Abstract
The aim of this study is to investigate the solution algorithm for solving the problem of determining reliable shortest paths in road networks with stochastic travel times. The availability of reliable shortest paths enables travelers, in the face of travel time uncertainty, to plan their trips with a pre-specified on-time arrival probability. In this study, the reliable shortest path between origin and destination nodes is determined using a multiple-criteria shortest path approach when link travel times follow normal distributions. The dominance conditions involved in such problems are established, thereby reducing the number of generated non-dominated paths during the search processes. Two solution algorithms, multi-criteria label-setting and A* algorithms, are proposed and their complexities analyzed. Computational results using large scale networks are presented. Numerical examples using data from a real-world advanced traveller information system is also given to illustrate the applicability of the solution algorithms in practice.
Similar content being viewed by others
References
Bellman R (1958) On a routing problem. Q Appl Math 16:87–90
Brumbaugh-Smith J, Shier D (1989) An empirical investigation of some bicriterion shortest path algorithms. Eur J Oper Res 43:216–224
Chan KS, Lam WHK, Tam ML (2009) Real-time estimation of arterial travel times with spatial travel time covariance relationships. Transp Res Rec 2121:102–109
Chen A, Ji ZW (2005) Path finding under uncertainty. J Adv Transp 39:19–37
Chen A, Yang C, Kongsomsaksakul S, Lee M (2007a) Network-based accessibility measures for vulnerability analysis of degradable transportation networks. Netw Spat Econ 7:241–256
Chen YY, Bell MGH, Bogenberger K (2007b) Reliable pretrip multipath planning and dynamic adaptation for a centralized road navigation system. IEEE Trans Intell Transp Syst 8:14–20
Chen BY, Lam WHK, Tam ML (2008) Using a multi-criteria approach to solve path finding problem in road network with uncertainty. In: Proceeding of 13th Conference of the Hong Kong Society for Transportation Studies. Hong Kong, China, pp 229–238
Chen BY, Lam WHK, Sumalee A, Shao H (2011) An efficient solution algorithm for solving multi-class reliability-based traffic assignment problem. Math Comput Model 54:1428–1439
Chen BY, Lam WHK, Sumalee A, Li QQ, Li ZC (2012a) Vulnerability analysis for large-scale and congested road networks with demand uncertainty. Transp Res A Policy Pract 46:501–516
Chen BY, Lam WHK, Sumalee A, Li ZL (2012b) Reliable shortest path finding in stochastic networks with spatial correlated link travel times. Int J Geogr Inf Sci 26:365–386
Chen BY, Lam WHK, Sumalee A, Li QQ, Tam ML (2012c) Reliable shortest path problems in stochastic time-dependent networks. J Intell Transp Syst. Accepted for publication
Fan YY, Nie Y (2006) Optimal routing for maximizing the travel time reliability. Netw Spat Econ 6:333–344
Frank H (1969) Shortest paths in probabilistic graphs. Oper Res 17:583–599
Fredman ML, Tarjan RE (1987) Fibonacci heaps and their uses in improved network optimization algorithms. J ACM 34:596–615
Fu LP (2001) An adaptive routing algorithm for in-vehicle route guidance systems with real-time information. Transp Res B Methodol 35:749–765
Hart EP, Nilsson NJ, Raphael B (1968) A formal basis for the heuristic determination of minimum cost paths. IEEE Trans Syst Man Cybern Syst Hum 4:100–107
Hutson KR, Shier DR (2009) Extended dominance and a stochastic shortest path problem. Comput Oper Res 36:584–596
Lam TC, Small KA (2001) The value of time and reliability: measurement from a value pricing experiment. Transp Res E Logist Transp Rev 37:231–251
