Abstract
We develop a novel optimal control algorithm to change the phase of an oscillator using a minimum energy input, which also minimizes the oscillator’s transversal distance to the uncontrolled periodic orbit. Our algorithm uses a two-dimensional reduction technique based on both isochrons and isostables. We develop a novel method to eliminate cardiac alternans by connecting our control algorithm with the underlying physiological problem. We also describe how the devised algorithm can be used for spike timing control which can potentially help with motor symptoms of essential and parkinsonian tremor, and aid in treating jet lag. To demonstrate the advantages of this algorithm, we compare it with a previously proposed optimal control algorithm based on standard phase reduction for the Hopf bifurcation normal form, and models for cardiac pacemaker cells, thalamic neurons, and circadian gene regulation cycle in the suprachiasmatic nucleus. We show that our control algorithm is effective even when a large phase change is required or when the nontrivial Floquet multiplier is close to unity; in such cases, the previously proposed control algorithm fails.
Similar content being viewed by others
References
Benabid A, Pollak P, Hoffmann D, Gervason C, Hommel M, Perret J, De Rougemont J, Gao D (1991) Long-term suppression of tremor by chronic stimulation of the ventral intermediate thalamic nucleus. Lancet 337(8738):403–406. https://doi.org/10.1016/0140-6736(91)91175-T
Benabid A, Chabardes S, Mitrofanis J, Pollak P (2009) Deep brain stimulation of the subthalamic nucleus for the treatment of Parkinson’s disease. Lancet Neurol 8(1):67–81. https://doi.org/10.1016/S1474-4422(08)70291-6
Brown E, Moehlis J, Holmes P (2004) On the phase reduction and response dynamics of neural oscillator populations. Neural Comp 16:673–715. https://doi.org/10.1162/089976604322860668’
Castejón O, Guillamon A, Huguet G (2013) Phase-amplitude response functions for transient-state stimuli. J Math Neurosci 3(1):1. https://doi.org/10.1186/2190-8567-3-13
Christini D, Riccio M, Culianu C, Fox J, Karma A, Gilmour R Jr (2006) Control of electrical alternans in canine cardiac Purkinje fibers. Phys Rev Lett 96(10):104101. https://doi.org/10.1103/PhysRevLett.96.104101
Czeisler C, Kronauer R, Allan J, Duffy J, Jewett M, Brown E, Ronda J (1989) Bright light induction of strong (type 0) resetting of the human circadian pacemaker. Science 2:4. https://doi.org/10.1126/science.2734611
Czeisler C, Johnson M, Duffy J, Brown E, Ronda J, Kronauer R (1990) Exposure to bright light and darkness to treat physiologic maladaptation to night work. N Engl J Med 322(18):1253–1259. https://doi.org/10.1056/NEJM199005033221801
Dean D, Forger D, Klerman E et al (2009) Taking the lag out of jet lag through model-based schedule design. PLoS Comput Biol 5(6):E1000418. https://doi.org/10.1371/journal.pcbi.1000418
Eastman C, Martin S (1999) How to use light and dark to produce circadian adaptation to night shift work. Ann Med 31(2):87–98. https://doi.org/10.3109/07853899908998783
Efimov D, Sacré P, Sepulchre R (2009) Controlling the phase of an oscillator: a phase response curve approach. In: Proceedings of the 48h IEEE conference on decision and control (CDC) held jointly with 2009 28th Chinese control conference, pp 7692–7697. https://doi.org/10.1109/CDC.2009.5400901
Ermentrout B (2002) Simulating, analyzing, and animating dynamical systems: a guide to XPPAUT for researchers and students, vol 14. SIAM, Philadelphia
