EqualMotion: Diverse Motion Control Approaches

• EqualMotion is a multifaceted term defining distinct frameworks that achieve controlled motion relative to target geometries or energy distributions.
• It spans topics from engineered swift equilibration in stochastic thermodynamics to adaptive, accessible wearable motion-capture systems.
• These approaches leverage novel control forces, moving meshes, and co-design principles to enable precise tracking and inclusivity in dynamic environments.

EqualMotion denotes several distinct research constructs in recent technical literature rather than a single unified framework. In the supplied arXiv corpus, the term is attached to or used to interpret four different objects: engineered swift equilibration for overdamped Brownian systems on arbitrary curved configuration spaces (Frim et al., 2020); Lyapunov analysis of trajectories relative to time-varying nonisolated equilibrium families in nonautonomous ODEs (Saoud, 25 Jun 2026); energy-equidistributed moving sampling for PINNs solving conservative PDEs (Gao et al., 27 Aug 2025); and a body-agnostic wearable motion-capture system for disabled creative practitioners in the creative industries (Hilton et al., 11 Jul 2025). A likely source of ambiguity is that these uses share a common emphasis on motion relative to a target geometry, distribution, or representation, but they are methodologically and disciplinarily distinct.

1. Term, scope, and conceptual family

In stochastic thermodynamics, EqualMotion names the problem of forcing a stochastic system to move between equilibrium states as if it were evolving quasistatically, but in finite time, by adding suitable control forces (Frim et al., 2020). In dynamical-systems theory, the closely related phrase “equilibria in motion” refers to stability, tracking, and convergence relative to a moving equilibrium family $\{E(t)\}_{t \ge t_0}$, where the target is not a fixed point but a time-varying set (Saoud, 25 Jun 2026). In scientific machine learning, EqualMotion has been used to describe collocation points that move so that energy is equally distributed across the mesh, yielding energy-equidistributed moving sampling PINNs (Gao et al., 27 Aug 2025). In HCI and creative technology, “EqualMotion” is the title of a body-agnostic, wearable motion-capture system designed through a disability-centred co-design approach (Hilton et al., 11 Jul 2025).

These meanings should not be conflated. The first three concern control, tracking, or redistribution relative to evolving equilibrium or energy structures; the fourth concerns inclusive motion-capture hardware, calibration, and representation. This suggests a family resemblance at the level of “motion made commensurate with a target structure,” but not a shared technical formalism.

2. EqualMotion in stochastic thermodynamics

In "Engineered swift equilibration for arbitrary geometries" (Frim et al., 2020), the relevant problem is overdamped Brownian motion under rapid parameter variation. For a controlled potential $V(x;\lambda(t))$, the instantaneous equilibrium distribution is

$$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$

Engineered swift equilibration introduces an auxiliary term, equivalently an auxiliary force $\mathbf{F}_{\mathrm{ext}}$, such that the actual distribution satisfies

$\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$

for all $t$, not only at the endpoints. The paper generalizes this from earlier one-dimensional and Euclidean settings to generic overdamped Brownian systems on an arbitrary compact Riemannian manifold $M$ with metric $g$ (Frim et al., 2020).

The geometric formulation is expressed in differential forms. Defining the $1$-form

$$P=\rho_{\mathrm{eq}}\,\mathbf{F}_{\mathrm{ext}}\cdot d\mathbf{x},$$

the core constraint becomes

$V(x;\lambda(t))$0

Using the Hodge decomposition,

$V(x;\lambda(t))$1

with $V(x;\lambda(t))$2 a $V(x;\lambda(t))$3-form, $V(x;\lambda(t))$4 a $V(x;\lambda(t))$5-form, and $V(x;\lambda(t))$6 a harmonic $V(x;\lambda(t))$7-form, the ESE condition reduces to the generalized Poisson equation

$V(x;\lambda(t))$8

In the simplest gauge $V(x;\lambda(t))$9, the auxiliary force is

$$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$0

The significance of this construction is twofold. First, it establishes solvability on arbitrary compact manifolds because normalization implies

$$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$1

which is the Poisson solvability condition. Second, it shows that full temporal configurational control is, in principle, achievable along any smooth equilibrium trajectory generated by a time-dependent potential on compact $$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$2 (Frim et al., 2020).