Lam WHK, Chan KS, Shi JWZ (2002) A traffic flow simulator for short-term travel time forecasting. J Adv Transp 36:265–291
Li ZC, Huang HJ, Lam WHK (2012) Modelling heterogeneous drivers’ response to route guidance and parking information system in stochastic and time-dependent networks. Transportmetrica 8:105–129
Miller-Hooks ED, Mahmassani HS (2003) Path comparisons for a priori and time-adaptive decisions in stochastic, time-varying networks. Eur J Oper Res 146:67–82
Mirchandani PB (1976) Shortest distance and reliability of probabilistic networks. Comput Oper Res 3:347–355
Nie Y, Wu X (2009) Shortest path problem considering on-time arrival probability. Transp Res B Methodol 43:597–613
Nie Y, Wu X (2011) Optimal path problems with second-order stochastic dominance constraints. Netw Spat Econ. doi:10.1007/s11067-011-9167-6
Nikolova E (2006) Stochastic shortest paths via quasi-convex maximization. In: Proceedings of 2006 European Symposium of Algorithms. Zurich, Switzerland, pp 552–563
Nikolova E (2010) High-performance heuristics for optimization in stochastic traffic engineering problems. In: Lirkov I, Margenov S, Wasniewski J (eds) Large-scale scientific computing. Springer, New York, pp 352–360
Rakha H, El-Shawarby I, Arafeh M, Dion F (2006) Estimating path travel time reliability. In: Proceedings of 2006 IEEE Intelligent Transportation Systems Conference. Toronto, Canada, pp 236–241
Reinhardt LB, Pisinger D (2011) Multi-objective and multi-constrained non-additive shortest path problems. Comput Oper Res 38:605–616
Sen S, Pillai R, Joshi S, Rathi AK (2001) A mean-variance model for route guidance in advanced traveler information systems. Transp Sci 35:37–49
Shao H, Lam WHK, Chan KS (2004) The problem of searching the reliable path for transportation networks with uncertainty. In: Proceeding of 9th Conference of the Hong Kong Society for Transportation Studies. Hong Kong, China, pp 226–234
Shao H, Lam WHK, Tam ML (2006) A reliability-based stochastic traffic assignment model for network with multiple user classes under uncertainty in demand. Netw Spat Econ 6:173–204
Skriver AJV, Andersen KA (2000) A label correcting approach for solving bicriterion shortest-path problems. Comput Oper Res 27:507–524
Tam ML, Lam WHK (2008) Using automatic vehicle identification data for travel time estimation in Hong Kong. Transportmetrica 4:179–194
Tam ML, Lam WHK, Lo HP (2008) Modeling air passenger travel behavior on airport ground access mode choices. Transportmetrica 4:135–153
Tian LJ, Huang HJ, Gao ZY (2012) A cumulative perceived value-based dynamic user equilibrium model considering the travelers’ risk evaluation on arrival time. Netw Spat Econ. doi:10.1007/s11067-011-9168-5
Waller ST, Ziliaskopoulos AK (2002) On the online shortest path problem with limited arc cost dependencies. Networks 40:216–227
Wu X, Nie Y (2011) Modeling heterogeneous risk-taking behavior in route choice: a stochastic dominance approach. Transp Res A Policy Pract 45:896–915
Acknowledgments
The authors are thankful to the anonymous referees for their comments and suggestions that improved this paper. The work described in this paper was jointly supported by a competitive earmarked research grant from the Research Grant Council of the Hong Kong Special Administration Region (PolyU 5195/07E), an internal research grant J-BB7Q from the Research Committee of the Hong Kong Polytechnic University, research grants from National Natural Science Foundation of China (70901073, 40830530 and 41071285), the National High-tech R&D Program of China (863 program) (2012AA12A211) and the School of Traffic & Transportation of Beijing Jiaotong University (2010RC005).
