Loading [MathJax]/jax/output/SVG/jax.js
Research article

On the existence and stability of minimizers for generalized Tikhonov functionals with general similarity data

  • Received: 21 July 2020 Accepted: 29 November 2020 Published: 06 January 2021
  • MSC : 47J06, 65J20, 49J27

  • We study a unified form of Tikhonov regularization for ill-posed problems with a general data similarity term. We discuss sufficient conditions on this generalized Tikhonov functional that guarantee existence and stability of solutions. Furthermore, we show that some particular cases of similarity functionals and regularization techniques can be cast into a unified theoretical framework. In particular, we consider the cases of p-power norm, Bregman distance and mutual information as examples of a data similarity term, and Tikhonov regularization of order one or using a Sobolev norm, total variation penalization and powers of semi-norms associated to closed operators as examples of a regularization term. Finally, we take up the case of the image registration problem.

    Citation: El Mostafa Kalmoun, Fatimah Allami. On the existence and stability of minimizers for generalized Tikhonov functionals with general similarity data[J]. AIMS Mathematics, 2021, 6(3): 2764-2777. doi: 10.3934/math.2021169

    Related Papers:

    [1] Paola Goatin, Chiara Daini, Maria Laura Delle Monache, Antonella Ferrara . Interacting moving bottlenecks in traffic flow. Networks and Heterogeneous Media, 2023, 18(2): 930-945. doi: 10.3934/nhm.2023040
    [2] Felisia Angela Chiarello, Paola Goatin . Non-local multi-class traffic flow models. Networks and Heterogeneous Media, 2019, 14(2): 371-387. doi: 10.3934/nhm.2019015
    [3] Jan Friedrich, Oliver Kolb, Simone Göttlich . A Godunov type scheme for a class of LWR traffic flow models with non-local flux. Networks and Heterogeneous Media, 2018, 13(4): 531-547. doi: 10.3934/nhm.2018024
    [4] Abraham Sylla . Influence of a slow moving vehicle on traffic: Well-posedness and approximation for a mildly nonlocal model. Networks and Heterogeneous Media, 2021, 16(2): 221-256. doi: 10.3934/nhm.2021005
    [5] Christophe Chalons, Paola Goatin, Nicolas Seguin . General constrained conservation laws. Application to pedestrian flow modeling. Networks and Heterogeneous Media, 2013, 8(2): 433-463. doi: 10.3934/nhm.2013.8.433
    [6] Caterina Balzotti, Simone Göttlich . A two-dimensional multi-class traffic flow model. Networks and Heterogeneous Media, 2021, 16(1): 69-90. doi: 10.3934/nhm.2020034
    [7] Dong Li, Tong Li . Shock formation in a traffic flow model with Arrhenius look-ahead dynamics. Networks and Heterogeneous Media, 2011, 6(4): 681-694. doi: 10.3934/nhm.2011.6.681
    [8] Raimund Bürger, Kenneth H. Karlsen, John D. Towers . On some difference schemes and entropy conditions for a class of multi-species kinematic flow models with discontinuous flux. Networks and Heterogeneous Media, 2010, 5(3): 461-485. doi: 10.3934/nhm.2010.5.461
    [9] Raimund Bürger, Christophe Chalons, Rafael Ordoñez, Luis Miguel Villada . A multiclass Lighthill-Whitham-Richards traffic model with a discontinuous velocity function. Networks and Heterogeneous Media, 2021, 16(2): 187-219. doi: 10.3934/nhm.2021004
    [10] Alexander Kurganov, Anthony Polizzi . Non-oscillatory central schemes for traffic flow models with Arrhenius look-ahead dynamics. Networks and Heterogeneous Media, 2009, 4(3): 431-451. doi: 10.3934/nhm.2009.4.431
  • We study a unified form of Tikhonov regularization for ill-posed problems with a general data similarity term. We discuss sufficient conditions on this generalized Tikhonov functional that guarantee existence and stability of solutions. Furthermore, we show that some particular cases of similarity functionals and regularization techniques can be cast into a unified theoretical framework. In particular, we consider the cases of p-power norm, Bregman distance and mutual information as examples of a data similarity term, and Tikhonov regularization of order one or using a Sobolev norm, total variation penalization and powers of semi-norms associated to closed operators as examples of a regularization term. Finally, we take up the case of the image registration problem.



    Macroscopic traffic flow models based on fluid-dynamics equations have been introduced in the transport engineering literature since the mid-fifties of last century, with the celebrated Lighthill, Whitham [11] and Richards [13] (LWR) model. Since then, the engineering and applied mathematical literature on the subject has considerably grown, addressing the need for more sophisticated models better capturing traffic flow characteristics. Indeed, the LWR model is based on the assumption that the mean traffic speed is a function of the traffic density, which is not experimentally verified in congested regimes. To overcome this issue, the so-called "second order" models (e.g. Payne-Whitham [12,15] and Aw-Rascle-Zhang [3,16]) consist of a mass conservation equation for the density and an acceleration balance law for the speed, thus considering the two quantities as independent.

    More recently, "non-local" versions of the LWR model have been proposed in [5,14], where the speed function depends on a weighted mean of the downstream vehicle density to better represent the reaction of drivers to downstream traffic conditions.

    Another limitation of the standard LWR model is the first-in first-out rule, not allowing faster vehicles to overtake slower ones. To address this and other traffic heterogeneities, "multi-class" models consist of a system of conservation equations, one for each vehicle class, coupled in the speed terms, see [4] and references therein for more details.

