Introduction The Poissonian City Variance and efficiency Flows The Poissonian City Wilfrid Kendall w.s.kendall@warwick.ac.uk Mathematics of Phase Transitions Past, Present, Future 13 November 2009 Conclusion References Introduction The Poissonian City Variance and efficiency Flows Conclusion References A problem in frustrated optimization Consider N cities x (N) = {x1 , . . . , xN } in square side √ N. Introduction The Poissonian City Variance and efficiency Flows Conclusion References A problem in frustrated optimization Consider N cities x (N) = {x1 , . . . , xN } in square side Assess road network G = G(x (N) ) √ N. connecting cities by: Introduction The Poissonian City Variance and efficiency Flows Conclusion References A problem in frustrated optimization Consider N cities x (N) = {x1 , . . . , xN } in square side Assess road network G = G(x (N) ) network total road length len(G) √ N. connecting cities by: Introduction The Poissonian City Variance and efficiency Flows Conclusion References A problem in frustrated optimization Consider N cities x (N) = {x1 , . . . , xN } in square side Assess road network G = G(x (N) ) √ N. connecting cities by: network total road length len(G) (minimized by Steiner minimum tree ST(x (N) )); Introduction The Poissonian City Variance and efficiency Flows Conclusion References A problem in frustrated optimization Consider N cities x (N) = {x1 , . . . , xN } in square side Assess road network G = G(x (N) ) √ N. connecting cities by: network total road length len(G) (minimized by Steiner minimum tree ST(x (N) )); versus average network distance between two random cities, average(G) = XX 1 distG (xi , xj ) , N(N − 1) i≠j Introduction The Poissonian City Variance and efficiency Flows Conclusion References A problem in frustrated optimization Consider N cities x (N) = {x1 , . . . , xN } in square side Assess road network G = G(x (N) ) √ N. connecting cities by: network total road length len(G) (minimized by Steiner minimum tree ST(x (N) )); versus average network distance between two random cities, average(G) = XX 1 distG (xi , xj ) , N(N − 1) i≠j (minimized by laying tarmac for complete graph). Introduction The Poissonian City Variance and efficiency Flows Conclusion References Questions Aldous and Kendall (2008) provide answers for the First Question √ Consider a configuration x (N) of N cities in [0, N]2 as above, and a well-chosen connecting network G = G(x (N) ). How does large-N trade-off between len(G) and average(G) behave? (And how clever do we have to be to get a good trade-off?) Introduction The Poissonian City Variance and efficiency Flows Conclusion References Questions Aldous and Kendall (2008) provide answers for the First Question √ Consider a configuration x (N) of N cities in [0, N]2 as above, and a well-chosen connecting network G = G(x (N) ). How does large-N trade-off between len(G) and average(G) behave? (And how clever do we have to be to get a good trade-off?) Note: len(ST(x (N) )) is no more than O(N) (Steele 1997, §2.2); Introduction The Poissonian City Variance and efficiency Flows Conclusion References Questions Aldous and Kendall (2008) provide answers for the First Question √ Consider a configuration x (N) of N cities in [0, N]2 as above, and a well-chosen connecting network G = G(x (N) ). How does large-N trade-off between len(G) and average(G) behave? (And how clever do we have to be to get a good trade-off?) Note: len(ST(x (N) )) is no more than O(N) (Steele 1997, §2.2); Average Euclidean distance √ between two randomly chosen cities is at most 2N; Introduction The Poissonian City Variance and efficiency Flows Conclusion References Questions Aldous and Kendall (2008) provide answers for the First Question √ Consider a configuration x (N) of N cities in [0, N]2 as above, and a well-chosen connecting network G = G(x (N) ). How does large-N trade-off between len(G) and average(G) behave? (And how clever do we have to be to get a good trade-off?) Note: len(ST(x (N) )) is no more than O(N) (Steele 1997, §2.2); Average Euclidean distance √ between two randomly chosen cities is at most 2N; Perhaps increasing total network length by const × N α might achieve average network distance no more than order N β longer than average Euclidean distance? Introduction The Poissonian City Variance and efficiency Flows Conclusion References Further Questions Today I focus on answers to yet further questions including: Question about fluctuations Given a good compromise between average(G) and len(G), how might the variance behave? Introduction The Poissonian City Variance and efficiency Flows Conclusion References Further Questions Today I focus on answers to yet further questions including: Question about fluctuations Given