F_1, F_2\}.\) In this subcase, by Mechanism 1, agent i does not have the incentive to lie as the two possible outputs \((x_1,x_n)\) and \((x_n,x_1)\) are the same in his perspective.
Case 2: \(d> x_n-x_1\) and \(x_1\le 1-d.\) In this case, without loss of generality, suppose \({\mathcal {M}}=(y_1,y_2)=(x_1,x_1+d)\), and
We have the following three subcases for \(p_i\).
-
1.
\(p_i=\{F_1\}.\) In this subcase, by Mechanism 1, agent i with preference \(\{F_1\}\) has his cost \(c_i((x_1,x_1+d)|(x_i,\{F_1\}))=x_i-x_1.\) After agent i misreports his preference to \(p_i'=\{F_2\}\) or \(p_i'=\{F_1,F_2\},\) if \((y_1',y_2')\) remains unchanged, agent i ’s cost is unchanged. Otherwise, we have \((y_1',y_2')=(x_1+d,x_1).\) Agent i’s cost changes to \(c_i((x_1+d,x_1)|(x_i,\{F_1\}))=x_1+d-x_i.\) Further, it holds that
$$\begin{aligned}&MC((x_1,x_1+d)|{{\textbf {x}}}, {\textbf {p}}')>MC((x_1+d,x_1)|{{\textbf {x}}}, {{\textbf {p}}}'). \end{aligned}$$(A7)Denote \(\alpha _3=MC((x_1,x_1+d)|{{\textbf {x}}}_{-i}, {\textbf {p}}_{-i})\) and \(\alpha _4=MC\) \(((x_1+d,x_1)|{{\textbf {x}}}_{-i}, {{\textbf {p}}}_{-i}).\) We have two sub-subcases for \(p_i'.\)
-
(a)
\(p_i'=\{F_2\}.\) We can simplify (A6) and (A7) as
$$\begin{aligned}&\max \{\alpha _3,x_i-x_1\}\le \max \{\alpha _4, x_1+d-x_i\}, \end{aligned}$$(A8)$$\begin{aligned}&\max \{\alpha _3,x_1+d-x_i\}> \max \{\alpha _4, x_i-x_1\}, \end{aligned}$$(A9)respectively. Assume by contradiction that \(x_i-x_1>x_1+d-x_i.\) By (A8), we have
$$\begin{aligned}&\max \{\alpha _3,x_1+d-x_i\}\le \max \{\alpha _3,x_i-x_1\}\nonumber \\ \le&\max \{\alpha _4, x_1+d-x_i\}\le \max \{\alpha _4, x_i-x_1\}, \end{aligned}$$which contradicts (A9). Hence, \(x_i-x_1\le x_1+d-x_i,\) which implies that agent i cannot decrease his cost.
-
(b)
\(p_i'=\{F_1, F_2\}.\) We can simplify (A7) as
$$\begin{aligned}&\max \{\alpha _3,\min \{x_1+d-x_i, x_i-x_1\}\}\nonumber \\ >&\max \{\alpha _4, \min \{x_1+d-x_i, x_i-x_1\}\}. \end{aligned}$$(A10)Assume by contradiction that \(x_i-x_1>x_1+d-x_i.\) By (A10), we have
$$\begin{aligned}&\max \{\alpha _3, x_i-x_1\}\ge \max \{\alpha _3, x_1+d-x_i\}\\ =&\max \{\alpha _3,\min \{x_1+d-x_i, x_i-x_1\}\} \\ >&\max \{\alpha _4, \min \{x_1+d-x_i, x_i-x_1\}\} =\max \{\alpha _4, x_1+d-x_i\}, \end{aligned}$$which contradicts (A6) and (A8). Hence, \(x_i-x_1\le x_1+d-x_i,\) which implies that agent i cannot decrease his cost.
-
(a)
-
2.
\(p_i=\{F_2\}.\) Due to symmetry, this subcase is similar to Subcase 1.
-
3.
\(p_i=\{F_1, F_2\}.\) In this subcase, by Mechanism 1, agent i does not have the incentive to lie as the two possible outputs \((x_1,x_1+d)\) and \((x_1+d,x_1)\) are the same in his perspective.
Case 3: \(d> x_n-x_1,\) \(x_1> 1-d\), and \(x_n\ge d.\) Due to symmetry, this case is similar to Case 2.