    In this paper, we consider the following class of non-local systems of M conservation laws in one space dimension:

    tρi(t,x)+x(ρi(t,x)vi((rωi)(t,x)))=0,i=1,...,M, (1)

    where

    r(t,x):=Mi=1ρi(t,x), (2)
    vi(ξ):=vmaxiψ(ξ), (3)
    (rωi)(t,x):=x+ηixr(t,y)ωi(yx)dy, (4)

    and we assume:

    (H1) The convolution kernels ωiC1([0,ηi];R+), ηi>0, are non-increasing functions such that ηi0ωi(y)dy=Ji. We set W0:=maxi=1,,Mωi(0).

    (H2) vmaxi are the maximal velocities, with 0<vmax1vmax2vmaxM.

    (H3) ψ:R+R+ is a smooth non-increasing function such that ψ(0)=1 and ψ(r)=0 for r1 (for simplicity, we can consider the function ψ(r)=max{1r,0}).

    We couple (1) with an initial datum

    ρi(0,x)=ρ0i(x),i=1,,M. (5)

    Model (1) is obtained generalizing the n-populations model for traffic flow described in [4] and it is a multi-class version of the one dimensional scalar conservation law with non-local flux proposed in [5], where ρi is the density of vehicles belonging to the i-th class, ηi is proportional to the look-ahead distance and Ji is the interaction strength. In our setting, the non-local dependence of the speed functions vi describes the reaction of drivers that adapt their velocity to the downstream traffic, assigning greater importance to closer vehicles, see also [7,9]. We allow different anisotropic kernels for each equation of the system. The model takes into account the distribution of heterogeneous drivers and vehicles characterized by their maximal speeds and look-ahead visibility in a traffic stream.

    Due to the possible presence of jump discontinuities, solutions to (1), (5) are intended in the following weak sense.

    Definition 1.1. A function ρ=(ρ1,,ρM)(L1L)([0,T[×R;RM), T>0, is a weak solution of (1), (5) if

    T0(ρitφ+ρivi(rωi)xφ)(t,x)dxdt+ρ0i(x)φ(0,x)dx=0

    for all φC1c(],T[×R;R), i=1,,M.

    The main result of this paper is the proof of existence of weak solutions to (1), (5), locally in time. We remark that, since the convolution kernels ωi are not smooth on R, the results in [1] cannot be applied due to the lack of L-bounds on their derivatives.

    Theorem 1.2. Let ρ0i(x)(BVL)(R;R+), for i=1,,M, and assumptions (H1) - (H3) hold. Then the Cauchy problem (1), (5) admits a weak solution on [0,T[×R, for some T>0 sufficiently small.

    In this work, we do not address the question of uniqueness of the solutions to (1). Indeed, even if discrete entropy inequalities can be derived as in [5,Proposition 3], in the case of systems this is in general not sufficient to single out a unique solution.

    The paper is organized as follows. Section 2 is devoted to prove uniform L and BV estimates on the approximate solutions obtained through an approximation argument based on a Godunov type numerical scheme, see [8]. We have to point out that these estimates heavily rely on the monotonicity properties of the kernel functions ωi. In Section 3 we prove the existence in finite time of weak solutions applying Helly's theorem and a Lax-Wendroff type argument, see [10]. In Section 4 we present some numerical simulations for M=2. In particular, we consider the case of a mixed flow of cars and trucks on a stretch of road, and the flow of mixed autonomous and non-autonomous vehicles on a circular road. In this latter case, we analyze two cost functionals measuring the traffic congestion, depending on the penetration ratio of autonomous vehicles. The final Appendix contains alternative L and BV estimates, based on approximate solutions constructed via a Lax-Friedrichs type scheme, which is commonly used in the framework of non-local equations, see [1,2,5].

    First of all, we extend ωi(x)=0 for x>ηi. For jZ and nN, let xj+1/2=jΔx be the cell interfaces, xj=(j1/2)Δx the cells centers and tn=nΔt the time mesh. We aim at constructing a finite volume approximate solution ρΔx=(ρΔx1,,ρΔxM), with ρΔxi(t,x)=ρni,j for (t,x)Cnj=[tn,tn+1[×]xj1/2,xj+1/2] and i=1,...,M.

    To this end, we approximate the initial datum ρ0i for i=1,...,M with a piecewise constant function

    ρ0i,j=1Δxxj+1/2xj1/2ρ0i(x)dx,jZ.

    Similarly, for the kernel, we set

    ωki:=1Δx(k+1)ΔxkΔxω0i(x)dx,kN,

    so that Δx+k=0ωki=ηi0ωi(x)dx=Ji (the sum is indeed finite since ωki=0 for kNi sufficiently large). Moreover, we set rnj+k=Mi=1ρni,j+k for kN and

    Vni,j:=vmaxiψ(Δx+k=0ωkirnj+k),i=1,,M,jZ. (6)

    We consider the following Godunov-type scheme adapted to (1), which was introduced in [8] in the scalar case:

    ρn+1i,j=ρni,jλ(ρni,jVni,j+1ρni,j1Vni,j) (7)

    where we have set λ=ΔtΔx.

    We provide here the necessary estimates to prove the convergence of the sequence of approximate solutions constructed via the Godunov scheme (7).

    Lemma 2.1. (Positivity) For any T>0, under the CFL condition

    λ1vmaxMψ, (8)

    the scheme (7) is positivity preserving on [0,T]×R.

    Proof. Let us assume that ρni,j0 for all jZ and i1,...,M. It suffices to prove that ρn+1i,j in (7) is non-negative. We compute

    ρn+1i,j=ρni,j(1λVni,j+1)+λρni,j1Vni,j0 (9)

    under assumption (8).