Forger D, Paydarfar D (2004) Starting, stopping, and resetting biological oscillators: in search of optimum perturbations. J Theor Biol 230(4):521–532. https://doi.org/10.1016/j.jtbi.2004.04.043
Gad A, David G, Markus S, Hans R, Charna D, Florian K, Rau IM, Frederick W, Ueli S (2008) SIRT1 regulates circadian clock gene expression through PER2 deacetylation. Cell 134(2):317–328. https://doi.org/10.1016/j.cell.2008.06.050
Glendinning P (1994) Stability, instability and chaos: an introduction to the theory of nonlinear differential equations. Cambridge texts in applied mathematics. Cambridge University Press, Cambridge. https://doi.org/10.1017/CBO9780511626296
Gonze D, Bernard S, Waltermann C, Kramer A, Herzel H (2005) Spontaneous synchronization of coupled circadian oscillators. Biophys J 89(1):120–129. https://doi.org/10.1529/biophysj.104.058388
Guckenheimer J (1975) Isochrons and phaseless sets. J Math Biol 1(3):259–273. https://doi.org/10.1007/BF01273747
Guckenheimer J (1995) Phase portraits of planar vector fields: computer proofs. Exp Math 4(2):153–165. https://doi.org/10.1080/10586458.1995.10504316
Guckenheimer J, Holmes PJ (1983) Nonlinear oscillations. Dynamical systems and bifurcations of vector fields. Springer, New York. https://doi.org/10.1007/978-1-4612-1140-2
Guevara M, Ward G, Shrier A, Glass L (1984) Electrical alternans and period doubling bifurcations. IEEE Comp Cardiol 562:167–170
Guillamon A, Huguet G (2009) A computational and geometric approach to phase resetting curves and surfaces. SIAM J Appl Dyn Syst 8(3):1005–1042. https://doi.org/10.1137/080737666
Hall G, Gauthier D (2002) Experimental control of cardiac muscle alternans. Phys Rev Lett 88(19):198102. https://doi.org/10.1103/PhysRevLett.88.198102
Hall K, Christini D, Tremblay M, Collins J, Glass L, Billette J (1997) Dynamic control of cardiac alternans. Phys Rev Lett 78(23):4518. https://doi.org/10.1103/PhysRevLett.78.4518
Honeycutt R (1992) Stochastic runge–kutta algorithms. i. White noise. Phys Rev A 45:600–603. https://doi.org/10.1103/PhysRevA.45.600
Izhikevich EM (2007) Dynamical systems in neuroscience. MIT Press, Cambridge
Kane A, Hutchison W, Hodaie M, Lozano A, Dostrovsky J (2009) Enhanced synchronization of thalamic theta band local field potentials in patients with essential tremor. Exp Neurol 217(1):171–176. https://doi.org/10.1016/j.expneurol.2009.02.005
Klerman E (2005) Clinical aspects of human circadian rhythms. J Biol Rhythms 20(4):375–386. https://doi.org/10.1177/0748730405278353
Kopell N, Howard L (1973) Plane wave solutions to reaction–diffusion equations. Stud Appl Math 52(4):291–328. https://doi.org/10.1002/sapm1973524291
Kühn A, Tsui A, Aziz T, Ray N, Brücke C, Kupsch A, Schneider GH, Brown P (2009) Pathological synchronisation in the subthalamic nucleus of patients with Parkinson’s disease relates to both bradykinesia and rigidity. Exp Neurol 215(2):380–387. https://doi.org/10.1016/j.expneurol.2008.11.008
Kuramoto Y (1984) Chemical oscillations, waves, and turbulence. Springer, Berlin
Kuramoto Y (1997) Phase-and center-manifold reductions for large populations of coupled oscillators with application to non-locally coupled systems. Int J Bifurc Chaos 7(04):789–805. https://doi.org/10.1142/S0218127497000595
Malkin I (1949) Methods of Poincare and Liapunov in the theory of nonlinear oscillations. Gostexizdat, Moscow
Marsden J, McCracken M (2012) The Hopf bifurcation and its applications, vol 9. Springer, New York. https://doi.org/10.1007/978-1-4612-6374-6