Case 4: \(d>x_n-x_1,\) \(x_1> 1-d,\) \(x_n< d,\) and \(x_1+x_n\ge 1.\) In this case, we have \(d>1/2\) and all agents’ locations must be included between \(F_1\) and \(F_2.\) Thus, we can use similar steps in Case 2 to prove that Mechanism 1 is strategyproof.
Case 5: \(d>x_n-x_1,\) \(x_1> 1-d,\) \(x_n< d,\) and \(x_1+x_n< 1.\) Due to symmetry, this case is similar to Case 4.
In conclusion, by considering all cases, Mechanism 1 is strategyproof.
\(\square \)
Appendix B: Proof of Theorem 4
Proof
Let \({\mathcal {M}}\) be Mechanism 1. Suppose that agent i misreports his preference from \(p_i\) to \(p_i'\). Denote preference set \({{\textbf {p}}}'=\{p_i',{{\textbf {p}}}_{-i}\}.\) Denote \((y_1,y_2)=\mathcal M({{\textbf {x}}}, {{\textbf {p}}})\in D_d\) and \((y_1',y_2')={\mathcal {M}}({\textbf {x}}, {{\textbf {p}}}')\in D_d.\) We have five cases.
Case 1: \(d\le x_n-x_1.\) In this case, by Mechanism 1, without loss of generality, we assume that
and thus \((y_1,y_2)=(x_1,x_n).\) We have three subcases.
-
1.
\(p_i=\{F_1\}.\) In this subcase, by Mechanism 1, agent i with preference \(\{F_1\}\) has his cost \(c_i((x_1,x_n)|(x_i,\{F_1\}))=x_i-x_1.\) After agent i misreports his preference to \(p_i'=\{F_2\}\) or \(p_i'=\{F_1,F_2\},\) if \((y_1',y_2')\) remains unchanged, agent i ’s cost is unchanged. Otherwise, we have \((y_1',y_2')=(x_n,x_1).\) Agent i’s cost changes to \(c_i((x_n,x_1)|(x_i,\{F_1\}))=x_n-x_i.\) Further, it holds that
$$\begin{aligned}&MC((x_1,x_n)|{{\textbf {x}}}, {{\textbf {p}}}')> MC((x_n,x_1)|{\textbf {x}}, {{\textbf {p}}}'). \end{aligned}$$(B12)Denote \(\alpha _1=MC((x_1,x_n)|{{\textbf {x}}}_{-i}, {{\textbf {p}}}_{-i})\) and \(\alpha _2=MC\) \(((x_n,x_1)|{{\textbf {x}}}_{-i}, {{\textbf {p}}}_{-i}).\) We have two sub-subcases for \(p_i'.\)
-
(a)
\(p_i'=\{F_2\}.\) This sub-subcase is similar to Sub-subcase 1.1.a in the proof of Theorem 1.
-
(b)
\(p_i'=\{F_1, F_2\}.\) We can simplify (B11) and (B12) as
$$\begin{aligned}&\max \{\alpha _1,x_i-x_1\} \le \max \{\alpha _2, x_n-x_i\}, \end{aligned}$$(B13)$$\begin{aligned}&\max \{\alpha _1,\max \{x_n-x_i, x_i-x_1\}\} > \max \{\alpha _2, \max \{x_n-x_i, x_i-x_1\}\}, \end{aligned}$$(B14)respectively. Assume by contradiction that \(x_i-x_1>x_n-x_i.\) By (B14), we have
$$\begin{aligned} \max \{\alpha _1,x_i-x_1\}&=\max \{\alpha _1,\max \{x_n-x_i, x_i-x_1\}\}\\&>\max \{\alpha _2, \max \{x_n-x_i, x_i-x_1\}\}\\&=\max \{\alpha _2, x_i-x_1\}\ge \max \{\alpha _2, x_n-x_i\}, \end{aligned}$$which contradicts (B13). Hence, \(x_i-x_1\le x_n-x_i,\) which implies that agent i cannot decrease his cost.
-
(a)
-
2.
\(p_i=\{F_2\}.\) Due to symmetry, this subcase is similar to Subcase 1.
-
3.
\(p_i=\{F_1, F_2\}.\) In this subcase, by Mechanism 1, agent i does not have the incentive to lie as the two possible outputs \((x_1,x_n)\) and \((x_n,x_1)\) are the same in his perspective.