a good compromise between average(G) and len(G), how might the variance behave? Question about true geodesics The upper bound is obtained by controlling non-geodesic paths. How might true geodesics behave? Introduction The Poissonian City Variance and efficiency Flows Conclusion References Further Questions Today I focus on answers to yet further questions including: Question about fluctuations Given a good compromise between average(G) and len(G), how might the variance behave? Question about true geodesics The upper bound is obtained by controlling non-geodesic paths. How might true geodesics behave? Question about flows Consider a network which exhibits good trade-offs. What can be said about flows in this network? Introduction The Poissonian City Variance and efficiency Flows Conclusion References Answer to first question (I) Augment Steiner tree by a low-intensity invariant Poisson line process Π. Introduction The Poissonian City Variance and efficiency Flows Conclusion References Answer to first question (I) Augment Steiner tree by a low-intensity invariant Poisson line process Π. Unit intensity is 1 2 d r d θ: we will use this and scale. Introduction The Poissonian City Variance and efficiency Flows Conclusion References Answer to first question (I) Augment Steiner tree by a low-intensity invariant Poisson line process Π. 1 2 d r d θ: we will use this and scale. √ Pick two cities x and y at distance n = N units apart. Unit intensity is Introduction The Poissonian City Variance and efficiency Flows Conclusion References Answer to first question (I) Augment Steiner tree by a low-intensity invariant Poisson line process Π. 1 2 d r d θ: we will use this and scale. √ Pick two cities x and y at distance n = N units apart. Unit intensity is Remove lines separating the two cities; Introduction The Poissonian City Variance and efficiency Flows Conclusion References Answer to first question (I) Augment Steiner tree by a low-intensity invariant Poisson line process Π. 1 2 d r d θ: we will use this and scale. √ Pick two cities x and y at distance n = N units apart. Unit intensity is Remove lines separating the two cities; focus on cell Cx,y containing the two cities. Introduction The Poissonian City Variance and efficiency Flows Conclusion Answer to first question (II) Upper-bound “network distance” between two cities by References Introduction The Poissonian City Variance and efficiency Flows Conclusion Answer to first question (II) Upper-bound “network distance” between two cities by mean semi-perimeter of cell, i 1 h 4 5 E len ∂Cx,y = n+ log n + γ + + ... . 2 3 8 References Introduction The Poissonian City Variance and efficiency Flows Conclusion References Answer to first question (II) Upper-bound “network distance” between two cities by mean semi-perimeter of cell, i 1 h 4 5 E len ∂Cx,y = n+ log n + γ + + ... . 2 3 8 Aldous and Kendall (2008) apply this to resolve our First Question, and use other methods from stochastic geometry to show that the resolution is nearly optimal. Introduction The Poissonian City Variance and efficiency Flows Conclusion References Links to random metric spaces The study of the metric space generated by the line process forms a chapter in the theory of random metric spaces: Introduction The Poissonian City Variance and efficiency Flows Conclusion References Links to random metric spaces The study of the metric space generated by the line process forms a chapter in the theory of random metric spaces: Vershik (2004) builds random metric spaces out of random distance matrices (compare MDS in statistics); almost all such metric spaces are isometric to Urysohn’s celebrated universal metric space. Introduction The Poissonian City Variance and efficiency Flows Conclusion References Links to random metric spaces The study of the metric space generated by the line process forms a chapter in the theory of random metric spaces: Vershik (2004) builds random metric spaces out of random distance matrices (compare MDS in statistics); almost all such metric spaces are isometric to Urysohn’s celebrated universal metric space. But these spaces are definitely not finite-dimensional! Introduction The Poissonian City Variance and efficiency Flows Conclusion References Links to random metric spaces The study of the metric space generated by the line process forms a chapter in the theory of random metric spaces: Vershik (2004) builds random metric spaces out of random distance matrices (compare MDS in statistics); almost all such metric spaces are isometric to Urysohn’s celebrated universal metric space. But these spaces are definitely not finite-dimensional! The Brownian map has been introduced as the limit of random quadrangulations of the 2-sphere (for