Advertisement

Improving Selfish Routing for Risk-Averse Players

  • 17 Accesses

Abstract

We investigate how and to which extent one can exploit risk-aversion and modify the perceived cost of the players in selfish routing so that the Price of Anarchy (PoA) wrt. the total latency is improved. The starting point is to introduce some small random perturbations to the edge latencies so that the expected latency does not change, but the perceived cost of the players increases, due to risk-aversion. We adopt the simple model of γ-modifiable routing games, a variant of selfish routing games with restricted tolls. We prove that computing the best γ-enforceable flow is NP-hard for parallel-link networks with affine latencies and two classes of heterogeneous risk-averse players. On the positive side, we show that for parallel-link networks with heterogeneous players and for series-parallel networks with homogeneous players, there exists a nicely structured γ-enforceable flow whose PoA improves fast as γ increases. We show that the complexity of computing such a γ-enforceable flow is determined by the complexity of computing a Nash flow for the original game. Moreover, we prove that the PoA of this flow is best possible in the worst-case, in the sense that there are instances where (i) the best γ-enforceable flow has the same PoA, and (ii) considering more flexible modifications does not lead to any improvement on the PoA.

This is a preview of subscription content, log in to check access.

Access options

Buy single article

Instant unlimited access to the full article PDF.

US$ 39.95

Price includes VAT for USA

Subscribe to journal

Immediate online access to all issues from 2019. Subscription will auto renew annually.

US$ 99

This is the net price. Taxes to be calculated in checkout.

Notes

  1. 1.

    Despite our brief discussion about how such an upper bound γe can be determined, we deliberately avoid getting into the details of how γe’s are calculated. This depends crucially (and not always in a simple way) on the particular practical application and cannot be incorporated into a theoretical model.

  2. 2.

    To simplify the model and make it easily applicable to general networks, we assume that the latency modifications (and the resulting individual costs of the players) are separable. This is a relatively standard simplifying assumption (see e.g., [16, 18]) on the structure of risk-averse individual costs in networks and only affects the extension of our results to series-parallel networks.

  3. 3.

    Note that if ℓe(x) is not strictly increasing, xe may not be uniquely defined (and it may be \(L^{\prime } = L\)). Then, for each eE1, we let xe be the largest value such that \(f^{\prime }_{e} + x_{e} \leq o_{e}\) and \(L^{\prime } = \ell _{e}(f^{\prime }_{e}+x_{e})\) (i.e., if \(L^{\prime } = \ell _{e}(o_{e})\), xe becomes \(o_{e} - f^{\prime }_{e}\) so that e moves from E1 to E2). For each eE3, we let xe be the smallest value such that \(L^{\prime } = \ell _{e}(f^{\prime }_{e}+x_{e})\).

  4. 4.

    Before rerouting, the equilibrium cost for the links in E1E3 (with traffic rate \({\sum }_{e \in E_{1} \cup E_{3}} f^{\prime }_{e}\)) is \(L < L^{\prime }\) and e(om) > L. After we reroute x units of flow from link m to E1E3, the equilibrium cost for the links in E1E3 (with traffic rate \({\sum }_{e \in E_{1} \cup E_{3}} (f^{\prime }_{e} + x_{e})\)) is \(L^{\prime } > L\) and \(\ell _{e}(o_{m}-x) < L^{\prime } < L\). Hence, by continuity and due to the parallel link structure of the network, the unique equilibrium flow has equilibrium cost \(L^{\prime \prime } \in (L, L^{\prime })\).

  5. 5.

    For that, recall that the flow that minimizes the total latency is an equilibrium for the game where the latency on every edge e has been changed from e(x) to (xe(x)). But since all latency functions of edges in X2 are affine and share the same constant (recall i(x) = (x/si) + 4), equal latencies with respect to (xe(x)) is equivalent to equal latencies with respect to e(x). Thus the optimal total latency is achieved when the edges in X2 are at equilibrium, sharing the same latency.

  6. 6.

    Note that such edges will remain non-optimal, because in later steps, they may only lose more flow.

  7. 7.

    Note that such edges will remain non-optimal, because in later steps, they may only gain more flow.

  8. 8.

    For completeness, one can give a detailed proof of this statement by induction on the decomposition of \(G^{\mathbf {\Gamma }_{1}}_{1}\) and \(G^{\mathbf {\Gamma }_{2}}_{2}\).

  9. 9.

    Any (unused) path p with an unused edge has \(\ell _{p}(o) \geq \max \nolimits _{p: o_{p} > 0} \ell _{p}(o)\). Moreover, the perceived cost of p can only increase due to edge modifications. Since the modifications corresponding to the solution of (Oγ) make the perceived cost of all used paths equal to \(\max \nolimits _{p: o_{p} > 0} \ell _{p}(o)\), o becomes a Nash flow of \(\mathcal {G}^{\mathbf {\Gamma }}\).

  10. 10.

    Property (d) requires that \(\mathcal {D}\) should be closed under addition of constants, as long as the resulting function remains nonnegative.

References

  1. 1.

    Angelidakis, H., Fotakis, D., Lianeas, T.: Stochastic congestion games with risk-averse players. In: Proc. of the 6th international symposium on algorithmic game theory (SAGT ’13), vol. 8146 of LNCS, pp. 86–97 (2013)

  2. 2.

    Bonifaci, V., Salek, M., Schäfer, G.: Efficiency of restricted tolls in non-atomic network routing games. In: Proceedings of the 4th international symposium on algorithmic game theory (SAGT ’10), volume 6982 of LNCS, pp. 302–313. Springer-Verlag (2011)

  3. 3.

    Correa, J.R., Schulz, A.S., Stier Moses, N.E.: Selfish routing in capacitated networks. Math. Oper. Res. 29(4), 961–976 (2004)

  4. 4.