Case 2: \(d> x_n-x_1\) and \(x_1\le 1-d.\) In this case, without loss of generality, suppose \({\mathcal {M}}=(y_1,y_2)=(x_1,x_1+d)\), and
We have the following three subcases.
-
1.
\(p_i=\{F_1\}.\) In this subcase, by Mechanism 1, agent i with preference \(\{F_1\}\) has his cost \(c_i((x_1,x_1+d)|(x_i,\{F_1\}))=x_i-x_1.\) After agent i misreports his preference to \(p_i'=\{F_2\}\) or \(p_i'=\{F_1,F_2\},\) if \((y_1',y_2')\) remains unchanged, agent i ’s cost is unchanged. Otherwise, we have \((y_1',y_2')=(x_1+d,x_1).\) Agent i’s cost changes to \(c_i((x_1+d,x_1)|(x_i,\{F_1\}))=x_1+d-x_i.\) Further, it holds that
$$\begin{aligned}&MC((x_1,x_1+d)|{{\textbf {x}}}, {{\textbf {p}}}')> MC((x_1+d,x_1)|{{\textbf {x}}}, {{\textbf {p}}}'). \end{aligned}$$(B16)Denote \(\alpha _3=MC((x_1,x_1+d)|{{\textbf {x}}}_{-i}, {\textbf {p}}_{-i})\) and \(\alpha _4=MC((x_1+d,x_1)|{{\textbf {x}}}_{-i}, {\textbf {p}}_{-i}).\) We have two sub-subcases for \(p_i'.\)
-
(a)
\(p_i'=\{F_2\}.\) This sub-subcase is similar to Sub-subcase 2.1.a in the proof of Theorem 1.
-
(b)
\(p_i'=\{F_1, F_2\}.\) We can simplify (B15) and (B16) as
$$\begin{aligned}&\max \{\alpha _3,x_i-x_1\} \le \max \{\alpha _4, x_1+d-x_i\}, \end{aligned}$$(B17)$$\begin{aligned}&\max \{\alpha _3,\max \{x_1+d-x_i, x_i-x_1\}\}\nonumber \\&> \max \{\alpha _4, \max \{x_1+d-x_i, x_i-x_1\}\}, \end{aligned}$$(B18)respectively. Assume by contradiction that \(x_i-x_1>x_1+d-x_i.\) By (B18), we have
$$\begin{aligned} \max \{\alpha _3, x_i-x_1\} =&\max \{\alpha _3,\max \{x_1+d-x_i, x_i-x_1\}\} \\ >&\max \{\alpha _4, \max \{x_1+d-x_i, x_i-x_1\}\}\\ =&\max \{\alpha _4, x_i-x_1\}\ge \max \{\alpha _4, x_1+d-x_i\}, \end{aligned}$$which contradicts (B17). Hence, \(x_i-x_1\le x_1+d-x_i,\) which implies that agent i cannot decrease his cost.
-
(a)
-
2.
\(p_i=\{F_2\}.\) Due to symmetry, this subcase is similar to Subcase 1.
-
3.
\(p_i=\{F_1, F_2\}.\) In this subcase, by Mechanism 1, agent i does not have the incentive to lie as the two possible outputs \((x_1,x_1+d)\) and \((x_1+d,x_1)\) are the same in his perspective.
Case 3: \(d> x_n-x_1,\) \(x_1> 1-d\), and \(x_n\ge d.\) Due to symmetry, this case is similar to Case 2.
Case 4: \(d> x_n-x_1,\) \(x_1> 1-d\), \(x_n<d,\) and \(x_1+x_n< 1.\) In this case, we have \(d>1/2\) and all agents’ locations must be included between \(F_1\) and \(F_2.\) Thus, we can use similar steps in Case 2 to prove that Mechanism 1 is strategyproof.
Case 5: \(d>x_n-x_1,\) \(x_1> 1-d,\) \(x_n< d,\) and \(x_1+x_n< 1.\) Due to symmetry, this case is similar to Case 4.
In conclusion, by considering all cases, Mechanism 1 is strategyproof.