example, Le Gall 2009). Introduction The Poissonian City Variance and efficiency Flows Conclusion References Links to random metric spaces The study of the metric space generated by the line process forms a chapter in the theory of random metric spaces: Vershik (2004) builds random metric spaces out of random distance matrices (compare MDS in statistics); almost all such metric spaces are isometric to Urysohn’s celebrated universal metric space. But these spaces are definitely not finite-dimensional! The Brownian map has been introduced as the limit of random quadrangulations of the 2-sphere (for example, Le Gall 2009). But these spaces are definitely not flat! Introduction The Poissonian City Variance and efficiency Flows Conclusion References Links to random metric spaces The study of the metric space generated by the line process forms a chapter in the theory of random metric spaces: Vershik (2004) builds random metric spaces out of random distance matrices (compare MDS in statistics); almost all such metric spaces are isometric to Urysohn’s celebrated universal metric space. But these spaces are definitely not finite-dimensional! The Brownian map has been introduced as the limit of random quadrangulations of the 2-sphere (for example, Le Gall 2009). But these spaces are definitely not flat! Baccelli, Tchoumatchenko, and Zuyev (2000) link to 4 geometric spanners; they exhibit π -spanner paths in Poisson-Delaunay triangulations. Introduction The Poissonian City Variance and efficiency Flows Conclusion References Links to random metric spaces The study of the metric space generated by the line process forms a chapter in the theory of random metric spaces: Vershik (2004) builds random metric spaces out of random distance matrices (compare MDS in statistics); almost all such metric spaces are isometric to Urysohn’s celebrated universal metric space. But these spaces are definitely not finite-dimensional! The Brownian map has been introduced as the limit of random quadrangulations of the 2-sphere (for example, Le Gall 2009). But these spaces are definitely not flat! Baccelli, Tchoumatchenko, and Zuyev (2000) link to 4 geometric spanners; they exhibit π -spanner paths in Poisson-Delaunay triangulations. Famous conjecture (late 1940’s) by D. G. Kendall: large cells in Poisson line process tessellation are nearly circular. Introduction The Poissonian City Variance and efficiency Flows Conclusion References Links to random metric spaces The study of the metric space generated by the line process forms a chapter in the theory of random metric spaces: Vershik (2004) builds random metric spaces out of random distance matrices (compare MDS in statistics); almost all such metric spaces are isometric to Urysohn’s celebrated universal metric space. But these spaces are definitely not finite-dimensional! The Brownian map has been introduced as the limit of random quadrangulations of the 2-sphere (for example, Le Gall 2009). But these spaces are definitely not flat! Baccelli, Tchoumatchenko, and Zuyev (2000) link to 4 geometric spanners; they exhibit π -spanner paths in Poisson-Delaunay triangulations. Famous conjecture (late 1940’s) by D. G. Kendall: large cells in Poisson line process tessellation are nearly circular. Now known to be true (Miles, Kovalenko). Introduction The Poissonian City Variance and efficiency Flows Conclusion Introducing the Poissonian city Consider an idealized network in a disk of radius n: References Introduction The Poissonian City Variance and efficiency Flows Conclusion Introducing the Poissonian city Consider an idealized network in a disk of radius n: long-range connections use Poisson line process; References Introduction The Poissonian City Variance and efficiency Flows Conclusion Introducing the Poissonian city Consider an idealized network in a disk of radius n: long-range connections use Poisson line process; connect points x and y thus: References Introduction The Poissonian City Variance and efficiency Flows Conclusion Introducing the Poissonian city Consider an idealized network in a disk of radius n: long-range connections use Poisson line process; connect points x and y thus: proceed from x in opposite direction to y, References Introduction The Poissonian City Variance and efficiency Flows Conclusion Introducing the Poissonian city Consider an idealized network in a disk of radius n: long-range connections use Poisson line process; connect points x and y thus: proceed from x in opposite direction to y, till one hits the line process, References Introduction The Poissonian City Variance and efficiency Flows Conclusion Introducing the Poissonian city Consider an idealized network in a disk of radius n: long-range connections use Poisson line