    Fiat, A., Papadimitriou, C.H.: When the players are not expectation maximizers. In: Proceedings of the 3th international symposium on algorithmic game theory (SAGT ’10), volume 6386 of LNCS, pp. 1–14 (2010)

  5. 5.

    Fleischer, L.: Linear Tolls Suffice: New Bounds and Algorithms for Tolls in Single Source Networks. Theor. Comput. Sci. 348, 217–225 (2005)

  6. 6.

    Fleischer, L., Jain, K., Mahdian, M.: Tolls for heterogeneous selfish users in multicommodity networks and generalized congestion games. In: Proc. of the 45th IEEE Symp. on Foundations of Computer Science (FOCS ’04), pp. 277–285 (2004)

  7. 7.

    Fotakis, D., Spirakis, P.: Cost-balancing tolls for atomic network congestion games. In: Proceedings of the 3rd international workshop on internet and network economics (WINE ’07), pp. 179–190. Springer-Verlag (2007)

  8. 8.

    Fotakis, D., Kalimeris, D., Lianeas, T.: Improving selfish routing for risk-averse players. In: Evangelos Markakis and Guido Schȧfer, editors, Web and Internet Economics - 11th International Conference, WINE 2015, Amsterdam, The Netherlands, December 9-12 Proceedings, volume 9470 of Lecture Notes in Computer Science, pp. 328–342. Springer (2015)

  9. 9.

    Hall, M.A.: Properties of the Equilibrium State in Transportation Networks. Transp. Sci. 12(3), 208–216 (1978)

  10. 10.

    Hoefer, M., Olbrich, L., Skopalik, A.: Taxing subnetworks. In: Proceedings of the 4th international workshop on internet and network economics (WINE ’08), vol. 5385 of LNCS, pp. 286–294. Springer-Verlag (2008)

  11. 11.

    Jelinek, T., Klaas, M., Schäfer, G.: Computing optimal tolls with arc restrictions and heterogeneous players. In: Proceedings of the 31st annual symposium on theoretical aspects of computer science (STACS ’14), volume 25 of LIPIcs, pages 433–444. Schloss Dagstuhl - Leibniz-Zentrum fuer Informatik (2014)

  12. 12.

    Karakostas, G., Kolliopoulos, S.: Edge pricing of multicommodity networks for heterogeneous users. In: Proceedings of the 45th IEEE Symp. on Foundations of Computer Science (FOCS ’04), pp. 268–276 (2004)

  13. 13.

    Kleer, P., Schäfer, G.: The impact of worst-case deviations in non-atomic network routing games. In: Martin Gairing and Rahul Savani, editors, Algorithmic Game Theory - 9th International Symposium, SAGT 2016, Liverpool, UK, September 19-21 Proceedings, volume 9928 of Lecture Notes in Computer Science, pp. 129–140. Springer (2016)

  14. 14.

    Lianeas, T., Nikolova, E., Stier Moses, N.E.: Asymptotically tight bounds for inefficiency in risk-averse selfish routing. In: Kambhampati, S. (ed.) Proceedings of the 25th international joint conference on artificial intelligence, IJCAI 2016, pp 338–344. IJCAI/AAAI Press, New York (2016)

  15. 15.

    Nikolova, E., Stier Moses, N.: Stochastic selfish routing. In: Proceedings of the 4th international symposium on algorithmic game theory (SAGT ’11), volume 6982 of LNCS, pp. 314–325 (2011)

  16. 16.

    Nikolova, E., Stier-Moses, N.: The burden of risk aversion in mean-risk selfish routing. In: Proceedings of the 16th ACM conference on electronic commerce (EC ’15), pp. 489–506 (2015)

  17. 17.

    Ordóñez, F., Stier Moses, N.: Wardrop equilibria with risk-averse users. Transp. Sci. 44(1), 63–86 (2010)

  18. 18.

    Piliouras, G., Nikolova, E., Shamma, J.S.: Risk sensitivity of price of anarchy under uncertainty. In: Proceedings of the 14th ACM conference on electronic commerce (EC ’13), pp. 715–732 (2013)

  19. 19.

    Roughgarden, T.: Stackelberg scheduling strategies. SIAM J.Comput. 33(2), 332–350 (2004)

  20. 20.

    Roughgarden, T.: Selfish routing and the price of anarchy. MIT Press (2005)

  21. 21.

    Tversky, A., Kahneman, D.: Prospect theory: An analysis of decision under risk. Econometrica 47(2), 263–291 (1979)

  22. 22.

    Valdez, J., Tarjan, R.E., Lawler, E.L.: The Recognition of Series-Parallel Digraphs. SIAM J. Comput. 11(2), 298–313 (1982)

Download references

Author information

Correspondence to Thanasis Lianeas.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

An extended abstract of this work [8] has appeared in the 11th Conference on Web and Internet Economics (WINE 2016). Research was supported by the project Algorithmic Game Theory, co-financed by the European Union (European Social Fund) and Greek national funds, through the Operational Program “Education and Lifelong Learning” of the National Strategic Reference Framework -Research Funding Program: THALES, investing in knowledge society through the European Social Fund, and by grant NSF CCF 1216103. The majority of this work was done while the second author was at the Department of Informatics and Telecommunications of the University of Athens.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Fotakis, D., Kalimeris, D. & Lianeas, T. Improving Selfish Routing for Risk-Averse Players. Theory Comput Syst 64, 339–370 (2020). https://doi.org/10.1007/s00224-019-09946-8

Download citation

Keywords

  • Selfish routing
  • Uncertainty
  • Risk-aversion
  • Restricted tolls
  • Price of anarchy