    Corollary 1. (L1-bound) For any nN, under the CFL condition (8) the approximate solutions constructed via the scheme (7) satisfy

    ρni1=ρ0i1,i=1,,M, (10)

    where ρni1:=Δxj|ρni,j| denotes the L1 norm of the i-th component of ρΔx.

    Proof. Thanks to Lemma 2.1, for all i{1,...,M} we have

    ρn+1i1=Δxjρn+1i,j=Δxj(ρni,jλρni,jVni,j+1+λρni,j1Vni,j)=Δxjρni,j,

    proving (10).

    Lemma 2.2. (L-bound) If ρ0i,j0 for all jZ and i=1,...,M, and (8) holds, then the approximate solution ρΔx constructed by the algorithm (7) is uniformly bounded on [0,T]×R for any T such that

    T<(Mρ0vmaxMψW0)1.

    Proof. Let ˉρ=max{ρni,j1,ρni,j}. Then we get

    ρn+1i,j=ρni,j(1λVni,j+1)+λρni,j1Vni,jˉρ(1+λ(Vni,jVni,j+1)) (11)

    and

    |Vni,jVni,j+1|=vmaxi|ψ(Δx+k=0ωkirnj+k)ψ(Δx+k=0ωkirnj+k+1)|vmaxiψΔx|+k=0ωki(rnj+k+1rnj+k)|=vmaxiψΔx|ω0irnj++k=1(ωk1iωki)rnj+k|vmaxiψΔxMρnωi(0) (12)

    where ρ=(ρ1,,ρM)=maxi,j|ρi,j|. Let now K>0 be such that ρK, =0,,n. From (11) and (12) we get

    ρn+1ρn(1+MKvmaxMψW0Δt),

    which implies

    ρnρ0eCnΔt,

    with C=MKvmaxMψW0. Therefore we get that ρ(t,)K for

    t1MKvmaxMψW0ln(Kρ0)1Meρ0vmaxMψW0,

    where the maximum is attained for K=eρ0.

    Iterating the procedure, at time tm, m1 we set K=emρ0 and we get that the solution is bounded by K until tm+1 such that

    tm+1tm+mMemρ0vmaxMψW0.

    Therefore, the approximate solution remains bounded, uniformly in Δx, at least for tT with

    T1Mρ0vmaxMψW0+m=1mem1Mρ0vmaxMψW0.

    Remark 1. Figure 1 shows that the simplex

    Figure 1.  Numerical simulation illustrating that the simplex S is not an invariant domain for (1). We take M=2 and we consider the initial conditions ρ1(0,x)=0.9χ[0.5,0.3] and ρ2(0,x)=0.1χ],0]+χ]0,+[ depicted in (a), the constant kernels ω1(x)=ω2(x)=1/η, η=0.5, and the speed functions given by vmax1=0.2, vmax2=1, ψ(ξ)=max{1ξ,0} for ξ0. The space and time discretization steps are Δx=0.001 and Δt=0.4Δx. Plots (b) and (c) show the density profiles of ρ1, ρ2 and their sum r at times t=1.8, 2.8. The function maxxRr(t,x) is plotted in (d), showing that r can take values greater than 1, even if r(0,x)=ρ1(0,x)+ρ2(0,x)1.
    S:={ρRM:Mi=1ρi1,ρi0fori=1,,M}

    is not an invariant domain for (1), unlike the classical multi-population model [4]. Indeed, let us consider the system

    tρi(t,x)+x(ρi(t,x)vi(r(t,x)))=0,i=1,...,M, (13)

    where r and vi are as in (2) and (3), respectively. We have the following:

    Lemma 2.3. Under the CFL condition

    λ1vmaxM(ψ+ψ),

    for any initial datum ρ0S the approximate solutions to (13) computed by the upwind scheme

    ρn+1j=ρnjλ[F(ρnj,ρnj+1)F(ρnj1,ρnj)], (14)

    with F(ρnj,ρnj+1)=ρnjψ(rnj+1), satisfy the following uniform bounds:

    ρnjSjZ,nN. (15)

    Proof. Assuming that ρnjS for all jZ, we want to prove that ρn+1jS. Rewriting (14), we get

    ρn+1i,j=ρni,jλ[vmaxiρni,jψ(rnj+1)vmaxiρni,j1ψ(rnj)].

    Summing on the index i=1,,M, gives

    rn+1j=Mi=1ρn+1i,j=Mi=1ρni,jλMi=1[vmaxiρni,jψ(rnj+1)vmaxiρni,j1ψ(rnj)]=rnj+λψ(rnj)Mi=1vmaxiρni,j1λψ(rnj+1)Mi=1vmaxiρni,j.

    Defining the following function of ρnj

    Φ(ρn1,j,,ρnM,j)=rnj+λψ(rnj)Mi=1vmaxiρni,j1λψ(rnj+1)Mi=1vmaxiρni,j,

    we observe that

    Φ(0,,0)=λψ(0)Mivmaxiρni,j1λψvmaxM1

    if λ1/vmaxMψ and

    Φ(ρn1,j,...,ρnM,j)=1λψ(rnj+1)Mi=1vmaxiρni,j1

    for ρnjS such that rnj=Mi=1ρni,j=1. Moreover

    Φρni,j(ρnj)=1+λψ(rnj)Mi=1vmaxiρni,j1λψ(rnj+1)vmaxi0

    if λ1/vmaxM(ψ+ψ). This proves that rn+1j1. To prove the positivity of (14), we observe that

    ρn+1i,j=ρni,j(1λvmaxiψ(rnj+1))+λvmaxiρni,j1ψ(rnj)0

    if λ1/vmaxMψ.

