betaFeedback

WorksMathematics

Quadrature-Induced Local Minima in Structure-Preserving Lotka-Volterra Control

Abstract

Harvest quadrature can change a discrete optimal control problem even when the state scheme stays fixed. We study this effect for a positivity-preserving Lotka–Volterra scheme. The difference appears in a model with harvesting alone: rewarding the scheme's removal flux gives a strictly convex reduced cost, whereas using prey density at the start of each step overestimates harvest and can create negative curvature and a strict local minimum that is not global. Rational root isolation certifies the two-step global minimizers. With diffusion in one space dimension, nonnegative bounded H1H^1 initial data and positive diffusivities, the costs, weighted gradients and Hessians differ by O(τ)O(τ) uniformly in the spatial mesh, where ττ is the time step. A uniform positive lower bound on the reduced Hessian gives the same order for the distance between nearby controls. A large enough effort penalty gives unique global discrete minimizers. We also prove positivity, exact population balance and a common variational limit, and derive exact discrete adjoints and Hessian actions. Numerical tests trace changes in curvature and stationary controls across parameters and grids, with costs evaluated on a common reference using adaptive time stepping. The construction extends to weighted production–destruction systems, giving exact harvest accounting and smooth derivatives on fixed grids.

Introduction

Harvesting removes prey and leaves less food for predators, while diffusion spreads both effects in space [2, 3, 4, 5, 37]. At position xx and time tt, we write v1(x,t)v_1(x,t) and v2(x,t)v_2(x,t) for predator and prey densities, and u(x,t)u(x,t) for fishing effort. The objective balances population tracking, effort costs and harvest reward. Discretizing this problem means choosing how to approximate both the dynamics and the reward [32, 31, 29, 34, 30].

Implicit–explicit (IMEX) Euler treats diffusion implicitly and reactions explicitly, preserving nonnegativity under a time-step restriction [6, 7]. A Patankar production–destruction step preserves positivity for every time step [8]. How, then, should harvest be counted over a step? To separate the effects of the state scheme and the reward, we compare three discrete problems. I use IMEX and the prey density at the start of the step. P0P_0 and P1P_1 share the same triangular positive scheme but use prey density at the start and end of the step, respectively. These are the old-state and new-state rules. Comparing II with P0P_0 isolates the state scheme, and comparing P0P_0 with P1P_1 isolates harvest quadrature.

The new-state rule rewards exactly the removal flux in the population balance. In the model with harvesting alone, the old-state rule overestimates removal. That extra reward can create negative curvature in directions that redistribute effort between time steps, add a strict local minimum, and favor extreme allocations at fixed total effort. For the full discrete problem, we derive state bounds uniform in the control and mesh, exact discrete adjoints, and Hessian-vector products that require no assembled Hessian matrix [10, 11]. Because the reward depends on the final prey state, its derivative enters the terminal adjoint condition.

For the positive scheme with diffusion in one dimension, we bound the gradient and Hessian differences uniformly in the mesh. Under coercivity, these bounds control the change in the optimal control. On Ω=(0,1)\Omega=(0,1), with nonnegative bounded H1H^1 initial data, all three problems have the same weak-* variational limit. Discrete controls whose costs approach the discrete global minimum have subsequences converging strongly in L2L^2 to continuous global minimizers, and discrete local minimizers approximate every strict L2L^2 local minimizer.

The numerical tests check the derivatives and compare four optimizers on a common reference. Parameter continuation and grid refinement then trace the effects of quadrature on temporal switching, sensitivity to the ini tial control and reduced curvature. For relaxed forward–backward sweep (FBS), we give an exact local contraction criterion: under strict complementarity and a positive definite Hessian on the free coordinates, the criterion depends on that Hessian’s smallest and largest eigenvalues.

This analysis builds on consistent discrete adjoints and second derivatives [12, 13, 14, 15], together with the positivity, conservation and accuracy of Patankar schemes [8, 16, 9]. Cacace et al. study production–destruction control by dynamic programming with a modified Patankar semi-Lagrangian scheme [17]. Here we examine how harvest quadrature changes curvature, derivatives and controls.