\(\square \)
Appendix C: Proof of Theorem 5
Proof
Let \({\mathcal {M}}\) be Mechanism 1 and \(\bar{{\mathcal {M}}}\) be the optimal solution for minimizing the maximum cost. Denote \((y_1,y_2)={\mathcal {M}}({{\textbf {x}}}, {{\textbf {p}}})\in D_d\) and \((y_1^\star ,y_2^\star )=\bar{{\mathcal {M}}}({{\textbf {x}}}, {{\textbf {p}}})\in D_d.\) Let agent \(m\in N\) be the agent where the maximum cost comes from in Mechanism 1. We have five cases.
Case 1: \(d\le |x_1-x_n|.\) According to the preference of agent 1 and agent n, there are five subcases.
-
1.
\(p_1=p_n=\{F_1\}.\) This subcase is similar to Subcase 1.1 in the proof of Theorem 2 and thus \(\gamma \le 2.\)
-
2.
\(p_1=\{F_1\}\) and \(p_n=\{F_2\}.\) In this subcase, by Mechanism 1, \({\mathcal {M}}=(y_1,y_2)=(x_1,x_n).\) According to agent m’s preference, we have three sub-subcases.
-
(a)
\(p_m=\{F_1\}.\) This sub-subcase is similar to Sub-subcase 1.2.a in the proof of Theorem 2 and thus \(\gamma \le 2.\)
-
(b)
\(p_m=\{F_2\}.\) Due to symmetry, this sub-subcase is similar to Sub-subcase 1.2.a.
-
(c)
\(p_m=\{F_1,F_2\}.\) Without loss of generality, we assume that \(x_n-x_m\ge x_m-x_1.\) For Mechanism 1, we have the maximum cost \(MC({\mathcal {M}}|{{\textbf {x}}},{{\textbf {p}}})=x_n-x_m.\) For the optimal solution \(\bar{{\mathcal {M}}},\) we have the maximum cost
$$\begin{aligned} MC(\bar{{\mathcal {M}}}|{{\textbf {x}}},{{\textbf {p}}}) \ge&\max \{|y_2^\star -x_n|, |y_2^\star -x_m|\}\ge ({x_n-x_m})/{2}. \end{aligned}$$Thus, the approximation ratio is \(\gamma \le 2\).
-
(a)
-
3.
\(p_1=p_n=\{F_2\}.\) The analysis is similar to Subcase 1.1 by symmetry.
-
4.
\(p_1=\{F_2\}\) and \(p_n=\{F_1\}.\) The analysis is similar to Subcase 1.2 by symmetry.
-
5.
\(\{F_1, F_2\} \in \{p_1, p_n\}.\) In this subcase, if agent 1 and agent n are of the same type (\(F_1\)-typed or \(F_2\)-typed), then the analysis will be similar to Subcase 1.1 or 1.3. If they are of different types, then the analysis will be similar to Subcase 1.2 or 1.4.
Case 2: \(d> x_n-x_1\) and \(x_1\le 1-d.\) According to the preference of agent 1 and agent n, there are five subcases.
-
1.
\(p_1=p_n=\{F_1\}.\) In this subcase, by Mechanism 1, \(\mathcal M=(y_1,y_2)=(x_1,x_1+d).\) According to agent m’s preference, we have three sub-subcases.
-
(a)
\(p_m=\{F_1\}.\) This sub-subcase is similar to Sub-subcase 2.1.a in the proof of Theorem 2 and thus \(\gamma \le 2.\)
-
(b)
\(p_m=\{F_2\}.\) This sub-subcase is similar to Sub-subcase 2.1.b in the proof of Theorem 2 and thus \(\gamma \le 2.\)
-
(c)
\(p_m=\{F_1, F_2\}.\) For Mechanism 1, we have
$$\begin{aligned} MC({\mathcal {M}}|{{\textbf {x}}},{{\textbf {p}}})=&\max \{x_m-x_1, x_1+d-x_m\}\\ \ge&c_n((x_1,x_1+d)|(x_n,\{F_1\}))=x_n-x_1. \end{aligned}$$If \(x_m=x_n,\) we have \(MC({\mathcal {M}}|{{\textbf {x}}},{\textbf {p}})=x_m-x_1.\) In this situation, for the optimal solution \(\bar{{\mathcal {M}}},\) we have the maximum cost