The paper works out explicit examples on $$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$3 and $$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$4. On the sphere, the system is an electric dipole in a time-dependent uniform field, with the solution constructed by spherical-harmonic expansion. On the torus, the system is two coupled pendula with time-dependent coupling, solved via Fourier modes. In both cases, simulations show that without ESE the distribution lags behind instantaneous equilibrium, whereas with the auxiliary force it tracks the instantaneous Boltzmann distribution closely during sigmoidal or periodic driving (Frim et al., 2020). The paper does not derive explicit general-manifold work or entropy-production formulae, but places the method in the context of prior finite-time thermodynamics and conjectures dissipation scaling of the form $$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$5 for short protocols.

3. EqualMotion as moving equilibrium families in nonautonomous dynamics

"Equilibria in Motion: Stability, Tracking, and Convergence" (Saoud, 25 Jun 2026) studies nonautonomous systems

$$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$6

with time-varying equilibrium sets

$$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$7

The central object is the tracking error

$$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$8

Unlike classical Lyapunov theory for fixed equilibria, the target set here moves and may be nonisolated, for example a manifold or affine hyperplane (Saoud, 25 Jun 2026).

The paper quantifies set motion through an equilibrium speed $$\rho_{\mathrm{eq}}(x;\lambda(t))=\frac{\exp[-\beta V(x;\lambda(t))]}{Z(\lambda(t))}.$$9 defined by local Hausdorff estimates. For every radius $\mathbf{F}_{\mathrm{ext}}$0,

$\mathbf{F}_{\mathrm{ext}}$1

This local formulation accommodates unbounded equilibrium families. If $\mathbf{F}_{\mathrm{ext}}$2, the total equilibrium travel is finite and the family admits a limiting geometry in the local Hausdorff sense (Saoud, 25 Jun 2026).

The Lyapunov framework uses two functions $\mathbf{F}_{\mathrm{ext}}$3, two nonnegative observables $\mathbf{F}_{\mathrm{ext}}$4, and coupled dissipation inequalities,

$\mathbf{F}_{\mathrm{ext}}$5

With the composite Lyapunov function

$\mathbf{F}_{\mathrm{ext}}$6

Theorem 3.1 yields a strict-decay estimate

$\mathbf{F}_{\mathrm{ext}}$7

where $\mathbf{F}_{\mathrm{ext}}$8. On a tube around the moving equilibrium family, the theory assumes an energy-distance comparison

$\mathbf{F}_{\mathrm{ext}}$9

This allows Lyapunov dissipation to be converted into geometric tracking bounds (Saoud, 25 Jun 2026).

The main quantitative estimate is

$\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$0

The first term is exponentially decaying memory of the initial error; the second is an exponentially weighted convolution of equilibrium motion and residual forcing. If $\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$1, then $\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$2. The paper also proves an ISS-type estimate for perturbed dynamics $\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$3, where the distance to the moving equilibrium family is bounded by an exponentially decaying transient plus contributions from $\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$4, $\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$5, and $\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$6 (Saoud, 25 Jun 2026).

The application to dynamic resource allocation makes the geometric meaning explicit. For demand $\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$7, the equilibrium set is

$\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$8

an affine hyperplane whose speed is controlled by

$\rho(x,t)=\rho_{\mathrm{eq}}(x;\lambda(t))$9

If $t$0, then $t$1 converges to $t$2, the equilibrium family converges to

$t$3

and tracking relative to $t$4 transfers to convergence relative to $t$5 (Saoud, 25 Jun 2026).

4. EqualMotion in energy-equidistributed PINNs

In "Energy-Equidistributed Moving Sampling Physics-informed Neural Networks for Solving Conservative Partial Differential Equations" (Gao et al., 27 Aug 2025), EqualMotion is interpreted as the movement of mesh or collocation points so that energy is equidistributed across them. The setting is a conservative PDE of Hamiltonian type,

$t$6

with energy functional

$t$7

where $t$8 is the energy density. In the conservative case, $t$9 (Gao et al., 27 Aug 2025).