Continuous problem

We write R\mathbb{R} for the real numbers and N:={1,2,…}\mathbb{N}:=\{1,2,\ldots\}. The symbol ≔\coloneqq introduces a definition. Let Ω⊂Rd\Omega\subset\mathbb{R}^{d} be a bounded Lipschitz domain, d∈Nd\in\mathbb{N}, and let T>0T>0 be the final time. On the space–time domain Q:=Ω×(0,T)Q:=\Omega\times(0,T), write v1,v2:Q→Rv_{1},v_{2}:Q\to\mathbb{R} for the densities and u:Q→Ru:Q\to\mathbb{R} for effort. In the following system, the evolution equations hold in QQ, the boundary conditions on ∂Ω×(0,T)\partial\Omega\times(0,T), and the initial conditions in Ω\Omega for i=1,2i=1,2:

{∂tv1−μ1Δv1=v1(−m+αv2),∂tv2−μ2Δv2=v2(r−βv1)−quv2,∂nv1=∂nv2=0,vi(0)=vi0.(1)\begin{cases} \partial_{t}v_{1}-\mu_{1}\Delta v_{1}=v_{1}(-m+\alpha v_{2}),\\ \partial_{t}v_{2}-\mu_{2}\Delta v_{2}=v_{2}(r-\beta v_{1})-quv_{2},\\ \partial_{n}v_{1}=\partial_{n}v_{2}=0,\\ v_{i}(0)=v_{i}^{0}. \tag*{(1)} \end{cases}

Here ∂Ω\partial\Omega is the boundary, ∂t\partial_{t} is the time derivative, Δ\Delta is the spatial Laplacian and ∂n\partial_{n} the outward normal derivative. We write vi(t)v_{i}(t) for the spatial profile vi(⋅,t)v_{i}(\cdot,t) and vi0v_{i}^{0} for its initial value. The model has positive diffusivities μi\mu_{i}, mortality mm, growth rate rr, interaction coefficients α,β\alpha,\beta, catchability qq, and effort bound Umax⁡U_{\max}. We assume μi,m,r,α,β,q,Umax⁡>0\mu_{i},m,r,\alpha,\beta,q,U_{\max}>0. The weights γi≥0\gamma_{i}\geq0, λ>0\lambda>0, and η≥0\eta\geq0 determine the penalties for tracking and effort, and the reward for harvest. For a domain DD, Lp(D)L^{p}(D), 1≤p<∞1\leq p<\infty, consists of measurable functions with finite norm

∥f∥Lp(D):=(∫D∣f∣p)1/p.\|f\|_{L^{p}(D)}:=\left(\int_{D}|f|^{p}\right)^{1/p}.

Thus L1L^{1} means integrable and L2L^{2} square-integrable. The space L∞(D)L^{\infty}(D) contains essentially bounded functions, with norm equal to the essential supremum of ∣f∣|f|. Functions equal almost everywhere are identified. The abbreviation “a.e.” means outside a set of measure zero. We write ∣D∣|D| for the measure of DD. We constrain effort to the admissible set

Uad:={u∈L∞(Q):0≤u≤Umax⁡ a.e.}.\mathcal{U}_{\mathrm{ad}}:=\{u\in L^{\infty}(Q):0\leq u\leq U_{\max}\ \text{a.e.}\}.

For target profiles vid∈L2(Q)v_{i}^{d}\in L^{2}(Q), write v:=(v1,v2)v:=(v_{1},v_{2}) and define the cost functional

J:L2(Q)2×L2(Q)⟶RJ:L^{2}(Q)^{2}\times L^{2}(Q)\longrightarrow\mathbb{R}
(v,u)⟼12∑i=12γi∥vi−vid∥L2(Q)2+λ2∥u∥L2(Q)2−ηq∫Quv2.(v,u)\longmapsto\frac{1}{2}\sum_{i=1}^{2}\gamma_{i}\|v_{i}-v_{i}^{d}\|_{L^{2}(Q)}^{2} +\frac{\lambda}{2}\|u\|_{L^{2}(Q)}^{2}-\eta q\int_{Q}uv_{2}.