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix A. Notations
- A :
-
A set of links
- |A|:
-
Number of links in the network
- a ij :
-
A link connecting node i and node j
- b :
-
Travel time budget
- d is :
-
Euclidean distance from node i to destination
- F(i):
-
Heuristic value function for the path set P rs
- f(i):
-
Minimum travel time budget of the path set P rs
- G :
-
Directed network
- h(i):
-
Estimation of travel time budget from node i to destination
- N :
-
A set of nodes
- |N|:
-
Number of nodes in the network
- P rs :
-
A set of paths from origin to destination
- |P|:
-
Number of non-dominated paths between origin and destination
- \( p_u^{{rs}} \) :
-
A path from origin to destination
- \( p_{ * }^{{rs}} \) :
-
Optimal path from origin to destination
- PDS(i):
-
A set of predecessor nodes of node i
- r :
-
Origin node
- s :
-
Destination node
- SCS(i):
-
A set of successor nodes of node i
- T ij :
-
Travel time distribution of link a ij
- t ij :
-
Mean travel time of link a ij
- \( T_u^{{rs}} \) :
-
Travel time distribution of path \( p_u^{{rs}} \)
- \( t_u^{{rs}} \) :
-
Mean travel time of path \( p_u^{{rs}} \)
- ν max :
-
Maximum travel speed in the network
- VOT :
-
Value of time
- Z α :
-
Inverse cumulative distribution function for standard normal distribution at α confidence level
- α :
-
Probability of on-time arrival
- σ ij :
-
Travel time standard deviation of link a ij
- \( \sigma_u^{{rs}} \) :
-
Travel time standard deviation of path \( p_u^{{rs}} \)
- \( x_{{ij}}^{{rs}} \) :
-
Link-path incidence relationship
- \( \tau_u^{{rs}} \) :
-
Toll charge of path \( p_u^{{rs}} \)
- τ ij :
-
Toll charge of link a ij
- \( {{f}_{{T_u^{{rs}}}}}\left( \cdot \right) \) :
-
Probability density function of path travel time distribution
- \( {{\Phi }_{{T_u^{{rs}}}}}\left( \cdot \right) \) :
-
Cumulative distribution function of path travel time distribution
- \( \Phi_{{T_u^{{rs}}}}^{{ - 1}}\left( \cdot \right) \) :
-
Inverse cumulative distribution function of path travel Time distribution
Appendix B. Proofs
Theorem 1
A sub-path of any non-dominated path must be a non-dominated path itself.
Proof
Suppose \( p_v^{{ri}} \in {{P}^{{ri}}} \) is a sub-path of a non-dominated path \( p_v^{{rw}} = p_v^{{ri}} \oplus p_v^{{iw}} \in {{P}^{{rw}}} \), and it is dominated by another path \( p_u^{{ri}} \in {{P}^{{ri}}} \). According to Definition 3.3, we have \( \Phi_{{T_u^{{rs}}}}^{{ - 1}}\left( \alpha \right) < \Phi_{{T_v^{{rs}}}}^{{ - 1}}\left( \alpha \right) \) for any path \( p^{is} ∈ P^{is} \). Therefore, there exists at least one path \( p_u^{{rw}} = p_u^{{ri}} \oplus p_v^{{iw}} \in {{P}^{{rw}}} \) satisfying \( \Phi_{{T_u^{{rs}}}}^{{ - 1}}\left( \alpha \right) < \Phi_{{T_v^{{rs}}}}^{{ - 1}}\left( \alpha \right) \) for any \( p^{ws} ∈ P^{ws} \), since any \( p_v^{{iw}} \oplus {{p}^{{ws}}} \in {{P}^{{is}}} \). Thus, \( p_u^{{ri}} \succ p_v^{{ri}} \). This contradicts the fact that \( p_v^{{ri}} \) is a non-dominated path. □
Let \( p_u^{{rw}} = p_u^{{ri}} \oplus {{p}^{{iw}}} \) and \( p_v^{{rw}} = p_v^{{ri}} \oplus {{p}^{{iw}}} \) be two paths from origin r to node w going through the same sub-path p iw. \( T_u^{{rw}} \) and \( T_v^{{rw}} \) denote travel time for these two paths respectively. We introduce following function to facilitate the proofs of Propositions 2 and 3.
The \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) \) has following properties:
Lemma 1
-
(i)
If \( {{Z}_{\alpha }}\sigma_u^{{ri}} < {{Z}_{\alpha }}\sigma_v^{{ri}} \), and thus \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) \) is a monotonic increasing function, \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) \in \left[ {\Phi_{{T_u^{{ri}}}}^{{ - 1}}\left( \alpha \right) - \Phi_{{T_v^{{ri}}}}^{{ - 1}}\left( \alpha \right),{ }t_u^{{ri}} - t_v^{{ri}}} \right) \);
-
(ii)
If \( {{Z}_{\alpha }}\sigma_u^{{ri}} > {{Z}_{\alpha }}\sigma_v^{{ri}} \), and thus \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) \) is a monotonic decreasing function, \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) \in \left( {t_u^{{ri}} - t_u^{{ri}},\Phi_{{T_u^{{ri}}}}^{{ - 1}}\left( \alpha \right) - \Phi_{{T_v^{{ri}}}}^{{ - 1}}\left( \alpha \right)} \right] \);
-
(iii)
If \( {{Z}_{\alpha }}\sigma_u^{{ri}} = {{Z}_{\alpha }}\sigma_v^{{ri}} \), and \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) = t_u^{{ri}} - t_v^{{ri}} \).