In one dimension, the discrete energy-equidistribution principle requires equal cell energies,

$M$0

The continuous form is

$M$1

which yields the equidistribution equation

$M$2

Differentiating gives the Euler–Lagrange form

$M$3

In multi-D, the construction is generalized through a variational functional and an SPD metric $M$4, leading to the 2D energy-equidistribution PDEs (Gao et al., 27 Aug 2025).

Rather than solving the elliptic system at every time, the paper introduces two EMMPDEs. The location-based form is

$M$5

and the velocity-based form is

$M$6

These moving-mesh laws are integrated with PINNs to form EEMS-PINNs. The architecture uses two networks: a PDE solution network $M$7 and a mesh-mapping network $M$8. The mesh network is trained by an EMMPDE loss, and the PDE network is then retrained on the moved physical collocation set $M$9 (Gao et al., 27 Aug 2025).

The monitor is the energy density itself. For the nonlinear wave example,

$g$0

This makes the adaptation energy-aware rather than residual-driven or gradient-driven. Reported numerical results are specific: for the 1D cubic Klein–Gordon equation, EEMS-PINN achieves relative $g$1 errors $g$2 for $g$3 points, improving over PINN ($g$4) and WAM-PINN ($g$5), and maintains relative energy error $g$6 across $g$7 (Gao et al., 27 Aug 2025). For the forced Klein–Gordon equation it reports $g$8 at $g$9. For the 1D sine–Gordon equation it reports $1$0 at $1$1. For the 2D sine–Gordon equation it reports $1$2 errors as low as $1$3 at $1$4 (Gao et al., 27 Aug 2025).

The paper also states several limitations. Two networks must be trained; extra losses and AD evaluations increase computational cost; and more rigorous analysis of stability and convergence of EMMPDE-driven PINNs is left to future work. It further notes that only one or two EqualMotion cycles are often needed in experiments, which partially limits the added overhead (Gao et al., 27 Aug 2025).

5. EqualMotion as accessible body-agnostic motion capture

The paper "EqualMotion: Accessible Motion Capture for the Creative Industries" (Hilton et al., 11 Jul 2025) defines EqualMotion as a body-agnostic, wearable motion-capture system designed specifically to be accessible to disabled creative practitioners. Technically, it is a modular IMU-based mocap suit using x-IMU3 sensors and a software pipeline; conceptually, it is a rethinking of mocap around disability-centred design rather than normative body models (Hilton et al., 11 Jul 2025).

The system is motivated by structural limitations in conventional mocap. Commercial platforms such as Vicon, Qualisys, Xsens, and Rokoko are described as assuming a fixed, ideal human skeleton with fixed joint topology, assumed limb presence, and symmetric full-range joint mobility. Standard IMU-based and optical calibration procedures frequently require T-pose, A-pose, or similar canonical poses, which may be impossible, unsafe, painful, or incompatible with seated or supine users. Harvey et al. are cited as showing that measurement and validation practices encode these assumptions at a deep level, while Gerling and Spiel characterize VR more broadly as being organized around a “corporeal standard” (Hilton et al., 11 Jul 2025).

EqualMotion responds through disability-centred co-design guided by the social model of disability and Charlton’s principle “nothing about us without us.” The team includes disabled and non-disabled researchers from HCI, critical disability studies, and engineering, and conducts co-design workshops with disabled movement practitioners, musicians, creative technologists, and mocap engineers. The design principles synthesized from this work are body-agnosticism, personalization, integration of mobility aids, respect for diverse movement vocabularies, and agency and comfort in calibration and representation (Hilton et al., 11 Jul 2025).

The hardware is an array of approximately 18 x-IMU3 modules in a kit with power banks and a network router. The wearable configuration is modular and node-based. A root IMU provides the central reference, for example on the torso or wheelchair frame, and additional IMUs are attached to user-relevant locations such as a residual limb, prosthetic, shoulder, wheel rim, or chair frame. Nodes are labeled using self-identified language rather than fixed anatomical categories. The software pipeline acquires

$1$5

estimates IMU orientations $1$6 by sensor fusion, and maps them not to a fixed skeleton but to user-defined segments with personalized relative transforms (Hilton et al., 11 Jul 2025).