Spaces, bounds, and existence

All spaces are real. The Sobolev space H1(D)H^{1}(D) consists of L2(D)L^{2}(D) functions whose first weak derivatives also lie in L2(D)L^{2}(D), with

∥f∥H1(D)2=∥f∥L2(D)2+∥∇f∥L2(D)2.\|f\|_{H^{1}(D)}^{2}=\|f\|_{L^{2}(D)}^{2}+\|\nabla f\|_{L^{2}(D)}^{2}.

Here ∇f\nabla f is the spatial gradient. Weak derivatives are defined by integration by parts against smooth test functions. For a Banach space XX, Lp(0,T;X)L^{p}(0,T;X) denotes strongly measurable XX-valued functions whose norm lies in scalar Lp(0,T)L^{p}(0,T). The space H1(0,T;X)H^{1}(0,T;X) also requires a weak time derivative in L2(0,T;X)L^{2}(0,T;X). The continuous dual X∗X^{*} consists of continuous linear functionals, with pairing ⟨ℓ,x⟩X∗,X=ℓ(x)\langle\ell,x\rangle_{X^{*},X}=\ell(x). We write XmX^{m} for the product of mm copies of XX. Set

W:=L2(0,T;H1(Ω))∩H1(0,T;H1(Ω)∗),\mathcal{W}:=L^{2}(0,T;H^{1}(\Omega))\cap H^{1}(0,T;H^{1}(\Omega)^{*}),
Y:=W∩L∞(Q),\mathcal{Y}:=\mathcal{W}\cap L^{\infty}(Q),

with norm

∥y∥Y:=∥y∥L2(0,T;H1)+∥∂ty∥L2(0,T;H1∗)+∥y∥L∞(Q).\|y\|_{\mathcal{Y}}:=\|y\|_{L^{2}(0,T;H^{1})}+\|\partial_{t}y\|_{L^{2}(0,T;H^{1*})}+\|y\|_{L^{\infty}(Q)}.

The control-to-state map and reduced objective are

S:Uad⟶YS:\mathcal{U}_{\mathrm{ad}}\longrightarrow\mathcal{Y}
u\longmapsto v\text{ solving \({xref}`(1)`}

and

j:Uad⟶Rj:\mathcal{U}_{\mathrm{ad}}\longrightarrow\mathbb{R}
u⟼J(S(u),u).u\longmapsto J(S(u),u).