    Lemma 2.4. (Spatial BV-bound) Let ρ0i(BVL)(R,R+) for all i=1,...,M. If (8) holds, then the approximate solution ρΔx(t,) constructed by the algorithm (7) has uniformly bounded total variation for t[0,T], for any T such that

    Tmini=1,,M 1H(TV(ρ0i)+1), (16)

    where H=ρvmaxMW0M(6MJ0ρψ+ψ).

    Proof. Subtracting the identities

    ρn+1i,j+1=ρni,j+1λ(ρni,j+1Vni,j+2ρni,jVni,j+1), (17)
    ρn+1i,j=ρni,jλ(ρni,jVni,j+1ρni,j1Vni,j), (18)

    and setting Δni,j+1/2=ρni,j+1ρni,j, we get

    Δn+1i,j+1/2=Δni,j+1/2λ(ρni,j+1Vni,j+22ρni,jVni,j+1+ρni,j1Vni,j).

    Now, we can write

    Δn+1i,j+1/2=(1λVni,j+2)Δni,j+1 (19)
    +λVni,jΔni,j1/2λρni,j(Vni,j+22Vni,j+1+Vni,j). (20)

    Observe that assumption (8) guarantees the positivity of (19). The term (20) can be estimated as

    Vni,j+22Vni,j+1+Vni,j==vmaxi(ψ(Δx+k=0ωkirnj+k+2)2ψ(Δx+k=0ωkirnj+k+1)+ψ(Δx+k=0ωkirnj+k))=vmaxiψ(ξj+1)Δx(+k=0ωkirnj+k+2+k=0ωkirnj+k+1)+vmaxiψ(ξj)Δx(+k=0ωkirnj+k+k=0ωkirnj+k+1)=vmaxiψ(ξj+1)Δx(+k=1(ωk1iωki)rnj+k+1ω0irnj+1)+vmaxiψ(ξj)Δx(+k=1(ωkiωk1i)rnj+k+ω0irnj)=vmaxi(ψ(ξj+1)ψ(ξj))Δx(+k=1(ωk1iωki)rnj+k+1ω0irnj+1)+vmaxiψ(ξj)Δx(+k=1(ωk1iωki)(rnj+k+1rnj+k)+ω0i(rnjrnj+1))=vmaxiψ(˜ξj+1/2)(ξj+1ξj)Δx(+k=1Mβ=1ωkiΔnβ,j+k+3/2)+vmaxiψ(ξj)Δx(Mβ=1N1k=1(ωk1iωki)Δnβ,j+k+1/2ω0iΔnβ,j+1/2),

    with ξjI(Δx+k=0ωkirnj+k,Δx+k=0ωkirnj+k+1) and ˜ξj+1/2I(ξj,ξj+1), where we set I(a,b)=[min{a,b},max{a,b}]. For some ϑ,μ[0,1], we compute

    ξj+1ξj=ϑΔx+k=0ωkiMβ=1ρnβ,j+k+2+(1ϑ)Δx+k=0ωkiMβ=1ρnβ,j+k+1μΔx+k=0ωkiMβ=1ρnβ,j+k+1(1μ)Δx+k=0ωkiMβ=1ρnβ,j+k=ϑΔx+k=1ωk1iMβ=1ρnβ,j+k+1+(1ϑ)Δx+k=0ωkiMβ=1ρnβ,j+k+1μΔx+k=0ωkiMβ=1ρnβ,j+k+1(1μ)Δx+k=1ωk+1iMβ=1ρnβ,j+k+1=Δx+k=1[ϑωk1i+(1ϑ)ωkiμωki(1μ)ωk+1i]Mβ=1ρnβ,j+k+1+(1ϑ)Δxω0iMβ=1ρnβ,j+1μΔxω0iMβ=1ρnβ,j+1(1μ)Δx(ω0iMβ=1ρnβ,j+ω1iMβ=1ρnβ,j+1).

    By monotonicity of ωi we have

    ϑωk1i+(1ϑ)ωkiμωki(1μ)ωk+1i0.

    Taking the absolute values we get

    |ξj+1ξj|Δx{+k=2[ϑωk1i+(1ϑ)ωkiμωki(1μ)ωk+1i]+4ω0i}MρnΔx{+k=2[ωk1iωk+1i]+4ω0i}MρnΔx6W0Mρn.

    Let now K1>0 be such that j|Δβ,j|K1 for β=1,,M, =0,,n. Taking the absolute values and rearranging the indexes, we have

    j|Δn+1i,j+1/2|j|Δni,j+1/2|(1λ(Vni,j+2Vni,j+1))+ΔtHK1,

    where H=ρvmaxMW0M(6MJ0ρψ"+ψ). Therefore, by (12) we get

    j|Δn+1i,j+1/2|j|Δni,j+1/2|(1+ΔtG)+ΔtHK1,

    with G=vmaxMψW0Mρ. We thus obtain

    j|Δni,j+1/2|eGnΔtj|Δ0i,j+1/2|+eHK1nΔt1,

    that we can rewrite as

    TV(ρΔxi)(nΔt,)eGnΔtTV(ρ0i)+eHK1nΔt1eHK1nΔt(TV(ρ0i)+1)1,

    since HG and it is not restrictive to assume K11. Therefore, we have that TV(ρΔxi)K1 for

    t1HK1ln(K1+1TV(ρ0i)+1),

    where the maximum is attained for some K1<e(TV(ρ0i)+1)1 such that

    ln(K1+1TV(ρ0i)+1)=K1K1+1.