process; connect points x and y thus: proceed from x in opposite direction to y, till one hits the line process, proceed clockwise or anti-clockwise round cell formed by lines not separating x from y, References Introduction The Poissonian City Variance and efficiency Flows Conclusion Introducing the Poissonian city Consider an idealized network in a disk of radius n: long-range connections use Poisson line process; connect points x and y thus: proceed from x in opposite direction to y, till one hits the line process, proceed clockwise or anti-clockwise round cell formed by lines not separating x from y, till y occludes x, References Introduction The Poissonian City Variance and efficiency Flows Conclusion Introducing the Poissonian city Consider an idealized network in a disk of radius n: long-range connections use Poisson line process; connect points x and y thus: proceed from x in opposite direction to y, till one hits the line process, proceed clockwise or anti-clockwise round cell formed by lines not separating x from y, till y occludes x, then proceed directly to y. References Introduction The Poissonian City Variance and efficiency Flows Conclusion Introducing the Poissonian city Consider an idealized network in a disk of radius n: long-range connections use Poisson line process; connect points x and y thus: proceed from x in opposite direction to y, till one hits the line process, proceed clockwise or anti-clockwise round cell formed by lines not separating x from y, till y occludes x, then proceed directly to y. Thus we can connect all pairs of points in the disk. References Introduction The Poissonian City Variance and efficiency Flows Conclusion References Density of point of maximal y-coordinate Density / intensity Using u = 2n x ∈ (−1, 1) and v = √y n > 0, 1 1 (sin β + sin γ − sin(β + γ)) exp − 2 (η − n) d x d y 4 ! v3 v2 ≈ du dv 2 exp − 1 − u2 1 − u2 Introduction The Poissonian City Variance and efficiency Flows Conclusion References Density of point of maximal y-coordinate Density / intensity Using u = 2n x ∈ (−1, 1) and v = √y n > 0, 1 1 (sin β + sin γ − sin(β + γ)) exp − 2 (η − n) d x d y 4 ! v3 v2 ≈ du dv 2 exp − 1 − u2 1 − u2 ASYMPTOTICALLY Location of maximum is uniformly distributed; Introduction The Poissonian City Variance and efficiency Flows Conclusion References Density of point of maximal y-coordinate Density / intensity Using u = 2n x ∈ (−1, 1) and v = √y n > 0, 1 1 (sin β + sin γ − sin(β + γ)) exp − 2 (η − n) d x d y 4 ! v3 v2 ≈ du dv 2 exp − 1 − u2 1 − u2 ASYMPTOTICALLY Location of maximum is uniformly distributed; Conditional height of maximum is length of Gaussian 4-vector. Introduction The Poissonian City Variance and efficiency Flows Conclusion Variance and growth process (I) Consider cell boundary as path with x-coordinate Xτ , parametrized by τ (excess over geodesic distance). References Introduction The Poissonian City Variance and efficiency Flows Conclusion Variance and growth process (I) Consider cell boundary as path with x-coordinate Xτ , parametrized by τ (excess over geodesic distance). Let Θ be angle with positive x-axis; References Introduction The Poissonian City Variance and efficiency Flows Conclusion Variance and growth process (I) Consider cell boundary as path with x-coordinate Xτ , parametrized by τ (excess over geodesic distance). Let Θ be angle with positive x-axis; 1 Θ jumps at Poisson point process intensity 2 ; References Introduction The Poissonian City Variance and efficiency Flows Conclusion Variance and growth process (I) Consider cell boundary as path with x-coordinate Xτ , parametrized by τ (excess over geodesic distance). Let Θ be angle with positive x-axis; 1 Θ jumps at Poisson point process intensity 2 ; A jump −∆Θ = Θ − Θ− obeys P [Θ− − Θ ≤ θ | Θ− ] = 1 − cos θ . 1 − cos Θ− References Introduction The Poissonian City Variance and efficiency Flows Conclusion Variance and growth process (II) Define σ (n) by Xσ (n) = n. References Introduction The Poissonian City Variance and efficiency Flows Conclusion References Variance and growth process (II) Define σ (n) by Xσ (n) = n. Then Mτ = 12 Xτ − 34 τ is almost a martingale, and Xσ (n) is almost a self-similar Markov process. Introduction The Poissonian City Variance and efficiency Flows Conclusion References Variance and growth process (II) Define σ (n) by Xσ (n) = n. Then Mτ = 12 Xτ − 34 τ is almost a martingale, and Xσ (n) is almost a self-similar Markov process. Hence (tautologically!) σ (n) = Θ2 Xσ (n) 2 log n + log 0 + log − 2Mσ (n) 3 2 n Introduction The Poissonian City Variance and efficiency Flows Conclusion References Variance and growth process (II) Define σ (n) by Xσ (n) = n. Then Mτ = 12 Xτ − 34 τ is almost a martingale, and Xσ (n) is almost