Proof
The gradient of \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) \) can be formulated as \( g_{{uv}}^{{ri}}\prime \left( {{{p}^{{iw}}}} \right) = \frac{{\partial g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right)}}{{\partial {{{\left( {{{\sigma }^{{iw}}}} \right)}}^2}}} = \frac{{{{Z}_{\alpha }}}}{2}\frac{{\sqrt {{{{{\left( {\sigma_v^{{ri}}} \right)}}^2} + {{{\left( {{{\sigma }^{{iw}}}} \right)}}^2}}} - \sqrt {{{{{\left( {\sigma_u^{{ri}}} \right)}}^2} + {{{\left( {{{\sigma }^{{iw}}}} \right)}}^2}}} }}{{\sqrt {{{{{\left( {\sigma_u^{{ri}}} \right)}}^2} + {{{\left( {{{\sigma }^{{iw}}}} \right)}}^2} * }} \sqrt {{{{{\left( {\sigma_v^{{ri}}} \right)}}^2} + {{{\left( {{{\sigma }^{{iw}}}} \right)}}^2}}} }} \). Therefore, when \( {{Z}_{\alpha }}\sigma_u^{{ri}} < {{Z}_{\alpha }}\sigma_v^{{ri}} \), we have \( g_{{uv}}^{{ri}}\prime \left( {{{p}^{{iw}}}} \right) > 0 \) and \( \Phi_{{T_u^{{ri}}}}^{{ - 1}}\left( \alpha \right) - \Phi_{{T_v^{{ri}}}}^{{ - 1}}\left( \alpha \right) \leqslant g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) < t_u^{{ri}} - t_v^{{ri}} \); when \( {{Z}_{\alpha }}\sigma_u^{{ri}} > {{Z}_{\alpha }}\sigma_v^{{ri}} \), we have \( g_{{uv}}^{{ri}}\prime \left( {{{p}^{{iw}}}} \right) < 0 \) and \( t_u^{{ri}} - t_v^{{ri}} < g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) \leqslant \Phi_{{T_u^{{ri}}}}^{{ - 1}}\left( \alpha \right) - \Phi_{{T_v^{{ri}}}}^{{ - 1}}\left( \alpha \right) \); when \( {{Z}_{\alpha }}\sigma_u^{{ri}} = {{Z}_{\alpha }}\sigma_v^{{ri}} \), we have \( g_{{uv}}^{{ri}}\prime \left( {{{p}^{{iw}}}} \right) = 0 \) and \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) = t_u^{{ri}} - t_v^{{ri}} \).□
According to Definition 3, \( p_u^{{ri}} \succ p_v^{{ri}} \) is equivalent to \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) < 0 \) for any path p iw ∈ P iw. We will prove \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) < 0 \) for following dominance conditions.
Proposition 2
(M-V dominance). Given an on-time arrival probability α and two paths \( p_u^{{ri}} \ne p_v^{{ri}} \in {{P}^{{ri}}} \), \( p_u^{{ri}} \succ p_v^{{ri}} \) if \( p_u^{{ri}} \) and \( p_v^{{ri}} \) satisfy either
-
(i)
\( t_u^{{ri}} \leqslant t_v^{{ri}} \) and \( {{Z}_{\alpha }}\sigma_u^{{ri}} < {{Z}_{\alpha }}\sigma_v^{{ri}} \) or
-
(ii)
\( t_u^{{ri}} < t_v^{{ri}} \) and \( {{Z}_{\alpha }}\sigma_u^{{ri}} \leqslant {{Z}_{\alpha }}\sigma_v^{{ri}} \)
Proof
It can be easily followed by Lemma 1.□
Proposition 3
(M-B dominance). Given an on-time arrival probability α and two paths \( p_u^{{ri}} \ne p_v^{{ri}} \in {{P}^{{ri}}} \), \( p_u^{{ri}} \succ p_v^{{ri}} \) if \( p_u^{{ri}} \) and \( p_v^{{ri}} \) satisfy \( t_u^{{ri}} \leqslant t_v^{{ri}} \) and \( \Phi_{{T_u^{{ri}}}}^{{ - 1}}\left( \alpha \right) < \Phi_{{T_v^{{ri}}}}^{{ - 1}}\left( \alpha \right) \).