$$\begin{aligned} MC(\bar{{\mathcal {M}}}|{{\textbf {x}}},{{\textbf {p}}}) \ge \max \{|y_1^\star -x_1|, |y_1^\star -x_n|\}\ge (x_n-x_1)/2 \end{aligned}$$and thus \(\gamma \le 2.\) Otherwise, we have \(MC({\mathcal {M}}|{{\textbf {x}}},{\textbf {p}})=x_1+d-x_m\ge x_n-x_1,\) if \(x_m<x_n.\) In this situation, for the optimal solution \(\bar{{\mathcal {M}}},\) we have the maximum cost
$$\begin{aligned} MC(\bar{{\mathcal {M}}}|{{\textbf {x}}},{{\textbf {p}}}) \ge&\max \{|y_1^\star -x_1|,|y_1^\star -x_n|, |y_1^\star -x_m|,|y_2^\star -x_m|\}\nonumber \\ =&\max \{|y_1^\star -x_1|,|y_1^\star -x_n|,|y_2^\star -x_m|\}\nonumber \\ \ge&\left\{ \begin{array}{ll} \frac{d+x_m-x_1}{2} &{}\text{ if } x_1\le x_m\le \frac{x_1+x_n}{2}\\ \frac{d+x_n-x_m}{2} &{}\text{ if } \frac{x_1+x_n}{2}<x_m\le x_n \end{array} \right. \ge&({d+x_1-x_m})/{2}, \end{aligned}$$(C19)where the equality in the second inequality follows when \((y_1^\star ,y_2^\star )=(\frac{x_1+x_m+d}{2},\) \(\frac{x_1+x_m-d}{2})\) if \(x_1\le x_m\le \frac{x_1+x_n}{2};\) or \((y_1^\star ,y_2^\star )=(\frac{x_m+x_n-d}{2},\frac{x_m+x_n+d}{2})\) if \(\frac{x_1+x_n}{2}<x_m\le x_n.\) Thus, the approximation ratio is \(\gamma \le 2\).
-
(a)
-
2.
\(p_1=\{F_1\}\) and \(p_n=\{F_2\}.\) In this subcase, by Mechanism 1, \({\mathcal {M}}=(y_1,y_2)=(x_1,x_1+d).\) According to agent m’s preference, we have three sub-subcases.
-
(a)
\(p_m=\{F_1\}.\) This sub-subcase is similar to Sub-subcase 2.2.a in the proof of Theorem 2 and thus \(\gamma \le 2.\)
-
(b)
\(p_m=\{F_2\}.\) This sub-subcase is similar to Sub-subcase 2.2.b in the proof of Theorem 2 and thus \(\gamma \le 2.\)
-
(c)
\(p_m=\{F_1, F_2\}.\) For Mechanism 1, we have
$$\begin{aligned} MC({\mathcal {M}}|{{\textbf {x}}},{{\textbf {p}}})=\max \{x_m-x_1,x_1+d-x_m\} \le d. \end{aligned}$$For the optimal solution \(\bar{{\mathcal {M}}},\) we have the maximum cost
$$\begin{aligned} MC(\bar{{\mathcal {M}}}|{{\textbf {x}}},{{\textbf {p}}}) \ge \max \{|y_1^\star -x_m|,|y_2^\star -x_m|\}\ge d/2. \end{aligned}$$Thus, the approximation ratio is \(\gamma \le 2.\)
-
(a)
-
3.
\(p_1=p_n=\{F_2\}.\) The analysis is similar to Subcase 2.1 by symmetry.
-
4.
\(p_1=\{F_2\}\) and \(p_n=\{F_1\}.\) The analysis is similar to Subcase 2.2 by symmetry.
-
5.
\(\{F_1, F_2\} \in \{p_1, p_n\}.\) In this subcase, if agent 1 and agent n are of the same type (\(F_1\)-typed or \(F_2\)-typed), then the analysis will be similar to Subcase 2.1 or 2.3. If they are of different types, then the analysis will be similar to Subcase 2.2 or 2.4.
Case 3: \(d> x_n-x_1,\) \(x_1> 1-d\), and \(x_n\ge d.\) Due to symmetry, this case is similar to Case 2.
Case 4: \(d> x_n-x_1,\) \(x_1> 1-d\), \(x_n<d,\) and \(x_1+x_n\ge 1.\) In this case, by \(1-d<x_1\le x_n<d,\) d should satisfy \(d>1/2.\) Note that in any mechanism, all agents’ locations must be between \(F_1\) and \(F_2.\) According to the preference of agent 1 and agent n, there are five subcases.
-
1.