a self-similar Markov process. Hence σ (n) = Dispose of log integration; Xσ (n) 2 log n + O(1) + log − 2Mσ (n) 3 n Θ02 2 contribution to second moment by Introduction The Poissonian City Variance and efficiency Flows Conclusion References Variance and growth process (II) Define σ (n) by Xσ (n) = n. Then Mτ = 12 Xτ − 34 τ is almost a martingale, and Xσ (n) is almost a self-similar Markov process. Hence σ (n) = X 2 log n + O(1) − 2Mσ (n) 3 Dispose of log σn(n) contribution to second moment using Lamperti transformation and work of Bertoin and Yor (2005); Introduction The Poissonian City Variance and efficiency Flows Conclusion References Variance and growth process (II) Define σ (n) by Xσ (n) = n. Then Mτ = 12 Xτ − 34 τ is almost a martingale, and Xσ (n) is almost a self-similar Markov process. Hence σ (n) = 2 log n + O(1) − 2Mσ (n) 3 Mσ (n) derives from a uniformly integrable L2 martingale; Introduction The Poissonian City Variance and efficiency Flows Conclusion References Variance and growth process (II) Define σ (n) by Xσ (n) = n. Then Mτ = 12 Xτ − 34 τ is almost a martingale, and Xσ (n) is almost a self-similar Markov process. Hence σ (n) = 2 log n + O(1) − 2Mσ (n) 3 and thus E [σ (n)] = Var [σ (n)] = 2 log n + O(1) 3 q 20 log n + O log n . 27 Introduction The Poissonian City Variance and efficiency Flows Conclusion References Variance and growth process (II) Define σ (n) by Xσ (n) = n. Then Mτ = 12 Xτ − 34 τ is almost a martingale, and Xσ (n) is almost a self-similar Markov process. Hence σ (n) = 2 log n + O(1) − 2Mσ (n) 3 and thus E [σ (n)] = Var [σ (n)] = 2 log n + O(1) 3 q 20 log n + O log n . 27 We pdeduce that perimeter length fluctuations are O log n . Introduction The Poissonian City Variance and efficiency Flows Conclusion References What about true geodesics? The above fluctuation theory shows that true geodesics typically have smaller excess than our paths; Introduction The Poissonian City Variance and efficiency Flows Conclusion References What about true geodesics? The above fluctuation theory shows that true geodesics typically have smaller excess than our paths; however Introduction The Poissonian City Variance and efficiency Flows Conclusion References What about true geodesics? The above fluctuation theory shows that true geodesics typically have smaller excess than our paths; however the number of “top-to-bottom” crossings of a true geodesic is stochastically bounded in any region [na, nb] ⊆ (0, n), so all but a stochastically bounded number of “short-cuts” must be within O(n/ log n) of start or end, affecting coefficient of log(n) but no more; Introduction The Poissonian City Variance and efficiency Flows Conclusion References What about true geodesics? The above fluctuation theory shows that true geodesics typically have smaller excess than our paths; however the number of “top-to-bottom” crossings of a true geodesic is stochastically bounded in any region [na, nb] ⊆ (0, n), so all but a stochastically bounded number of “short-cuts” must be within O(n/ log n) of start or end, affecting coefficient of log(n) but no more; indeed Introduction The Poissonian City Variance and efficiency Flows Conclusion References What about true geodesics? The above fluctuation theory shows that true geodesics typically have smaller excess than our paths; however the number of “top-to-bottom” crossings of a true geodesic is stochastically bounded in any region [na, nb] ⊆ (0, n), so all but a stochastically bounded number of “short-cuts” must be within O(n/ log n) of start or end, affecting coefficient of log(n) but no more; indeed we can prove that any path of length n built using the Poisson lines must have mean excess at least 5 log 4 − log n + o(log n) . 4 Introduction The Poissonian City Variance and efficiency Flows Conclusion How much better can a true geodesic be? Methods: References Introduction The Poissonian City Variance and efficiency Flows Conclusion References How much better can a true geodesic be? Methods: Z n E (sec θx − 1) d x ≥ 1 1 2 Zn 1 h i E θx2 d x Introduction The Poissonian City Variance and efficiency Flows Conclusion References How much better can a true geodesic be? Methods: Z n E (sec θx − 1) d x ≥ 1 P [|θx | ≥ u] ≥ 2 E exp(−uLx+ ) 1 2 Zn 1 h i E θx2 d x Introduction The Poissonian City Variance and efficiency Flows Conclusion References How much better can a true geodesic be? Methods: Z n E (sec θx − 1) d x ≥ 1 P [|θx | ≥ u] 2 q p2 + 4 + p ≤ ep 2 /2 ≥ Z∞ p 1 2 Zn 1 h i E θx2 d x 2 E exp(−uLx+ ) e−s 2 /2 ds ≤ 4 q p2 + 8 + 3p Introduction The Poissonian City Variance and efficiency Flows Conclusion References How much better can a true geodesic be? Methods: Z n E (sec θx − 1) d x ≥ 1 P [|θx | ≥ u] 2 q p2 + 4 + p Birnbaum (1942) ≤ ep 2 /2 ≥ Z∞ p 1 2 Zn 1 h i E θx2 d x 2 E exp(−uLx+ ) e−s 2 /2 ds ≤ 4 q p2 + 8 + 3p Sampford (1953) Introduction The Poissonian City Variance and efficiency Flows Conclusion Flow in a model network (I) Consider an idealized network in a disk of radius n, conditioned to have a line passing through the origin: Consider the region in 4-space determined by pairs of points for which such a connection passes through o. References Introduction The Poissonian City Variance and efficiency Flows Conclusion Flow in a model network (I) Consider an idealized network in a disk of radius n, conditioned to have a line passing through the origin: Consider the region in 4-space determined by pairs of points for which such a connection passes through o. What is the limiting distribution of the 4-volume of this region as n → ∞? References Introduction The Poissonian City Variance and efficiency Flows Conclusion Flow in a model network (II) Mean value of Vn is asymptotically 2n3 . References Introduction The Poissonian City Variance and efficiency Flows Conclusion Flow in a model network (II) Mean value of Vn is asymptotically 2n3 . Dominant contribution comes from opposing pairs of points nearly aligned with conditioned line ` (angle of order about √1n ) . References Introduction The Poissonian City Variance and efficiency Flows Conclusion References Flow in a model network (II) Mean value of Vn is asymptotically 2n3 . Dominant contribution comes from opposing pairs of points nearly aligned with conditioned line ` (angle of order about √1n ) . Very heavy variance calculations show, second moment of Vn /n3 is uniformly bounded ("no congestion"). Introduction The Poissonian City Variance and efficiency Flows Conclusion References Flow in a model network (II) Mean value of Vn is asymptotically 2n3 . Dominant contribution comes from opposing pairs of points nearly aligned with conditioned line ` (angle of order about √1n ) . Very heavy variance calculations show, second moment of Vn /n3 is uniformly bounded ("no congestion"). In fact a limiting construction, involving a network based on an improper anisotropic line process, generates a non-trivial limiting distribution for Vn /n3 . Introduction The Poissonian City Variance and efficiency Flows Conclusion References Improper Anisotropic Line Process (I) Limiting process: scale x-axis by n and y-axis by resulting improper line process: √ n. The has an improper rose of directions (there is a singularity at vertical directions); Introduction The Poissonian City Variance and efficiency Flows Conclusion References Improper Anisotropic Line Process (I) Limiting process: scale x-axis by n and y-axis by resulting improper line process: √ n. The has an improper rose of directions (there is a singularity at vertical directions); possesses a special affine symmetry group; Introduction The Poissonian City Variance and efficiency Flows Conclusion References Improper Anisotropic Line Process (I) Limiting process: scale x-axis by n and y-axis by resulting improper line process: √ n. The has an improper rose of directions (there is a singularity at vertical directions); possesses a special affine symmetry group; can be easily shown to yield non-degenerate flow volume. Introduction The Poissonian City Variance and efficiency Flows Conclusion Improper Anisotropic Line Process (II) References Introduction The Poissonian City Variance and efficiency Flows Conclusion Improper Anisotropic Line Process (II) References Introduction The Poissonian City Variance and efficiency Flows Conclusion Improper Anisotropic Line Process (II) References Introduction The Poissonian City Variance and efficiency Flows Conclusion Improper Anisotropic Line Process (II) References Introduction The Poissonian City Variance and efficiency Flows Conclusion Improper Anisotropic Line Process (II) References Introduction The Poissonian City Variance and efficiency Flows Conclusion Improper Anisotropic Line Process (II) Open questions: References Introduction The Poissonian City Variance and efficiency Flows Conclusion References Improper Anisotropic Line Process (II) Open questions: analytical characterization of flow volume distribution? Introduction The Poissonian City Variance and efficiency Flows Conclusion References Improper Anisotropic Line Process (II) Open questions: analytical characterization of flow volume distribution? good simulation methodology? Introduction The Poissonian City Variance and efficiency Flows What I don’t know yet Computing User Equilibrium for flows. Conclusion References Introduction The Poissonian City Variance and efficiency Flows Conclusion What I don’t know yet Computing User Equilibrium for flows. We can in principle calculate asymptotic mean flow at any