    Therefore the total variation is uniformly bounded for

    t1He(TV(ρ0i)+1).

    Iterating the procedure, at time tm, m1 we set K1=em(TV(ρ0i)+1)1 and we get that the solution is bounded by K1 until tm+1 such that

    tm+1tm+mHem(TV(ρ0i)+1). (21)

    Therefore, the approximate solution has bounded total variation for tT with

    T1H(TV(ρ0i)+1).

    Corollary 2. Let ρ0i(BVL)(R;R+). If (8) holds, then the approximate solution ρΔx constructed by the algorithm (7) has uniformly bounded total variation on [0,T]×R, for any T satisfying (16).

    Proof. If TΔt, then TV(ρΔxi;[0,T]×R)TTV(ρ0i). Let us assume now that T>Δt. Let nTN{0} such that nTΔt<T(nT+1)Δt. Then

    TV(ρΔxi;[0,T]×R)=nT1n=0jZΔt|ρni,j+1ρni,j|+(TnTΔt)jZ|ρnTi,j+1ρnTi,j|Tsupt[0,T]TV(ρΔxi)(t,)+nT1n=0jZΔx|ρn+1i,jρni,j|.

    We then need to bound the term

    nT1n=0jZΔx|ρn+1i,jρni,j|.

    From the definition of the numerical scheme (7), we obtain

    ρn+1i,jρni,j=λ(ρni,j1Vni,jρni,jVni,j+1)=λ(ρni,j1(Vni,jVni,j+1)+Vni,j+1(ρni,j1ρni,j)).

    Taking the absolute values and using (12) we obtain

    |ρn+1i,jρni,j|λ(vmaxiψMρnωi(0)Δx|ρni,j1|+vmaxiψ|ρni,j1ρni,j|).

    Summing on j, we get

    jZΔx|ρn+1i,jρni,j|=vmaxiψMρnωi(0)ΔtjZΔx|ρni,j1|+vmaxiψΔtjZ|ρni,j1ρni,j|,

    which yields

    nT1n=0jZΔx|ρn+1i,jρni,j|vmaxMψTsupt[0,T]TV(ρΔxi)(t,)+vmaxMψMW0Tsupt[0,T]ρΔxi(t,)1ρΔxi(t,)

    that is bounded by Corollary 1, Lemma 2.2 and Lemma 2.4.

    To complete the proof of the existence of solutions to the problem (1), (5), we follow a Lax-Wendroff type argument as in [5], see also [10], to show that the approximate solutions constructed by scheme (7) converge to a weak solution of (1). By Lemma 2.2, Lemma 2.4 and Corollary 2, we can apply Helly's theorem, stating that for i=1,,M, there exists a subsequence, still denoted by ρΔxi, which converges to some ρi(L1BV)([0,T]×R;R+) in the L1loc-norm. Let us fix i{1,,M}. Let φC1c([0,T[×R) and multiply (7) by φ(tn,xj). Summing over jZ and n{0,,nT} we get

    nT1n=0jφ(tn,xj)(ρn+1i,jρni,j)=λnT1n=0jφ(tn,xj)(ρni,jVni,j+1ρni,j1Vni,j).

    Summing by parts we obtain

    jφ((nT1)Δt,xj)ρnTi,j+jφ(0,xj)ρ0i,j+nT1n=1j(φ(tn,xj)φ(tn1,xj))ρni,j+λnT1n=0j(φ(tn,xj+1)φ(tn,xj))Vni,j+1ρni,j=0. (22)

    Multiplying by Δx we get

    Δxjφ((nT1)Δt,xj)ρnTi,j+Δxjφ(0,xj)ρ0i,j (23)
    +ΔxΔtnT1n=1j(φ(tn,xj)φ(tn1,xj))Δtρni,j (24)
    +ΔxΔtnT1n=0j(φ(tn,xj+1)φ(tn,xj))ΔxVni,j+1ρni,j=0. (25)

    By L1loc convergence of ρΔxiρi, it is straightforward to see that the terms in (23), (24) converge to

    R(ρ0i(x)φ(0,x)ρi(T,x)φ(T,x))dx+T0Rρi(t,x)tφ(t,x)dxdt, (26)

    as Δx0. Concerning the last term (25), we can rewrite

    ΔxΔtnT1n=0jφ(tn,xj+1)φ(tn,xj)ΔxVni,j+1ρni,j=ΔxΔtnT1n=0jφ(tn,xj+1)φ(tn,xj)Δx(ρni,jVni,j+1ρni,jVni,j)+ΔxΔtnT1n=0jφ(tn,xj+1)φ(tn,xj)Δxρni,jVni,j. (27)

    By (12) we get the estimate

    ρni,jVni,j+1ρni,jVni,jvmaxiψΔxMρ2ωi(0).

    Set R>0 such that φ(t,x)=0 for |x|>R and j0,j1Z such that R]xj012,xj0+12] and R]xj112,xj1+12], then

    ΔxΔtnTn=0jφ(tn,xj+1)φ(tn,xj)Δx(ρni,jVni,j+1ρni,jVni,j)ΔxΔtxφnTn=0j1j=j0vmaxiψMρ2ωi(0)ΔxxφvmaxiψMρ2ωi(0)Δx2RT,

    which goes to zero as Δx0.

    Finally, again by the L1loc convergence of ρΔxiρi, we have that

    ΔxΔtnT1n=0j(φ(tn,xj+1)φ(tn,xj))Δxρni,jVni,j12T0Rxφ(t,x)ρi(t,x)vi(rωi)dxdt.