\(p_1=p_n=\{F_1\}.\) In this subcase, by Mechanism 1 and \(x_1+x_n\ge 1\ge d\), \({\mathcal {M}}=(y_1,y_2)=(d,0).\) According to agent m’s preference, we have three sub-subcases.
-
(a)
\(p_m=\{F_1\}.\) This sub-subcase is similar to Sub-subcase 4.1.a in the proof of Theorem 2 and thus \(\gamma =1.\)
-
(b)
\(p_m=\{F_2\}.\) This sub-subcase is similar to Sub-subcase 4.1.a in the proof of Theorem 2 and thus \(\gamma \le 2.\)
-
(c)
\(p_m=\{F_1, F_2\}.\) For Mechanism 1, we have
$$\begin{aligned} MC({\mathcal {M}}|{{\textbf {x}}},{{\textbf {p}}}) =&\max \{x_m, d-x_m\} \ge c_1((d,0)|(x_n,\{F_1\}))=d-x_1. \end{aligned}$$If \(x_m=x_1,\) we have \(MC({\mathcal {M}}|{{\textbf {x}}},{\textbf {p}})=d-x_1\ge x_1.\) In this situation, for the optimal solution \(\bar{{\mathcal {M}}},\) we have the maximum cost
$$\begin{aligned} MC(\bar{{\mathcal {M}}}|{{\textbf {x}}},{{\textbf {p}}}) \ge \max \{|y_1^\star -x_1|, |y_1^\star -x_n|\}\ge d-x_1 \end{aligned}$$and thus \(\gamma =1.\) Otherwise, we have \(MC({\mathcal {M}}|{{\textbf {x}}},{{\textbf {p}}})=x_m\ge d-x_1.\) In this situation, for the optimal solution \(\bar{{\mathcal {M}}},\) (C19) still holds here, and further by (C19), we have the maximum cost \( MC(\bar{{\mathcal {M}}}|{{\textbf {x}}},{{\textbf {p}}})\ge x_m/2. \) Thus, the approximation ratio is \(\gamma \le 2\).
-
(a)
-
2.
\(p_1=\{F_1\}\) and \(p_n=\{F_2\}.\) In this subcase, by Mechanism 1, \({\mathcal {M}}=(y_1,y_2)=(0,d).\) According to agent m’s preference, we have three sub-subcases.
-
(a)
\(p_m=\{F_1\}.\) This sub-subcase is similar to Sub-subcase 4.2.a in the proof of Theorem 2 and thus \(\gamma \le 2.\)
-
(b)
\(p_m=\{F_2\}.\) This sub-subcase is similar to Sub-subcase 4.2.b in the proof of Theorem 2 and thus \(\gamma \le 2.\)
-
(c)
\(p_m=\{F_1, F_2\}.\) For Mechanism 1, we have the maximum cost
$$\begin{aligned} MC({\mathcal {M}}|{{\textbf {x}}},{{\textbf {p}}})=\max \{x_m,d-x_m\}\le d. \end{aligned}$$For the optimal solution \(\bar{{\mathcal {M}}},\) we have the maximum cost
$$\begin{aligned}&MC(\bar{{\mathcal {M}}}|{{\textbf {x}}},{{\textbf {p}}}) \ge \max \{|y_1^\star -x_m|,|y_2^\star -x_m|\}\ge d/2. \end{aligned}$$Thus, the approximation ratio is \(\gamma \le 2.\)
-
(a)
-
3.
\(p_1=p_n=\{F_2\}.\) The analysis is similar to Subcase 4.1 by symmetry.
-
4.
\(p_1=\{F_2\}\) and \(p_n=\{F_1\}.\) The analysis is similar to Subcase 4.2 by symmetry.
-
5.
\(\{F_1, F_2\} \in \{p_1, p_n\}.\) In this subcase, if agent 1 and agent n are of the same type (\(F_1\)-typed or \(F_2\)-typed), then the analysis will be similar to Subcase 4.1 or 4.3. If they are of different types, then the analysis will be similar to Subcase 4.2 or 4.4.
In conclusion, by considering all cases, Mechanism 1 has an approximation ratio \(\gamma =2.\)
\(\square \)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Xu, X., Zhang, J. & Xie, L. Minmax for facility location game with optional preference under minimum distance requirement. J Comb Optim 46, 23 (2023). https://doi.org/10.1007/s10878-023-01087-6
Accepted:
Published:
DOI: https://doi.org/10.1007/s10878-023-01087-6