particular point in the disk, and hence compute User / Wardrop / Nash Equilibria so as to minimize objective including distance travelled and traffic encountered if References Introduction The Poissonian City Variance and efficiency Flows Conclusion What I don’t know yet Computing User Equilibrium for flows. We can in principle calculate asymptotic mean flow at any particular point in the disk, and hence compute User / Wardrop / Nash Equilibria so as to minimize objective including distance travelled and traffic encountered if users can choose which of two routes (left or right) to travel; References Introduction The Poissonian City Variance and efficiency Flows Conclusion What I don’t know yet Computing User Equilibrium for flows. We can in principle calculate asymptotic mean flow at any particular point in the disk, and hence compute User / Wardrop / Nash Equilibria so as to minimize objective including distance travelled and traffic encountered if users can choose which of two routes (left or right) to travel; or users can choose to use only some of the available lines (thinning the Poisson line process). References Introduction The Poissonian City Variance and efficiency Flows Conclusion What I don’t know yet Computing User Equilibrium for flows. We can in principle calculate asymptotic mean flow at any particular point in the disk, and hence compute User / Wardrop / Nash Equilibria so as to minimize objective including distance travelled and traffic encountered if users can choose which of two routes (left or right) to travel; or users can choose to use only some of the available lines (thinning the Poisson line process). Moving away from lines. References Introduction The Poissonian City Variance and efficiency Flows Conclusion References What I don’t know yet Computing User Equilibrium for flows. We can in principle calculate asymptotic mean flow at any particular point in the disk, and hence compute User / Wardrop / Nash Equilibria so as to minimize objective including distance travelled and traffic encountered if users can choose which of two routes (left or right) to travel; or users can choose to use only some of the available lines (thinning the Poisson line process). Moving away from lines. results will not change qualitatively if lines replaced by long segments. How long? Introduction The Poissonian City Variance and efficiency Flows Conclusion References What I don’t know yet Computing User Equilibrium for flows. We can in principle calculate asymptotic mean flow at any particular point in the disk, and hence compute User / Wardrop / Nash Equilibria so as to minimize objective including distance travelled and traffic encountered if users can choose which of two routes (left or right) to travel; or users can choose to use only some of the available lines (thinning the Poisson line process). Moving away from lines. results will not change qualitatively if lines replaced by long segments. How long? results will not change qualitatively if line segments replaced by curves of low curvature. How low? Introduction The Poissonian City Variance and efficiency Flows Conclusion Aldous and Kendall (2008) showed that Conclusion References Introduction The Poissonian City Variance and efficiency Flows Conclusion Conclusion Aldous and Kendall (2008) showed that √ the “N cities in [0, N]2 ” connection problem can be resolved using a Poisson line process to gain nearly Euclidean efficiency at negligible cost; References Introduction The Poissonian City Variance and efficiency Flows Conclusion Conclusion Aldous and Kendall (2008) showed that √ the “N cities in [0, N]2 ” connection problem can be resolved using a Poisson line process to gain nearly Euclidean efficiency at negligible cost; conversely any configuration which is not too concentrated cannot be treated much more efficiently. References Introduction The Poissonian City Variance and efficiency Flows Conclusion Conclusion Aldous and Kendall (2008) showed that √ the “N cities in [0, N]2 ” connection problem can be resolved using a Poisson line process to gain nearly Euclidean efficiency at negligible cost; conversely any configuration which is not too concentrated cannot be treated much more efficiently. (and Poisson line processes are not computationally hard!) References Introduction The Poissonian City Variance and efficiency Flows Conclusion Conclusion Aldous and Kendall (2008) showed that √ the “N cities in [0, N]2 ” connection problem can be resolved using a Poisson line process to gain nearly Euclidean efficiency at negligible cost; conversely any configuration which is not too concentrated cannot be treated much more efficiently. (and Poisson line processes are not computationally hard!) Random variation