    In this section we perform some numerical simulations to illustrate the behaviour of solutions to (1) for M=2 modeling two different scenarios. In the following, the space mesh is set to Δx=0.001.

    In this example, we consider a stretch of road populated by cars and trucks. The space domain is given by the interval [2,3] and we impose absorbing conditions at the boundaries, adding N1=η1/Δx ghost cells for the first population and N2=η2/Δx for the second one at the right boundary, and just one ghost cell for both populations at the left boundary, where we extend the solution constantly equal to the last value inside the domain. The dynamics is described by the following 2×2 system

    {tρ1(t,x)+x(ρ1(t,x)vmax1ψ((rω1)(t,x)))=0,tρ2(t,x)+x(ρ2(t,x)vmax2ψ((rω2)(t,x)))=0, (28)

    with

    ω1(x)=2η1(1xη1),η1=0.3,ω2(x)=2η2(1xη2),η2=0.1,ψ(ξ)=max{1ξ,0},ξ0,vmax1=0.8,vmax2=1.3.

    In this setting, ρ1 represents the density of trucks and ρ2 is the density of cars on the road. Trucks moves at lower maximal speed than cars and have grater view horizon, but of the same order of magnitude. Figure 2 describes the evolution in time of the two population densities, correspondent to the initial configuration

    Figure 2.  Density profiles of cars and trucks at increasing times corresponding to the non-local model (28).
    {ρ1(0,x)=0.5χ[1.1,1.6],ρ2(0,x)=0.5χ[1.6,1.9],

    in which a platoon of trucks precedes a group of cars. Due to their higher speed, cars overtake trucks, in accordance with what observed in the local case [4].

    The aim of this test is to study the possible impact of the presence of Connected Autonomous Vehicles (CAVs) on road traffic performances. Let us consider a circular road modeled by the space interval [1,1] with periodic boundary conditions at x=±1. In this case, we assume that autonomous and non-autonomous vehicles have the same maximal speed, but the interaction radius of CAVs is two orders of magnitude grater than the one of human-driven cars. Moreover, we assume CAVs have constant convolution kernel, modeling the fact that they have the same degree of accuracy on information about surrounding traffic, independent from the distance. In this case, model (1) reads

    {tρ1(t,x)+x(ρ1(t,x)vmax1ψ((rω1)(t,x)))=0,tρ2(t,x)+x(ρ2(t,x)vmax2ψ((rω2)(t,x)))=0,ρ1(0,x)=β(0.5+0.3sin(5πx)),ρ2(0,x)=(1β)(0.5+0.3sin(5πx)), (29)

    with

    ω1(x)=1η1,η1=1,ω2(x)=2η2(1xη2),η2=0.01,ψ(ξ)=max{1ξ,0},ξ0,vmax1=vmax2=1.

    Above ρ1 represents the density of autonomous vehicles, ρ2 the density of non-autonomous vehicles and β[0,1] is the penetration rate of autonomous vehicle. Figure 3 displays the traffic dynamics in the case β=0.9.

    Figure 3.  Density profiles corresponding to the non-local problem (29) with β=0.9 at different times.

    As a metric of traffic congestion, given a time horizon T>0, we consider the two following functionals:

    J(β)=T0d|xr|dt, (30)
    Ψ(β)=T0[ρ1(t,ˉx)vmax1ψ((rω1)(t,ˉx))+ρ2(t,ˉx)vmax2ψ((rω2)(t,ˉx))]dt, (31)

    where ˉx=x00. The functional J measures the integral with respect to time of the spatial total variation of the total traffic density, see [6]. The functional Ψ measures the integral with respect to time of the traffic flow at a given point ˉx, corresponding to the number of cars that have passed through ˉx in the studied time interval. Figure 4 displays the values of the functionals J and Ψ for different values of β=0,0.1,0.2,,1. We can notice that the functionals are not monotone and present minimum and maximum values. The traffic evolution patterns corresponding these stationary values are reported in Figure 5, showing the (t,x)-plots of the total traffic density r(t,x) corresponding to these values of β.

    Figure 4.  Functional J (left) and Ψ (right).
    Figure 5.  (t,x)-plots of the total traffic density r(t,x)=ρ1(t,x)+ρ2(t,x) in (29) corresponding to different values of β: (a) no autonomous vehicles are present; (b) point of minimum for Ψ; (c) point of minimum for J; (d) point of maximum for J.

    The authors are grateful to Luis M. Villada for suggesting the non-local multi-class traffic model studied in this paper.

    We provide here alternative estimates for (1), based on approximate solutions constructed via the following adapted Lax-Friedrichs scheme:

    ρn+1i,j=ρni,jλ(Fni,j+1/2Fni,j1/2), (32)

    with

    Fni,j+1/2:=12ρni,jVni,j+12ρni,j+1Vni,j+1+α2(ρni,jρni,j+1), (33)

    where α1 is the viscosity coefficient and λ=ΔtΔx. The proofs are very similar to those exposed for Godunov approximations.

    Lemma A.1. For any T>0, under the CFL conditions

    λα<1, (34)
    αvmaxMψ, (35)

    the scheme (33)-(32) is positivity preserving on [0,T]×R.

    Lemma A.2. (L-bound) If ρ0i,j0 for all jZ and i=1,...,M, and the CFL conditions (34)- (35) hold, the approximate solution ρΔx constructed by the algorithm (33)- (32) is uniformly bounded on [0,T]×R for any T such that

    T<(Mρ0vmaxMψW0)1. (36)

    Lemma A.3. (BV estimates) Let ρ0i(BVL)(R,R+) for all i=1,...,M. If (35) holds and

    Δt22α+ΔxψW0vmaxMρΔx, (37)

    then the solution constructed by the algorithm (33)-(32) has uniformly bounded total variation for any T such that

    Tmini=1,...,M1D(TV(ρ0i)+1), (38)

    where D=ρvmaxMW0M(3MJ0ρψ"+2ψ).