References

  1. [1]A. Autran, “Lotka–Volterra optimal control,” GitHub repository, 2026. Available: https://github.com/AleXXandreAutran/structure-preserving-lotka-volterra-control.
  2. [2]A. Autran, M. J. Gander, B. Chaudet-Dumas, Numerical Simulation and Optimal Control of Spatial Lotka–Volterra Systems, MSc research internship report, University of Geneva, 2026. Available: https://alexxandreautran.github.io/documents/research/msc/lotka-volterra-optimal-control-report.pdf.
  3. [3]J. D. Murray, Mathematical Biology I: An Introduction, 3rd ed., ser. Interdisciplinary Applied Mathematics, vol. 17. New York: Springer, 2002.DOI
  4. [4]I. Hrinça, “An optimal control problem for the Lotka–Volterra system with diffusion,” PanAmerican Mathematical Journal, vol. 12, no. 3, pp. 23–46, 2002.
  5. [5]N. C. Apreutesei, “An optimal control problem for a predator–prey reaction–diffusion system,” Mathematical Modelling of Natural Phenomena, vol. 5, no. 6, pp. 180–195, 2010.DOI
  6. [6]W. Hundsdorfer and J. G. Verwer, Numerical Solution of Time-Dependent Advection–Diffusion–Reaction Equations, ser. Springer Series in Computational Mathematics, vol. 33. Berlin: Springer, 2003.DOI
  7. [7]A. Quarteroni, R. Sacco, and F. Saleri, Numerical Mathematics, 2nd ed., ser. Texts in Applied Mathematics, vol. 37. Berlin: Springer, 2007.
  8. [8]H. Burchard, E. Deleersnijder, and A. Meister, “A high-order conservative Patankar-type discretisation for stiff systems of production–destruction equations,” Applied Numerical Mathematics, vol. 47, no. 1, pp. 1–30, 2003.DOI
  9. [9]D. Torlo, P. Öffner, and H. Ranocha, “Issues with positivity-preserving Patankar-type schemes,” Applied Numerical Mathematics, vol. 182, pp. 117–147, 2022.DOI
  10. [10]M. B. Giles and N. A. Pierce, “An introduction to the adjoint approach to design,” Flow, Turbulence and Combustion, vol. 65, no. 3–4, pp. 393–415, 2000.
  11. [11]M. Hinze, R. Pinnau, M. Ulbrich, and S. Ulbrich, Optimization with PDE Constraints, ser. Mathematical Modelling: Theory and Applications, vol. 23. Dordrecht: Springer, 2009.
  12. [12]W. W. Hager, “Runge–Kutta methods in optimal control and the transformed adjoint system,” Numerische Mathematik, vol. 87, no. 2, pp. 247–282, 2000. Available: https://people.clas.ufl.edu/hager/files/rk.pdf.DOI
  13. [13]M. Herty, L. Pareschi, and S. Steffensen, “Implicit-explicit Runge–Kutta schemes for numerical discretization of optimal control problems,” SIAM Journal on Numerical Analysis, vol. 51, no. 4, pp. 1875–1899, 2013. Available: https://arxiv.org/abs/1202.1166.
  14. [14]J. M. Sanz-Serna, “Symplectic Runge–Kutta schemes for adjoint equations, automatic differentiation, optimal control, and more,” SIAM Review, vol. 58, no. 1, pp. 3–33, 2016. Available: https://arxiv.org/abs/1503.04021.
  15. [15]S.-i. Ito, T. Matsuda, and Y. Miyatake, “Adjoint-based exact Hessian computation,” BIT Numerical Mathematics, vol. 61, pp. 503–522, 2021. Available: https://link.springer.com/article/10.1007/s10543-020-00833-0.arxiv.org/abs/1910.06524
  16. [16]S. Kopecz and A. Meister, “On order conditions for modified Patankar–Runge–Kutta schemes,” Applied Numerical Mathematics, vol. 123, pp. 159–179, 2018. Available: https://arxiv.org/abs/1702.04589.
  17. [17]S. Cacace, A. Oliviero, and M. Pezzella, “Modified Patankar semi-Lagrangian scheme for the optimal control of production-destruction systems,” Journal of Scientific Computing, vol. 106, Art. no. 43, 2026. Available: https://link.springer.com/article/10.1007/s10915-025-03167-7.DOI
  18. [18]F. Tröltzsch, Optimal Control of Partial Differential Equations: Theory, Methods and Applications, ser. Graduate Studies in Mathematics, vol. 112. Providence, RI: American Mathematical Society, 2010.
  19. [19]J. Simon, “Compact sets in the space Lᵖ(0,T;B),” Annali di Matematica Pura ed Applicata, vol. 146, pp. 65–96, 1987.
  20. [20]M. McAsey, L. Mou, and W. Han, “Convergence of the forward–backward sweep method in optimal control,” Computational Optimization and Applications, vol. 53, no. 1, pp. 207–226, 2012.DOI
  21. [21]H. F. Walker and P. Ni, “Anderson acceleration for fixed-point iterations,” SIAM Journal on Numerical Analysis, vol. 49, no. 4, pp. 1715–1735, 2011.