of network distance is controlled. References Introduction The Poissonian City Variance and efficiency Flows Conclusion Conclusion Aldous and Kendall (2008) showed that √ the “N cities in [0, N]2 ” connection problem can be resolved using a Poisson line process to gain nearly Euclidean efficiency at negligible cost; conversely any configuration which is not too concentrated cannot be treated much more efficiently. (and Poisson line processes are not computationally hard!) Random variation of network distance is controlled. “Near-geodesics” are pretty good. References Introduction The Poissonian City Variance and efficiency Flows Conclusion Conclusion Aldous and Kendall (2008) showed that √ the “N cities in [0, N]2 ” connection problem can be resolved using a Poisson line process to gain nearly Euclidean efficiency at negligible cost; conversely any configuration which is not too concentrated cannot be treated much more efficiently. (and Poisson line processes are not computationally hard!) Random variation of network distance is controlled. “Near-geodesics” are pretty good. Traffic flow in the network scales well. References Introduction The Poissonian City Variance and efficiency Flows Conclusion Conclusion Aldous and Kendall (2008) showed that √ the “N cities in [0, N]2 ” connection problem can be resolved using a Poisson line process to gain nearly Euclidean efficiency at negligible cost; conversely any configuration which is not too concentrated cannot be treated much more efficiently. (and Poisson line processes are not computationally hard!) Random variation of network distance is controlled. “Near-geodesics” are pretty good. Traffic flow in the network scales well. User equilibrium? Line segments or curves? References Introduction The Poissonian City Variance and efficiency Flows Conclusion Conclusion Aldous and Kendall (2008) showed that √ the “N cities in [0, N]2 ” connection problem can be resolved using a Poisson line process to gain nearly Euclidean efficiency at negligible cost; conversely any configuration which is not too concentrated cannot be treated much more efficiently. (and Poisson line processes are not computationally hard!) Random variation of network distance is controlled. “Near-geodesics” are pretty good. Traffic flow in the network scales well. User equilibrium? Line segments or curves? Same problem in 3-space or higher dimensions? References Introduction The Poissonian City Variance and efficiency Flows Conclusion Conclusion Aldous and Kendall (2008) showed that √ the “N cities in [0, N]2 ” connection problem can be resolved using a Poisson line process to gain nearly Euclidean efficiency at negligible cost; conversely any configuration which is not too concentrated cannot be treated much more efficiently. (and Poisson line processes are not computationally hard!) Random variation of network distance is controlled. “Near-geodesics” are pretty good. Traffic flow in the network scales well. User equilibrium? Line segments or curves? Same problem in 3-space or higher dimensions? QUESTIONS? References Introduction The Poissonian City Variance and efficiency Flows Conclusion References Bibliography This is a rich hypertext bibliography. Aldous, D. J. and W. S. Kendall (2008, March). Short-length routes in low-cost networks via Poisson line patterns. Advances in Applied Probability 40(1), 1–21, , and http://arxiv.org/abs/math.PR/0701140 . Baccelli, F., K. Tchoumatchenko, and S. Zuyev (2000). Markov paths on the Poisson-Delaunay graph with applications to routing in mobile networks. Advances in Applied Probability 32(1), 1–18. Bertoin, J. and M. Yor (2005). Exponential functionals of Lévy processes. Probability Surveys 2, 191–212 (electronic), . Birnbaum, Z. W. (1942). An inequality for Mill’s ratio. Annals of Mathematical Statistics 13, 245–246. Le Gall, J.-F. (2009). Geodesics in large planar maps and in the Brownian map. Acta Mathematica to appear. Introduction The Poissonian City Variance and efficiency Flows Conclusion Sampford, M. R. (1953). Some inequalities on Mill’s ratio and related functions. Annals of Mathematical Statistics 24, 130–132. Steele, J. M. (1997). Probability theory and combinatorial optimization, Volume 69 of CBMS-NSF Regional Conference Series in Applied Mathematics. Philadelphia, PA: Society for Industrial and Applied Mathematics (SIAM). Stoyan, D., W. S. Kendall, and J. Mecke (1995). Stochastic geometry and its applications (Second ed.). Chichester: John Wiley & Sons. (First edition in 1987 joint with Akademie Verlag, Berlin). Vershik, A. M. (2004). Random and universal metric spaces. In Dynamics and randomness II, Volume 10 of Nonlinear Phenom. Complex Systems, pp. 199–228. Dordrecht: Kluwer Acad. Publ. References