Mauroy A, Mezić I, Moehlis J (2013) Isostables, isochrons, and Koopman spectrum for the action-angle representation of stable fixed point dynamics. Phys D Nonlinear Phenom 261:19–30. https://doi.org/10.1016/j.physd.2013.06.004
Mines G (1913) On dynamic equilibrium in the heart. J Physiol 46(4–5):349–383. https://doi.org/10.1113/jphysiol.1913.sp001596
Minors D, Waterhouse J, Wirz-Justice A (1991) A human phase-response curve to light. Neurosci Lett 133(1):36–40. https://doi.org/10.1016/0304-3940(91)90051-T
Moehlis J, Shea-Brown E, Rabitz H (2006) Optimal inputs for phase models of spiking neurons. ASME J Comput Nonlinear Dyn 1(4):358–367. https://doi.org/10.1115/1.2338654
Monga B, Wilson D, Matchen T, Moehlis J (2018) Phase reduction and phase-based optimal control for biological systems: a tutorial (Under Review)
Nabi A, Moehlis J (2012) Time optimal control of spiking neurons. J Math Biol 64(6):981–1004. https://doi.org/10.1007/s00285-011-0441-5
Nabi A, Mirzadeh M, Gibou F, Moehlis J (2013a) Minimum energy desynchronizing control for coupled neurons. J Comput Neurosci 34(2):259–271. https://doi.org/10.1007/s10827-012-0419-3
Nabi A, Stigen T, Moehlis J, Netoff T (2013b) Minimum energy control for in vitro neurons. J Neural Eng 10(3):036005. https://doi.org/10.1088/1741-2560/10/4/049501
Netoff T, Schwemmer M, Lewis T (2012) Experimentally estimating phase response curves of neurons: theoretical and practical issues. In: Schultheiss N, Prinz A, Butera R (eds) Phase response curves in neuroscience. Springer, New York, pp 95–129. https://doi.org/10.1007/978-1-4614-0739-35
Nolasco J, Dahlen R (1968) A graphic method for the study of alternation in cardiac action potentials. J Appl Physiol 25(2):191–196
Pastore J, Girouard S, Laurita K, Akar F, Rosenbaum D (1999) Mechanism linking t-wave alternans to the genesis of cardiac fibrillation. Circulation 99(10):1385–1394. https://doi.org/10.1161/01.CIR.99.10.1385
Rappel WJ, Fenton F, Karma A (1999) Spatiotemporal control of wave instabilities in cardiac tissue. Phys Rev Lett 83(2):456. https://doi.org/10.1103/PhysRevLett.83.456
Rea M, Bierman A, Figueiro M, Bullough J (2008) A new approach to understanding the impact of circadian disruption on human health. J Circ Rhythms 6(1):7. https://doi.org/10.1186/1740-3391-6-7
Rubin J, Terman D (2004) High frequency stimulation of the subthalamic nucleus eliminates pathological thalamic rhythmicity in a computational model. J Comput Neurosci 16(3):211–235. https://doi.org/10.1023/B:JCNS.0000025686.47117.67
Shirasaka S, Kurebayashi W, Nakao H (2017) Phase-amplitude reduction of transient dynamics far from attractors for limit-cycling systems. Chaos 27(023):119
Snari R, Tinsley M, Wilson D, Faramarzi S, Netoff T, Moehlis J, Showalter K (2015) Desynchronization of stochastically synchronized chemical oscillators. Chaos Interdisc J Nonlinear Sci 25(12):123116. https://doi.org/10.1063/1.4937724
Stigen T, Danzl P, Moehlis J, Netoff T (2011) Controlling spike timing and synchrony in oscillatory neurons. J Neurophysiol 105(5):2074. https://doi.org/10.1186/1471-2202-12-S1-P223
Tass P (2003) A model of desynchronizing deep brain stimulation with a demand-controlled coordinated reset of neural subpopulations. Biol Cybernet 89(2):81–88. https://doi.org/10.1007/s00422-003-0425-7
Tass P (2007) Phase resetting in medicine and biology: stochastic modelling and data analysis, vol 172. Springer, Berlin. https://doi.org/10.1007/978-3-540-38161-7
Wang Q, Young LS (2003) Strange attractors in periodically-kicked limit cycles and Hopf bifurcations. Commun Math Phys 240(3):509–529. https://doi.org/10.1007/s00220-003-0902-9
Wataru K, Shirasaka S, Nakao H (2013) Phase reduction method for strongly perturbed limit cycle oscillators. Phys Rev Lett 111(214):101. https://doi.org/10.1103/PhysRevLett.111.214101
Wedgwood K, Lin K, Thul R, Coombes S (2013) Phase-amplitude descriptions of neural oscillator models. J Math Neurosci 3(1):2. https://doi.org/10.1186/2190-8567-3-2
Wever R (1985) Use of light to treat jet lag: differential effects of normal and bright artificial light on human circadian rhythms. Ann N Y Acad Sci 453(1):282–304. https://doi.org/10.1111/j.1749-6632.1985.tb11818.x
Wilson C, Beverlin B, Netoff T (2011) Chaotic desynchronization as the therapeutic mechanism of deep brain stimulation. Front Syst Neurosci. https://doi.org/10.3389/fnsys.2011.00050
Wilson D, Moehlis J (2014) Optimal chaotic desynchronization for neural populations. SIAM J Appl Dyn Syst 13(1):276. https://doi.org/10.1137/120901702
Wilson D, Moehlis J (2015a) Clustered desynchronization from high-frequency deep brain stimulation. PLoS Comput Biol 11(12):e1004673. https://doi.org/10.1371/journal.pcbi.1004673
Wilson D, Moehlis J (2015b) Extending phase reduction to excitable media: theory and applications. SIAM Rev 57(2):201–222. https://doi.org/10.1137/140952478
Wilson D, Moehlis J (2016) Isostable reduction of periodic orbits. Phys Rev E 94(5):052213. https://doi.org/10.1103/PhysRevE.94.052213
Winfree A (1967) Biological rhythms and the behavior of populations of coupled oscillators. J Theor Biol 16(1):15–42. https://doi.org/10.1016/0022-5193(67)90051-3
Winfree A (2001) The geometry of biological time, vol Second. Springer, New York. https://doi.org/10.1007/978-1-4757-3484-3
Yanagihara K, Noma A, Irisawa H (1980) Reconstruction of sino-atrial node pacemaker potential based on the voltage clamp experiments. Jpn J Physiol 30(6):841–857. https://doi.org/10.2170/jjphysiol.30.841
Zhang J, Wen J, Julius A (2012) Optimal circadian rhythm control with light input for rapid entrainment and improved vigilance. In: Proceedings of the 51st IEEE conference on decision and control (CDC). IEEE, pp 3007–3012. https://doi.org/10.1109/CDC.2012.6426226
Zlotnik A, Li J (2014) Optimal subharmonic entrainment of weakly forced nonlinear oscillators. SIAM J Appl Dyn Syst 13(4):1654–1693. https://doi.org/10.1137/140952211
Zlotnik A, Chen Y, Kiss I, Tanaka H, Li JS (2013) Optimal waveform for fast entrainment of weakly forced nonlinear oscillators. Phys Rev Lett 111(2):024102. https://doi.org/10.1103/PhysRevLett.111.024102
Acknowledgements
This work was supported by National Science Foundation Grants Nos. NSF-1363243 and NSF-1635542. We thank Dan Wilson for helpful discussions on numerical computation of the augmented phase reduction.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of Interest
The authors declare that they have no conflict of interest.
Additional information
Communicated by Auke Jan Ijspeert.
This article belongs to the Special Issue on Control Theory in Biology and Medicine. It derived from a workshop at the Mathematical Biosciences Institute, Ohio State University, Columbus, OH, USA.
Appendices
A Numerical Methods
In this appendix, we give details on the numerical methods we used to compute the Floquet multipliers, PRC, and IRC, and solve the Euler Lagrange equations and the full model equations.
1.1 A.1 Computation of PRC
For the normal form of the Hopf bifurcation, we can compute the PRC and its derivative w.r.t. \(\theta \) analytically, see, e.g., Brown et al. (2004). For computing the PRCs (and their derivatives w.r.t. \(\theta \)) of the YNI, thalamic neuron, and the clock gene regulation model, we use the XPP package (Ermentrout 2002), which is widely used by the community working on nonlinear oscillators. This package solves the appropriate adjoint equation backward in time along the periodic orbit to compute the PRC as a function of time. We scale the PRC computed by this package by \(\omega \), as we consider PRC as \(\mathcal {Z}(\theta )=\frac{\partial \theta }{\partial \mathbf {x}}\), whereas the computed PRC from the XPP package is \(\tilde{\mathcal {Z}}(t)=\frac{\partial t}{\partial \mathbf {x}}\). Note that the XPP computes the derivative of the PRC w.r.t. time \(\left( \dot{\tilde{\mathcal {Z}}}(t)=\frac{\partial ^2 t}{\partial \mathbf {x}\partial t}\right) \), which is numerically equivalent to its derivative w.r.t. \(\theta \)\(\left( \mathcal {Z}'(\theta )=\frac{\partial ^2 \theta }{ \partial \mathbf {x}\partial \theta }\right) \). The XPP package gives the PRC and its derivative as a time series. After appropriately scaling the time series, we write them as an analytical expression of \(\theta \) by approximating them as a finite Fourier series, to be used in the numerical computation of the Euler–Lagrange equations.
1.2 A.2 Computation of Floquet multipliers
Once the PRC has been computed, we choose an arbitrary point on the periodic orbit as \(\theta =0\) and approximate the isochron \(\varGamma _0\) as an \(n-1\) dimensional hyperplane orthogonal to the PRC at that point. To compute the Jacobian DF, we compute \(\mathbf {x}_\varGamma ^j\) (as defined beneath Eq. 8 in the main text) for a large j, for a number of initial conditions \(\mathbf {x}_0\) spread out on the isochron. Eigenvector decomposition of DF gives us the Floquet multipliers of the periodic orbit and the corresponding Floquet exponents \(k_i\). Note that for planar systems, the nontrivial Floquet exponent can be directly computed from the divergence of the vector field as (Glendinning 1994)
1.3 A.3 Two point boundary value problem with Newton iteration
We calculate the IRC and solve the Euler–Lagrange equations as a two point boundary value problem using Newton iteration, which we briefly summarize. Consider a general two point boundary value problem
with the linear boundary condition
To solve such a boundary value problem, we integrate Eq. (44) with the initial guess \(c=y(0)\) and calculate the function g(c):
where y(b) is the solution at time b with the initial condition c. If we had chosen the correct initial condition c, g(c) would be 0. Based on the current guess \(c^\nu \), and the \(g(c^\nu )\) value, we choose the next initial condition by the Newton Iteration as
We compute the Jacobian \(J=\left. \frac{\partial g}{\partial c}\right| _{c^\nu }\) numerically as
where
\(J_i\) is the \(i\mathrm{th}\) column of J, \(\epsilon \) is a small number, and \(e_i\) is a column vector with 1 in the \(i\mathrm{th}\) position and 0 elsewhere.
1.3.1 A.3.1 Computation of IRC
To calculate the IRC, we first compute and save the periodic solution \(\gamma (t)\) using Matlab’s ODE solver ode45 with a relative error tolerance of \(3e-12\), and an absolute error tolerance of \(1e-15\). The next step is to solve the adjoint equation
with periodic boundary conditions
We choose an initial guess \({\mathcal I}(0)\), and integrate the adjoint equation using Matlab’s ODE solver ode45 with a relative error tolerance of \(3e-12\), and an absolute error tolerance of \(1e-15\). For Newton iteration, we take
where I is the identity matrix, and J is the Jacobian matrix
which we compute numerically. We use Eqs. (46)–(48) together with Eq. (45) to compute the next initial condition. Once a periodic solution is obtained, the computed IRC is scaled by the normalization condition \(\nabla _{\mathbf {x_0}}\psi _i \cdot v_i=1\) (Wilson and Moehlis 2016). Its derivative w.r.t. \(\theta \) is obtained numerically by a central difference scheme
The obtained IRC and its derivative w.r.t. \(\theta \) are written as analytical expressions of \(\theta \) by a finite Fourier series approximation, which is used in the computation of the Euler–Lagrange equations.
1.3.2 A.3.2 Solving Euler–Lagrange equations
For Euler Lagrange equations based on augmented phase reduction, we set the boundary conditions as \(\theta (0)=0, \ \theta (T_1)=2\pi ,\ \psi (0)=0,\ \psi (T_1)=0\). We can write this as a two point boundary value problem with the function g as
Since \(\theta (0)\), and \(\psi (0)\) are fixed by our problem, g can be influenced by changing \(\lambda _1(0)\) and \(\lambda _2(0)\) only. So we get the following matrices for Newton Iteration:
In a similar way, we get the following matrices for Euler–Lagrange equations based on standard phase reduction:
All the integrations are done with Matlab ODE solver ode45 with relative error tolerance \(\le 1e-10\) and absolute error tolerance \(\le 1e-10\).
B Models
In this appendix, we give details of the mathematical models used and also their augmented and standard phase reduction models, which are necessary to reproduce the results of this article.
1.1 B.1 YNI model
Here we list the both full and reduced model parameters of the YNI model (Yanagihara et al. 1980) introduced in Sect. 4.2.2.
1.1.1 B.1.1 Full model equations and parameters
The full YNI model is given as
where
1.1.2 B.1.2 Reduced model equations and parameters
For the augmented and standard phase reduction of the YNI model, we get \(\omega =0.03088\), \(k=-0.00135\). Once the PRC, IRC, and their derivatives w.r.t. \(\theta \) are numerically computed (see “Appendix A”), we approximate them as finite Fourier series to be used as an analytical function in the numerical computation of Euler–Lagrange equations. \(\theta =0\) corresponds to the initial condition \(V=-19.2803, \ d= 0.6817, \ f=0.0236, \ m=0.8540, \ h=0.0013, \ q=0.0038, \ p= 0.6592\).
1.2 B.2 Thalamic neuron model
Here we list both full and reduced model parameters of the thalamus model (Rubin and Terman 2004) used in Section 4.3.2
1.2.1 B.2.1 Full model equations and parameters
The full thalamic neuron model is given as
where
1.2.2 B.2.2 Reduced model equations and parameters
For the augmented and standard phase reduction of the thalamic neuron model, we get \(\omega =0.7484\), \(k=-0.0225\). Once the PRC, IRC, and their derivatives w.r.t. \(\theta \) are numerically computed (see “Appendix A”), we approximate them as finite Fourier series to be used as an analytical function in the numerical computation of Euler–Lagrange equations. \(\theta =0\) corresponds to the initial condition \(v=-57.5298,\ h=0.1424, \ r= 0.0017\).
1.3 B.3 Clock gene regulation model
Here we list the both full and reduced model parameters of the clock gene regulation model (Gonze et al. 2005) used in Sect. 4.4.2.
1.3.1 Full model equations and parameters
The full thalamus model is given as
1.3.2 Reduced model equations and parameters
For the augmented and standard phase reduction of the clock gene regulation model, we get \(\omega =0.2669\), \(k=-0.0021\). Once the PRC, IRC, and their derivatives w.r.t. \(\theta \) are numerically computed (see “Appendix A”), we approximate them as finite Fourier series to be used as an analytical function in the numerical computation of Euler–Lagrange equations. \(\theta =0\) corresponds to the initial condition \(X=0.1948, \ Y=0.4154,\ Z=1.8530\).
Rights and permissions
About this article
Cite this article
Monga, B., Moehlis, J. Optimal phase control of biological oscillators using augmented phase reduction. Biol Cybern 113, 161–178 (2019). https://doi.org/10.1007/s00422-018-0764-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00422-018-0764-z