Variational Integrators

• Variational integrators are geometric numerical methods that discretize Hamilton's principle, preserving symplectic structure, momentum, and near energy conservation.
• They employ frameworks such as Galerkin, shooting-based, and Lie group methods to achieve high-order accuracy and structure preservation.
• Their applications span complex mechanical systems, robotics, optimal control, and physical field theories for long-time accurate simulations.

Variational integrators are geometric numerical methods for time integration of Lagrangian and Hamiltonian dynamical systems, constructed by discretizing Hamilton’s principle of stationary action rather than directly discretizing the equations of motion. As a consequence, these methods preserve crucial geometric structures of the continuous dynamics at the discrete level—including symplecticity, momentum conservation, and near conservation of energy—thus enabling long-time accurate simulations of complex mechanical systems, control problems, and physical field theories (Colombo et al., 8 Feb 2025, Leok et al., 2011, Leok et al., 2010, Hall et al., 2012, Hall et al., 2014).

1. Discretization of Hamilton's Principle

At the core of variational integrator (VI) construction is the discretization of Hamilton’s principle: $\delta \int_{0}^{T} L(q, \dot{q})\,dt = 0$ where $L: TQ \to \mathbb{R}$ is the Lagrangian on configuration manifold $Q$. This leads to the Euler–Lagrange equations in the continuous setting. The discrete analogue replaces the action integral by a discrete action sum using a discrete Lagrangian $L_d(q_k, q_{k+1}; h)$: $$S_d(q_0, ..., q_N) = \sum_{k=0}^{N-1} L_d(q_k, q_{k+1}; h)$$ Stationarity of $S_d$ under variations with fixed endpoints gives the discrete Euler–Lagrange (DEL) equations: $D_2 L_d(q_{k-1}, q_k) + D_1 L_d(q_k, q_{k+1}) = 0$ This defines a second-order implicit integrator, with the map $(q_{k-1}, q_k) \mapsto (q_k, q_{k+1})$ (Leok et al., 2011, Hall et al., 2012, Leok et al., 2010).

For Lie group configuration spaces, the principle generalizes using group-reduced coordinates and retraction maps $\mathcal{R}: \mathfrak{g} \to G$, preserving group structure exactly (Colombo et al., 8 Feb 2025, Hall et al., 2014).

2. Structure-Preserving Properties

Variational integrators are inherently geometric:

• Symplecticity: The discrete flow preserves a symplectic or multisymplectic structure derived from the variational principle, guaranteeing long-time bounded energy error (no secular drift) (Leok et al., 2011, Hall et al., 2012, Hall et al., 2014, Leok et al., 2010).
• Momentum Conservation: For systems with symmetries (Lie group actions), a discrete Noether theorem ensures exact preservation of discrete momentum maps when the discrete Lagrangian is group-invariant (Colombo et al., 8 Feb 2025, Leok et al., 2011, Leok et al., 2010, Hall et al., 2014).
• Energy Behavior: Although VIs are not generically energy-preserving, they exhibit near-conservation—energy errors remain oscillatory and bounded over exponentially long times. In special cases (quadratic Lagrangians, symplectic midpoint), exact energy conservation is achieved (Ober-Blöbaum et al., 2011, Hall et al., 2012, Hall et al., 2014).
• Backward Error Analysis: The discrete map corresponds to the exact flow of a modified Hamiltonian, implying qualitative accuracy over long integrations (Leok et al., 2011, Hall et al., 2012).

3. Construction Methodologies

Multiple systematic frameworks have been established for constructing variational integrators, each admitting high-order and structure-preserving generalizations (Leok et al., 2011, Hall et al., 2012, Hall et al., 2014):

Construction	Key Ideas and Features
Galerkin Variational	Approximate exact discrete Lagrangian via trial functions/basis, apply quadrature
Shooting-Based	Solve ODE boundary value problem using one-step schemes; approximate action via quadrature
Spectral/Galerkin	Use polynomial or spectral basis for solution curves; obtain geometric convergence in $n$
Lie Group Integrators	Interpolate in Lie algebra; map to group via exponential/Cayley; preserve group operations
Taylor Variational	Use Taylor expansions in shooting; gain an extra order of accuracy per Taylor step

For systems with external forcing or stochastic influences, extended variational principles (Lagrange–d’Alembert type) yield forced or stochastic VIs (Kraus et al., 2019, Holm et al., 2016, Ober-Blöbaum et al., 2011).

In higher-order or constrained systems (e.g., for underactuated control), the principle is posed on higher discrete jet bundles or includes Lagrange multipliers for constraints, preserving structure and constraints simultaneously (Colombo et al., 2013, Colombo et al., 2012).

4. Applications in Mechanics, Control, and Physical Systems

Variational integrators have been developed for a wide range of physical and engineering systems:

• Multibody and Robotic Systems: Efficient O(n)-time VIs with recursive Newton–Euler evaluation enable scalable simulation of high-DOF systems (e.g., humanoid robots), maintaining energy and momentum (Lee et al., 2016).
• Optimal Control: Direct collocation with VIs leads to nonlinear programs whose KKT systems are consistent with discrete Pontryagin optimality conditions; commutation results guarantee equivalence of dualization and discretization (Campos et al., 2015, Colombo et al., 8 Feb 2025).
• Underactuated and Symmetric Systems: VIs naturally accommodate systems on principal bundles, underactuation, and higher-order constraints via discrete reduction and DAH principles (Colombo et al., 2012, Colombo et al., 8 Feb 2025, Colombo et al., 2013).
• Electric Circuits: VIs applied to circuits yield good energy and frequency spectrum preservation even for degenerate Lagrangians under constraints (KCL/KVL) and dissipation (Ober-Blöbaum et al., 2011).
• Thermoelastic Solids and PDEs: VIs extend to continuum systems, obtaining multisymplectic integrators for field theories, exact conservation laws for discrete entropy, momentum, and nearly energy (even under high spatial discretization) (A et al., 2014, Kraus et al., 2014).
• Stochastic and Dissipative Systems: Stochastic variational integrators derive from discrete stochastic action principles, preserving symplecticity and Noether invariants in expectation, with superior long-time stability (Kraus et al., 2019, Holm et al., 2016).
• Contact Dynamics and Dissipative Systems: Contact variational integrators based on Herglotz's principle produce one-step maps that preserve a conformal contact structure and dissipate energy at correct rates (Vermeeren et al., 2019).
• Plasma Physics and Nonvariational PDEs: Extension via formal Lagrangians and adjoint fields enables VIs for equations lacking a classical variational form, preserving energy and momentum up to machine precision in guiding center, Vlasov, and MHD systems (Kraus, 2013, Kraus et al., 2014).
• Spectral and High-Order Integrators: Galerkin polynomial and spectral VIs provide arbitrarily high order by increasing basis degree, with optimal or geometric convergence rates and structure preservation (Hall et al., 2012, Hall et al., 2014).
• Time-Dependent (Nonautonomous) Systems: V