Proof
When \( {{Z}_{\alpha }}\sigma_u^{{ri}} < {{Z}_{\alpha }}\sigma_v^{{ri}} \) and \( t_u^{{ri}} \leqslant t_v^{{ri}} \) according to Lemma 1(i), we have \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) < \left( {t_u^{{ri}} - t_v^{{ri}}} \right) \leqslant 0 \). When \( {{Z}_{\alpha }}\sigma_u^{{ri}} > {{Z}_{\alpha }}\sigma_v^{{ri}} \) and \( t_u^{{ri}} \leqslant t_v^{{ri}} \) according to Lemma 1(ii), we have \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) < \left( {\Phi_{{T_u^{{ri}}}}^{{ - 1}}\left( \alpha \right) - \Phi_{{T_v^{{ri}}}}^{{ - 1}}\left( \alpha \right)} \right) \leqslant 0 \). When \( {{Z}_{\alpha }}\sigma_u^{{ri}} = {{Z}_{\alpha }}\sigma_v^{{ri}} \) and \( \Phi_{{T_u^{{ri}}}}^{{ - 1}}\left( \alpha \right) < \Phi_{{T_v^{{ri}}}}^{{ - 1}}\left( \alpha \right) \), we have \( t_u^{{ri}} < t_v^{{ri}} \). According to Lemma 1(iii), we have \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) = \left( {t_u^{{ri}} - t_v^{{ri}}} \right) < 0 \). Therefore, when \( t_u^{{ri}} \leqslant t_v^{{ri}} \) and \( \Phi_{{T_u^{{ri}}}}^{{ - 1}}\left( \alpha \right) < \Phi_{{T_v^{{ri}}}}^{{ - 1}}\left( \alpha \right) \), we have \( g_{{uv}}^{{ri}}\left( {{{p}^{{iw}}}} \right) < 0 \), ∀p iw ∈ P iw, ∀w ∈ N. □
Let \( {{f}_{{T_u^{{ri}}}}}(t) \) be the PDF of path travel time distribution \( T_u^{{ri}} \). It can be proved that the objective value \( \Phi_{{T_u^{{ri}}}}^{{ - 1}}\left( \alpha \right) \) monotonically increases with path extensions as follows:
Proposition 7
Given two paths \( p_u^{{ri}} \) and \( p_u^{{rj}} = p_u^{{ri}} \oplus {{a}_{{ij}}} \), the relationship \( \matrix{ {\Phi_{{T_u^{{rj}}}}^{{ - 1}}\left( \lambda \right) > \Phi_{{T_u^{{ri}}}}^{{ - 1}}\left( \lambda \right),} \hfill &{\forall \lambda \in \left( {0,1} \right)} \hfill \\ }<!end array> \) always holds.
Proof
\( \begin{array}{*{20}{c}} {{{\Phi }_{{T_{u}^{{rj}}}}}(b) - {{\Phi }_{{T_{u}^{{ri}}}}}(b) = \int_{0}^{b} {{{\Phi }_{{{{T}_{{ij}}}}}}\left( {b - t} \right){{f}_{{T_{u}^{{ri}}}}}(t)dt - \int_{0}^{b} {{{f}_{{T_{u}^{{ri}}}}}(t)dt} } } \hfill \\ { = \int_{0}^{b} {\left( {{{\Phi }_{{{{T}_{{ij}}}}}}\left( {b - t} \right) - 1} \right){{f}_{{T_{u}^{{ri}}}}}(t)dt < 0,} \quad \forall b \in {{R}^{ + }}} \hfill \\ { \Rightarrow \Phi _{{T_{u}^{{rj}}}}^{{ - 1}}\left( \lambda \right) > \Phi _{{T_{u}^{{ri}}}}^{{ - 1}}\left( \lambda \right),\quad \forall \lambda \in \left( {0,1} \right)} \hfill \\ \end{array} \). □
Rights and permissions
About this article
Cite this article
Chen, B.Y., Lam, W.H.K., Sumalee, A. et al. Finding Reliable Shortest Paths in Road Networks Under Uncertainty. Netw Spat Econ 13, 123–148 (2013). https://doi.org/10.1007/s11067-012-9175-1
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11067-012-9175-1