    [1] S. W. Anzengruber, The Discrepancy Principle for Tikhonov Regularization in Banach spaces: Regularization pPoperties and Rates of Convergence. Südwestdeutscher Verlag für Hochschulschriften, 2012.
    [2] L. M. Bregman, The relaxation method of finding the common point of convex sets and its application to the solution of problems in convex programming, USSR Comput. Math. Math. Phys., 7 (1976), 200–217.
    [3] M. Burger, S. Osher, Convergence rates of convex variational regularization, Inverse Probl., 20 (2004), 1411. doi: 10.1088/0266-5611/20/5/005
    [4] V. Caselles, A. Chambolle, M. Novaga, Total variation in imaging, Handb. Math. Methods Imaging, 2015 (2015), 1455–1499.
    [5] T. Chan, S. Esedoglu, F. Park, A. Yip, Recent developments in total variation image restoration, Math. Models Comput. Vision, 17 (2005).
    [6] I. Ekeland, R. Temam, Convex analysis and variational problems, 28 (1999).
    [7] H. W. Engl, M. Hanke, A. Neubauer, Regularization of Inverse Problems, Springer Science Business Media, 375 (1996).
    [8] J. Flemming, Theory and examples of variational regularization with non-metric fitting functionals, J. Inverse Ill-Posed Probl., 18 (2010), 677–699,
    [9] A. Gholami, S. M. Hosseini, A balanced combination of Tikhonov and total variation regularizations for reconstruction of piecewise-smooth signals, Signal Processing, 93 (2013), 1945–1960. doi: 10.1016/j.sigpro.2012.12.008
    [10] C. Groetsch, The theory of Tikhonov regularization for Fredholm equations. 104p, Boston Pitman Publication, 1984.
    [11] J. Hadamard, Lectures on Cauchy's problem in linear partial differential equations, yale univ, Press. New Haven, 1923.
    [12] B. Hofmann, B. Kaltenbacher, C. Poeschl, O. Scherzer, A convergence rates result for Tikhonov regularization in Banach spaces with non-smooth operators, Inverse Problems, 23 (2007), 987. doi: 10.1088/0266-5611/23/3/009
    [13] B. K. Horn, B. G. Schunck, Determining optical flow, Artif. Intel., 17 (1981), 185–203.
    [14] E. M. Kalmoun, An investigation of smooth tv-like regularization in the context of the optical flow problem, J. Imaging, 4 (2018), 31. doi: 10.3390/jimaging4020031
    [15] P. Kazemi, R. J. Renka, Tikhonov regularization using Sobolev metrics, Electron. J. Differ. Eq., 2014.
    [16] S. Kindermann, Convex Tikhonov regularization in Banach spaces: New results on convergence rates, J. Inverse Ill-posed Probl., 24 (2016), 341–350.
    [17] G. Mazzieri, R. Spies, K. Temperini, Existence, uniqueness and stability of minimizers of generalized Tikhonov–Phillips functionals, J. Math. Anal. Appl., 396 (2012), 396–411. doi: 10.1016/j.jmaa.2012.06.039
    [18] F. D. M. Neto, A. J. da Silva Neto, An Introduction to Inverse Problems with Applications, Springer Science & Business Media, 2012.
    [19] D. L. Phillips, A technique for the numerical solution of certain integral equations of the first kind, J. ACM (JACM), 9 (1962), 84–97. doi: 10.1145/321105.321114
    [20] C. Pöschl, Tikhonov Regularization with General Residual Term, PhD thesis, University of Innsbruck, Austria, 2008.
    [21] E. Resmerita, Regularization of ill-posed problems in Banach spaces: Convergence rates, Inverse Probl., 21 (2005), 1303. doi: 10.1088/0266-5611/21/4/007
    [22] L. I. Rudin, S. Osher, E. Fatemi, Nonlinear total variation based noise removal algorithms, Physica D: Nonlinear Phenom., 60 (1992), 259–268. doi: 10.1016/0167-2789(92)90242-F
    [23] S. Salzo, Variational Regularization for Image Registration: Theory and Algorithms, PhD thesis, Dipartimento di Informatica e Scienze dell'Informazione, 2012.
    [24] A. N. Tikhonov, Solution of incorrectly formulated problems and the regularization method, Soviet Math., 4 (1963), 1035–1038.
    [25] P. Viola, W. M. Wells III, Alignment by maximization of mutual information, Int. J. Comput. Vision, 24 (1997), 137–154. doi: 10.1023/A:1007958904918
  • This article has been cited by:

    1. Rinaldo M. Colombo, Magali Lecureux-Mercier, Mauro Garavello, 2020, Chapter 5, 978-3-030-50449-6, 83, 10.1007/978-3-030-50450-2_5
    2. Felisia Angela Chiarello, Paola Goatin, Luis Miguel Villada, Lagrangian-antidiffusive remap schemes for non-local multi-class traffic flow models, 2020, 39, 2238-3603, 10.1007/s40314-020-1097-9
    3. Alexandre Bayen, Jan Friedrich, Alexander Keimer, Lukas Pflug, Tanya Veeravalli, Modeling Multilane Traffic with Moving Obstacles by Nonlocal Balance Laws, 2022, 21, 1536-0040, 1495, 10.1137/20M1366654
    4. Felisia Angela Chiarello, 2021, Chapter 5, 978-3-030-66559-3, 79, 10.1007/978-3-030-66560-9_5
    5. Jan Friedrich, Simone Göttlich, Maximilian Osztfalk, Network models for nonlocal traffic flow, 2022, 56, 2822-7840, 213, 10.1051/m2an/2022002
    6. Zlatinka Dimitrova, Flows of Substances in Networks and Network Channels: Selected Results and Applications, 2022, 24, 1099-4300, 1485, 10.3390/e24101485
    7. Giuseppe Maria Coclite, Lorenzo di Ruvo, On the initial-boundary value problem for a non-local elliptic-hyperbolic system related to the short pulse equation, 2022, 3, 2662-2963, 10.1007/s42985-022-00208-w
    8. Kuang Huang, Qiang Du, Stability of a Nonlocal Traffic Flow Model for Connected Vehicles, 2022, 82, 0036-1399, 221, 10.1137/20M1355732
    9. Yanbing Wang, Daniel B. Work, Estimation for heterogeneous traffic using enhanced particle filters, 2022, 18, 2324-9935, 568, 10.1080/23249935.2021.1881186
    10. Felisia Angela Chiarello, Harold Deivi Contreras, Luis Miguel Villada, Nonlocal reaction traffic flow model with on-off ramps, 2022, 17, 1556-1801, 203, 10.3934/nhm.2022003
    11. Ioana Ciotir, Rim Fayad, Nicolas Forcadel, Antoine Tonnoir, A non-local macroscopic model for traffic flow, 2021, 55, 0764-583X, 689, 10.1051/m2an/2021006
    12. Maria Colombo, Gianluca Crippa, Marie Graff, Laura V. Spinolo, On the role of numerical viscosity in the study of the local limit of nonlocal conservation laws, 2021, 55, 0764-583X, 2705, 10.1051/m2an/2021073
    13. Alexander Keimer, Lukas Pflug, 2023, 15708659, 10.1016/bs.hna.2022.11.001
    14. Alexandre Bayen, Jean-Michel Coron, Nicola De Nitti, Alexander Keimer, Lukas Pflug, Boundary Controllability and Asymptotic Stabilization of a Nonlocal Traffic Flow Model, 2021, 49, 2305-221X, 957, 10.1007/s10013-021-00506-7
    15. F. A. CHIARELLO, J. FRIEDRICH, P. GOATIN, S. GÖTTLICH, O. KOLB, A non-local traffic flow model for 1-to-1 junctions, 2020, 31, 0956-7925, 1029, 10.1017/S095679251900038X
    16. Jan Friedrich, Simone Göttlich, Alexander Keimer, Lukas Pflug, 2024, Chapter 30, 978-3-031-55263-2, 347, 10.1007/978-3-031-55264-9_30
    17. Veerappa Gowda G. D., Sudarshan Kumar Kenettinkara, Nikhil Manoj, Convergence of a second-order scheme for non-local conservation laws, 2023, 57, 2822-7840, 3439, 10.1051/m2an/2023080
    18. Jan Friedrich, Sanjibanee Sudha, Samala Rathan, Numerical schemes for a class of nonlocal conservation laws: a general approach, 2023, 18, 1556-1801, 1335, 10.3934/nhm.2023058
    19. Jan Friedrich, Simone Göttlich, Michael Herty, Lyapunov Stabilization for Nonlocal Traffic Flow Models, 2023, 61, 0363-0129, 2849, 10.1137/22M152181X
    20. Felisia A. Chiarello, Harold D. Contreras, 2024, Chapter 26, 978-3-031-55263-2, 303, 10.1007/978-3-031-55264-9_26
    21. Jan Friedrich, Lyapunov stabilization of a nonlocal LWR traffic flow model, 2023, 23, 1617-7061, 10.1002/pamm.202200084
    22. Agatha Joumaa, Paola Goatin, Giovanni De Nunzio, 2023, A Macroscopic Model for Multi-Modal Traffic Flow in Urban Networks, 979-8-3503-9946-2, 411, 10.1109/ITSC57777.2023.10422168
    23. Saeed Mohammadian, Zuduo Zheng, Md. Mazharul Haque, Ashish Bhaskar, Continuum modeling of freeway traffic flows: State-of-the-art, challenges and future directions in the era of connected and automated vehicles, 2023, 3, 27724247, 100107, 10.1016/j.commtr.2023.100107
    24. Harold Deivi Contreras, Paola Goatin, Luis-Miguel Villada, A two-lane bidirectional nonlocal traffic model, 2025, 543, 0022247X, 129027, 10.1016/j.jmaa.2024.129027
    25. Felisia Angela Chiarello, Paola Goatin, 2023, Chapter 3, 978-3-031-29874-5, 49, 10.1007/978-3-031-29875-2_3
    26. Alexander Keimer, Lukas Pflug, Discontinuous nonlocal conservation laws and related discontinuous ODEs – Existence, Uniqueness, Stability and Regularity, 2023, 361, 1778-3569, 1723, 10.5802/crmath.490
    27. Archie J. Huang, Animesh Biswas, Shaurya Agarwal, Incorporating Nonlocal Traffic Flow Model in Physics-Informed Neural Networks, 2024, 25, 1524-9050, 16249, 10.1109/TITS.2024.3429029
    28. Rinaldo M. Colombo, Mauro Garavello, Claudia Nocita, General stability estimates in nonlocal traffic models for several populations, 2025, 32, 1021-9722, 10.1007/s00030-025-01034-w
  • Reader Comments
  • © 2021 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(1805) PDF downloads(61) Cited by(0)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog