-
Inverse Problems Over Probability Measure Space
Authors:
Qin Li,
Maria Oprea,
Li Wang,
Yunan Yang
Abstract:
Define a forward problem as $ρ_y = G_\#ρ_x$, where the probability distribution $ρ_x$ is mapped to another distribution $ρ_y$ using the forward operator $G$. In this work, we investigate the corresponding inverse problem: Given $ρ_y$, how to find $ρ_x$? Depending on whether $ G$ is overdetermined or underdetermined, the solution can have drastically different behavior. In the overdetermined case,…
▽ More
Define a forward problem as $ρ_y = G_\#ρ_x$, where the probability distribution $ρ_x$ is mapped to another distribution $ρ_y$ using the forward operator $G$. In this work, we investigate the corresponding inverse problem: Given $ρ_y$, how to find $ρ_x$? Depending on whether $ G$ is overdetermined or underdetermined, the solution can have drastically different behavior. In the overdetermined case, we formulate a variational problem $\min_{ρ_x} D( G_\#ρ_x, ρ_y)$, and find that different choices of the metric $ D$ significantly affect the quality of the reconstruction. When $ D$ is set to be the Wasserstein distance, the reconstruction is the marginal distribution, while setting $ D$ to be a $φ$-divergence reconstructs the conditional distribution. In the underdetermined case, we formulate the constrained optimization $\min_{\{ G_\#ρ_x=ρ_y\}} E[ρ_x]$. The choice of $ E$ also significantly impacts the construction: setting $ E$ to be the entropy gives us the piecewise constant reconstruction, while setting $ E$ to be the second moment, we recover the classical least-norm solution. We also examine the formulation with regularization: $\min_{ρ_x} D( G_\#ρ_x, ρ_y) + α\mathsf R[ρ_x]$, and find that the entropy-entropy pair leads to a regularized solution that is defined in a piecewise manner, whereas the $W_2$-$W_2$ pair leads to a least-norm solution where $W_2$ is the 2-Wasserstein metric.
△ Less
Submitted 26 April, 2025;
originally announced April 2025.
-
Symplectic Geometry in Hybrid and Impulsive Optimal Control
Authors:
William Clark,
Maria Oprea
Abstract:
Hybrid dynamical systems are systems which undergo both continuous and discrete transitions. The Bolza problem from optimal control theory is applied to these systems and a hybrid version of Pontryagin's maximum principle is presented. This hybrid maximum principle is presented to emphasize its geometric nature which makes its study amenable to the tools of geometric mechanics and symplectic geome…
▽ More
Hybrid dynamical systems are systems which undergo both continuous and discrete transitions. The Bolza problem from optimal control theory is applied to these systems and a hybrid version of Pontryagin's maximum principle is presented. This hybrid maximum principle is presented to emphasize its geometric nature which makes its study amenable to the tools of geometric mechanics and symplectic geometry. One explicit benefit of this geometric approach is that the symplectic structure (and hence the induced volume) is preserved. This allows for a hybrid analog of caustics and conjugate points. Additionally, an introductory analysis of singular solutions (beating and Zeno) is discussed geometrically. This work concludes on a biological example where beating can occur.
△ Less
Submitted 21 April, 2025;
originally announced April 2025.
-
The Distributional Koopman Operator for Random Dynamical Systems
Authors:
Maria Oprea,
Alex Townsend,
Yunan Yang
Abstract:
The Distributional Koopman Operator (DKO) is introduced as a way to perform Koopman analysis on random dynamical systems where only aggregate distribution data is available, thereby eliminating the need for particle tracking or detailed trajectory data. Our DKO generalizes the stochastic Koopman operator (SKO) to allow for observables of probability distributions, using the transfer operator to pr…
▽ More
The Distributional Koopman Operator (DKO) is introduced as a way to perform Koopman analysis on random dynamical systems where only aggregate distribution data is available, thereby eliminating the need for particle tracking or detailed trajectory data. Our DKO generalizes the stochastic Koopman operator (SKO) to allow for observables of probability distributions, using the transfer operator to propagate these probability distributions forward in time. Like the SKO, the DKO is linear with semigroup properties, and we show that the dynamical mode decomposition (DMD) approximation can converge to the DKO in the large data limit. The DKO is particularly useful for random dynamical systems where trajectory information is unavailable.
△ Less
Submitted 15 April, 2025;
originally announced April 2025.
-
Stochastic Inverse Problem: stability, regularization and Wasserstein gradient flow
Authors:
Qin Li,
Maria Oprea,
Li Wang,
Yunan Yang
Abstract:
Inverse problems in physical or biological sciences often involve recovering an unknown parameter that is random. The sought-after quantity is a probability distribution of the unknown parameter, that produces data that aligns with measurements. Consequently, these problems are naturally framed as stochastic inverse problems. In this paper, we explore three aspects of this problem: direct inversio…
▽ More
Inverse problems in physical or biological sciences often involve recovering an unknown parameter that is random. The sought-after quantity is a probability distribution of the unknown parameter, that produces data that aligns with measurements. Consequently, these problems are naturally framed as stochastic inverse problems. In this paper, we explore three aspects of this problem: direct inversion, variational formulation with regularization, and optimization via gradient flows, drawing parallels with deterministic inverse problems. A key difference from the deterministic case is the space in which we operate. Here, we work within probability space rather than Euclidean or Sobolev spaces, making tools from measure transport theory necessary for the study. Our findings reveal that the choice of metric -- both in the design of the loss function and in the optimization process -- significantly impacts the stability and properties of the optimizer.
△ Less
Submitted 30 September, 2024;
originally announced October 2024.
-
Measure-Theoretic Time-Delay Embedding
Authors:
Jonah Botvinick-Greenhouse,
Maria Oprea,
Romit Maulik,
Yunan Yang
Abstract:
The celebrated Takens' embedding theorem provides a theoretical foundation for reconstructing the full state of a dynamical system from partial observations. However, the classical theorem assumes that the underlying system is deterministic and that observations are noise-free, limiting its applicability in real-world scenarios. Motivated by these limitations, we rigorously establish a measure-the…
▽ More
The celebrated Takens' embedding theorem provides a theoretical foundation for reconstructing the full state of a dynamical system from partial observations. However, the classical theorem assumes that the underlying system is deterministic and that observations are noise-free, limiting its applicability in real-world scenarios. Motivated by these limitations, we rigorously establish a measure-theoretic generalization that adopts an Eulerian description of the dynamics and recasts the embedding as a pushforward map between probability spaces. Our mathematical results leverage recent advances in optimal transportation theory. Building on our novel measure-theoretic time-delay embedding theory, we have developed a new computational framework that forecasts the full state of a dynamical system from time-lagged partial observations, engineered with better robustness to handle sparse and noisy data. We showcase the efficacy and versatility of our approach through several numerical examples, ranging from the classic Lorenz-63 system to large-scale, real-world applications such as NOAA sea surface temperature forecasting and ERA5 wind field reconstruction.
△ Less
Submitted 13 September, 2024;
originally announced September 2024.
-
Optimal Control of Reduced Left-Invariant Hybrid Control Systems
Authors:
William Clark,
Maria Oprea
Abstract:
Optimal control is ubiquitous in many fields of engineering. A common technique to find candidate solutions is via Pontryagin's maximum principle. An unfortunate aspect of this method is that the dimension of system doubles. When the system evolves on a Lie group and the system is invariant under left (or right) translations, Lie-Poisson reduction can be applied to eliminate half of the dimensions…
▽ More
Optimal control is ubiquitous in many fields of engineering. A common technique to find candidate solutions is via Pontryagin's maximum principle. An unfortunate aspect of this method is that the dimension of system doubles. When the system evolves on a Lie group and the system is invariant under left (or right) translations, Lie-Poisson reduction can be applied to eliminate half of the dimensions (and returning the dimension of the problem to the back to the original number).
Hybrid control systems are an extension of (continuous) control systems by allowing for sudden changes to the state. Examples of such systems include the bouncing ball - the velocity instantaneously jumps during a bounce, the thermostat - controls switch to on or off, and a sailboat undergoing tacking. The goal of this work is to extend the idea of Lie-Poisson reduction to the optimal control of these systems. If $n$ is the dimension of the original system, $2n$ is the dimension of the system produced by the maximum principle. In the case of classical Lie-Poisson reduction, the dimension drops back down to $n$. This, unfortunately, is impossible in hybrid systems as there must be an auxiliary variable encoding whether or not an event occurs. As such, the analogous hybrid Lie-Poisson reduction results in a $n+1$ dimensional system. The purpose of this work is to develop and present this technique.
△ Less
Submitted 22 March, 2024;
originally announced March 2024.
-
Optimal Control of Hybrid Systems with Submersive Resets
Authors:
William Clark,
Maria Oprea,
Aden Shaw
Abstract:
Hybrid dynamical systems are systems which posses both continuous and discrete transitions. Assuming that the discrete transitions (resets) occur a finite number of times, the optimal control problem can be solved by gluing together the optimal arcs from the underlying continuous problem via the "Hamilton jump conditions." In most cases, it is assumed that the reset is a diffeomorphism (onto its i…
▽ More
Hybrid dynamical systems are systems which posses both continuous and discrete transitions. Assuming that the discrete transitions (resets) occur a finite number of times, the optimal control problem can be solved by gluing together the optimal arcs from the underlying continuous problem via the "Hamilton jump conditions." In most cases, it is assumed that the reset is a diffeomorphism (onto its image) and the corresponding Hamilton jump condition admits a unique solution. However, in many applications, the reset results in a drop in dimension and the corresponding Hamilton jump condition admits zero/infinitely many solutions. A geometric interpretation of this issue is explored in the case where the reset is a submersion (onto its image). Necessary conditions are presented for this case along with an accompanying numerical example.
△ Less
Submitted 19 March, 2024; v1 submitted 25 January, 2024;
originally announced January 2024.
-
A Study of the Long-Term Behavior of Hybrid Systems with Symmetries via Reduction and the Frobenius-Perron Operator
Authors:
Maria Oprea,
Aden Shaw,
Robi Huq,
Kaito Iwasaki,
Dora Kassabova,
W. Clark
Abstract:
Hybrid dynamical systems are systems which undergo both continuous and discrete transitions. As typical in dynamical analysis, an essential goal is to study the long-term behavior of these systems. In this work, we present two different novel approaches for studying these systems. The first approach is based on constructing an analog of the Frobenius-Perron (transport) operator for hybrid systems.…
▽ More
Hybrid dynamical systems are systems which undergo both continuous and discrete transitions. As typical in dynamical analysis, an essential goal is to study the long-term behavior of these systems. In this work, we present two different novel approaches for studying these systems. The first approach is based on constructing an analog of the Frobenius-Perron (transport) operator for hybrid systems. Rather than tracking the evolution of a single trajectory, this operator encodes the asymptotic nature of an ensemble of trajectories. The second approach presented applies to an important subclass of hybrid systems, mechanical impact systems. We develop an analog of Lie-Poisson(-Suslov) reduction for left-invariant impact systems on Lie groups. In addition to the Hamiltonian (and constraints) being left-invariant, the impact surface must also be a right coset of a normal subgroup. This procedure allows a reduction from a $2n$-dimensional system to an $(n+1)$-dimensional one. We conclude the paper by presenting numerical results on a diverse array of applications.
△ Less
Submitted 21 September, 2023;
originally announced September 2023.
-
Learning the Delay Using Neural Delay Differential Equations
Authors:
Maria Oprea,
Mark Walth,
Robert Stephany,
Gabriella Torres Nothaft,
Arnaldo Rodriguez-Gonzalez,
William Clark
Abstract:
The intersection of machine learning and dynamical systems has generated considerable interest recently. Neural Ordinary Differential Equations (NODEs) represent a rich overlap between these fields. In this paper, we develop a continuous time neural network approach based on Delay Differential Equations (DDEs). Our model uses the adjoint sensitivity method to learn the model parameters and delay d…
▽ More
The intersection of machine learning and dynamical systems has generated considerable interest recently. Neural Ordinary Differential Equations (NODEs) represent a rich overlap between these fields. In this paper, we develop a continuous time neural network approach based on Delay Differential Equations (DDEs). Our model uses the adjoint sensitivity method to learn the model parameters and delay directly from data. Our approach is inspired by that of NODEs and extends earlier neural DDE models, which have assumed that the value of the delay is known a priori. We perform a sensitivity analysis on our proposed approach and demonstrate its ability to learn DDE parameters from benchmark systems. We conclude our discussion with potential future directions and applications.
△ Less
Submitted 14 June, 2023; v1 submitted 3 April, 2023;
originally announced April 2023.
-
Optimality of Zeno Executions in Hybrid Systems
Authors:
William Clark,
Maria Oprea
Abstract:
A unique feature of hybrid dynamical systems (systems whose evolution is subject to both continuous- and discrete-time laws) is Zeno trajectories. Usually these trajectories are avoided as they can cause incorrect numerical results as the problem becomes ill-conditioned. However, these are difficult to justifiably avoid as determining when and where they occur is a non-trivial task. It turns out t…
▽ More
A unique feature of hybrid dynamical systems (systems whose evolution is subject to both continuous- and discrete-time laws) is Zeno trajectories. Usually these trajectories are avoided as they can cause incorrect numerical results as the problem becomes ill-conditioned. However, these are difficult to justifiably avoid as determining when and where they occur is a non-trivial task. It turns out that in optimal control problems, not only can they not be avoided, but are sometimes required in synthesizing the solutions. This work explores the pedagogical example of the bouncing ball to demonstrate the importance of "Zeno control executions."
△ Less
Submitted 3 October, 2022;
originally announced October 2022.
-
Optimal Control of Nonholonomic Systems via Magnetic Fields
Authors:
Maria Oprea,
Max Ruth,
Dora Kassabova,
William Clark
Abstract:
Geometric optimal control utilizes tools from differential geometry to analyze the structure of a problem to determine the control and state trajectories to reach a desired outcome while minimizing some cost function. For a controlled mechanical system, the control usually manifests as an external force which, if conservative, can be added to the Hamiltonian. In this work, we focus on mechanical s…
▽ More
Geometric optimal control utilizes tools from differential geometry to analyze the structure of a problem to determine the control and state trajectories to reach a desired outcome while minimizing some cost function. For a controlled mechanical system, the control usually manifests as an external force which, if conservative, can be added to the Hamiltonian. In this work, we focus on mechanical systems with controls added to the symplectic form rather than the Hamiltonian. In practice, this translates to controlling the magnetic field for an electrically charged system. We develop a basic theory deriving necessary conditions for optimality of such a system subjected to nonholonomic constraints. We consider the representative example of a magnetically charged Chaplygin Sleigh, whose resulting optimal control problem is completely integrable.
△ Less
Submitted 19 September, 2022;
originally announced September 2022.
-
How do we walk? Using hybrid holonomy to approximate non-holonomic systems
Authors:
Maria Oprea,
William Clark
Abstract:
Why do we move forward when we walk? Our legs undergo periodic motion and thus possess no net change in position; however, our bodies do possess a net change in position and we are propelled forward. From a geometric perspective, this phenomenon of periodic input producing non-periodic output is holonomy. To obtain non-zero holonomy and propel forward, we must alternate which leg is in contact wit…
▽ More
Why do we move forward when we walk? Our legs undergo periodic motion and thus possess no net change in position; however, our bodies do possess a net change in position and we are propelled forward. From a geometric perspective, this phenomenon of periodic input producing non-periodic output is holonomy. To obtain non-zero holonomy and propel forward, we must alternate which leg is in contact with the ground; a non-zero net motion can be obtained by concatenating arcs that would individually produce no net motion. We develop a framework for computing the holonomy group of hybrid systems and analyze their behavior in the limit as the number of impacts goes to infinity.
△ Less
Submitted 30 March, 2022;
originally announced March 2022.
-
A Geometric Approach to Optimal Control of Hybrid and Impulsive Systems
Authors:
William Clark,
Maria Oprea,
Andrew J. Graven
Abstract:
Hybrid dynamical systems are systems which undergo both continuous and discrete transitions. The Bolza problem from optimal control theory is applied to these systems and a hybrid version of Pontryagin's maximum principle is presented. This hybrid maximum principle is presented to emphasize its geometric nature which makes its study amenable to the tools of geometric mechanics and symplectic geome…
▽ More
Hybrid dynamical systems are systems which undergo both continuous and discrete transitions. The Bolza problem from optimal control theory is applied to these systems and a hybrid version of Pontryagin's maximum principle is presented. This hybrid maximum principle is presented to emphasize its geometric nature which makes its study amenable to the tools of geometric mechanics and symplectic geometry. One explicit benefit of this geometric approach is that Zeno behavior can be strongly controlled for "generic" control problems. Moreover, when the underlying control system is a mechanical impact system, additional structure is present which can be exploited and is thus explored. Multiple examples are presented for both mechanical and non-mechanical systems.
△ Less
Submitted 22 November, 2021;
originally announced November 2021.