Calibration departs sharply from standard skeletal calibration. Instead of enforcing a universal neutral pose, EqualMotion builds a user-specific motion profile from a root node and user-chosen accessible poses or movement arcs. The planned workflow includes root selection, node labeling, user-chosen calibration poses or movements, and estimation of relative transforms. Conceptually, for a segment $1$7,

$1$8

and a least-squares problem over calibration samples is used to estimate $1$9 (Hilton et al., 11 Jul 2025). The system is also designed to ship with a dataset of accessible calibration poses and profiles for different mobility configurations, such as wheelchair use or unilateral amputation.

Mobility aids are treated as first-class tracked entities rather than artifacts or noise. IMUs may be mounted on wheelchair frames and wheels, crutches, canes, walkers, and prosthetic limbs. For wheels, the paper anticipates explicit rotational degrees of freedom and notes the need to manage magnetic disturbance from ferrous wheelchair materials by carefully handling magnetometer data (Hilton et al., 11 Jul 2025). This design choice extends to visual representation. Instead of default humanoid avatars, EqualMotion proposes an inclusive abstract-to-specific visual language: pre-calibration visualization begins as a constellation of indistinguishable nodes; during calibration, relevant nodes become associated with segments and a minimal line-and-node figure emerges; after calibration, the visualization may remain abstract or be enriched according to user preference (Hilton et al., 11 Jul 2025).

Evaluation is ongoing rather than complete. The paper identifies two longitudinal case studies, one in dance and one in music, to investigate rehearsal and performance use, software bridges to existing creative tools, and effects on participation, agency, and creative empowerment. It explicitly states that concrete metrics and quantitative comparative performance data are not yet reported (Hilton et al., 11 Jul 2025).

6. Distinctions, limitations, and recurrent themes

The most important distinction is categorical. In (Hilton et al., 11 Jul 2025), EqualMotion is a named mocap system. In (Frim et al., 2020, Saoud, 25 Jun 2026), and (Gao et al., 27 Aug 2025), the label refers to an interpretive description of distinct mathematical or computational programs: rapid equilibration of stochastic systems, tracking of moving equilibrium families, and energy-equidistributed mesh motion, respectively. Treating them as a single technical framework would therefore be misleading.

At the same time, several recurrent themes are explicit across the corpus. Each formulation introduces motion relative to a changing target rather than a static reference: instantaneous Boltzmann equilibrium in ESE (Frim et al., 2020), a moving equilibrium family in nonautonomous ODEs (Saoud, 25 Jun 2026), evolving energy density in moving-sampling PINNs (Gao et al., 27 Aug 2025), and personalized bodily configuration plus mobility aids in accessible mocap (Hilton et al., 11 Jul 2025). Each also replaces a normative default with an adaptive construct: Euclidean geometry is generalized to compact Riemannian manifolds; fixed equilibria are generalized to nonisolated moving sets; fixed collocation meshes are replaced by EMMPDE-driven redistribution; fixed humanoid skeletons are replaced by body-agnostic node graphs. This suggests a shared methodological tendency toward target-relative adaptation, although the underlying mathematics, implementation substrates, and application domains remain different.

The limitations are likewise domain-specific. In ESE, very short protocols require stronger forces and may challenge the validity of the overdamped Langevin description (Frim et al., 2020). In the equilibria-in-motion framework, the theory depends on local Hausdorff continuity of equilibrium sets, suitable Lyapunov functions, and energy-distance comparison on a tube (Saoud, 25 Jun 2026). In EEMS-PINNs, training cost increases because two networks and additional AD computations are required, and rigorous convergence theory is incomplete (Gao et al., 27 Aug 2025). In accessible mocap, technical maturity, quantitative validation, magnetometer interference, and broader coverage of disability configurations remain open issues (Hilton et al., 11 Jul 2025).

Taken together, EqualMotion is best understood not as a singular doctrine but as a term that presently indexes multiple research programs concerned with controlled motion under changing constraints. In one case it is an inclusive wearable system for creative practice; in the others it is a mathematically structured strategy for tracking equilibrium, energy, or distributional targets under nonstationary conditions.