  22. [22]E. G. Birgin, J. M. Martínez, and M. Raydan, “Nonmonotone spectral projected gradient methods on convex sets,” SIAM Journal on Optimization, vol. 10, no. 4, pp. 1196–1211, 2000.DOI
  23. [23]R. H. Byrd, P. Lu, J. Nocedal, and C. Zhu, “A limited memory algorithm for bound constrained optimization,” SIAM Journal on Scientific Computing, vol. 16, no. 5, pp. 1190–1208, 1995.
  24. [24]C. Zhu, R. H. Byrd, P. Lu, and J. Nocedal, “Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound-constrained optimization,” ACM Transactions on Mathematical Software, vol. 23, no. 4, pp. 550–560, 1997.
  25. [25]G. Izzo, E. Messina, M. Pezzella, and A. Vecchio, “Modified Patankar linear multistep methods for production-destruction systems,” Journal of Scientific Computing, vol. 102, Art. no. 87, 2025. Available: https://link.springer.com/article/10.1007/s10915-025-02804-5.DOI
  26. [26]T. Izgin and P. Öffner, “A study of the local dynamics of modified Patankar DeC and higher order modified Patankar–RK methods,” ESAIM: Mathematical Modelling and Numerical Analysis, vol. 57, no. 4, pp. 2319–2348, 2023. Available: https://arxiv.org/abs/2206.07371v4.
  27. [27]H. Amann, “Existence and regularity for semilinear parabolic evolution equations,” Annali della Scuola Normale Superiore di Pisa, Classe di Scienze, vol. 11, no. 4, pp. 593–676, 1984.
  28. [28]L. C. Evans, Partial Differential Equations, 2nd ed., ser. Graduate Studies in Mathematics, vol. 19. Providence, RI: American Mathematical Society, 2010.
  29. [29]J. Liu and Z. Wang, “Non-commutative discretize-then-optimize algorithms for elliptic PDE-constrained optimal control problems,” Journal of Computational and Applied Mathematics, vol. 362, pp. 596–613, 2019. Available: https://doi.org/10.1016/j.cam.2018.07.028.
  30. [30]J. Harzer, J. De Schutter, and M. Diehl, “Integration error regularization in direct optimal control using embedded Runge Kutta methods,” in Proceedings of the 23rd European Control Conference (ECC), pp. 3065–3070, 2025. Available: https://publications.syscop.de/Harzer2025a.pdf.
  31. [31]C. M. Campos, S. Ober-Blöbaum, and E. Trélat, “High order variational integrators in the optimal control of mechanical systems,” Discrete and Continuous Dynamical Systems, vol. 35, no. 9, pp. 4193–4223, 2015. Available: https://doi.org/10.3934/dcds.2015.35.4193.
  32. [32]J. Burkardt, M. Gunzburger, and J. Peterson, “Insensitive functionals, inconsistent gradients, spurious minima, and regularized functionals in flow optimization problems,” International Journal of Computational Fluid Dynamics, vol. 16, no. 3, pp. 171–185, 2002. Available: https://doi.org/10.1080/10618560290034663.
  33. [33]S. Blanes, A. Iserles, and S. Macnamara, “Positivity-preserving methods for ordinary differential equations,” ESAIM: Mathematical Modelling and Numerical Analysis, vol. 56, no. 6, pp. 1843–1870, 2022. Available: https://doi.org/10.1051/m2an/2022042.
  34. [34]P. Shakya and R. K. Sinha, “Finite element method for parabolic optimal control problems with a bilinear state equation,” Journal of Computational and Applied Mathematics, vol. 367, Art. no. 112431, 2020. Available: https://doi.org/10.1016/j.cam.2019.112431.
  35. [35]E. Casas, K. Chrysafinos, and M. Mateos, “Error estimates for the discretization of bilinear control problems governed by semilinear elliptic PDEs,” Mathematical Control and Related Fields, vol. 15, no. 4, pp. 1320–1345, 2025. Available: https://doi.org/10.3934/mcrf.2024049.
  36. [36]J. V. Burke and J. J. Moré, “On the identification of active constraints,” SIAM Journal on Numerical Analysis, vol. 25, no. 5, pp. 1197–1211, 1988. Available: https://doi.org/10.1137/0725068.
  37. [37]A. W. Leung, “Optimal harvesting-coefficient control of steady-state prey-predator diffusive Volterra–Lotka systems,” Applied Mathematics and Optimization, vol. 31, pp. 219–241, 1995. Available: https://doi.org/10.1007/BF01182789.

Paper details

vlnc:W_a7mdtpgn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

AI tools

View statement
Looking for related papers…
Contents