Continuous RMPS Models Overview

• Continuous RMPS models are continuous frameworks that extend discrete methods to unify quantum, plasma, and robotic applications with rigorous mathematical structures.
• In quantum gravity, cRMPS extend MPS techniques using path-ordered exponentials and spectral gap analysis to capture all-order gravitational effects and wormhole phase transitions.
• In fusion plasmas and robotics, cRMPS enable robust ELM suppression and stable motion control by fusing overlapping magnetic island dynamics and continuous Riemannian policies.

Continuous RMPS (cRMPS) models encompass a diverse set of technical paradigms unified by the theme of utilizing continuous—rather than discrete—mathematical structures for modeling, control, or physical description. Despite the identical acronym, the term "cRMPS" is independently used in stringently different domains: (1) random matrix product state models in quantum gravity and quantum statistical physics (Jung et al., 12 Dec 2025), (2) magnetically-controlled edge-localized mode suppression in fusion plasma physics (Hu et al., 2019), and (3) continuous Riemannian motion policy synthesis in robotics and control (Ratliff et al., 2018). This article provides comprehensive technical coverage of each usage context, clarifying key definitions, underlying mathematics, and area-specific significance.

1. cRMPS in Quantum Gravity: Random Matrix Product State Models

Random matrix product state (RMPS) models in the continuum generalize discrete matrix product state (MPS) techniques to describe quantum field theoretic states prepared by gravitational path integrals with fixed gravitational boundary conditions but unconstrained matter sectors. These "gravitationally prepared states" encode all-order quantum gravitational effects in closed universes.

Discrete to Continuum Construction

• Discrete MPS: For a 1D chain of $L$ sites, each with local $\mathbb{C}^k$ Hilbert space, the MPS reads $$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$, where $A_i$ are $N\times N$ matrices.
• Continuum limit (cMPS): By dividing the circle of length $L$ into $L/\epsilon$ sites, setting $$A_i \to I_N + \epsilon Q \otimes I + \sqrt{\epsilon} R_i \otimes \psi_i^\dagger(x)$$ and letting $\epsilon \to 0$, this yields:

$$|\Psi\rangle = \mathrm{Tr}\, \mathcal{P} \exp \int_0^L dx \left\{ Q\otimes I + \sum_{i=1}^k R_i \otimes \psi_i^\dagger(x) \right\} |0\rangle$$

with $\mathbb{C}^k$0 bosonic fields and $\mathbb{C}^k$1 denoting $\mathbb{C}^k$2-ordering (Jung et al., 12 Dec 2025).

Transfer Hamiltonian and Inner Product

The cRMPS state with $\mathbb{C}^k$3 Hermitian and $\mathbb{C}^k$4 defines the inner product as:

$\mathbb{C}^k$5

A bra–ket disconnected average (corresponding to the absence of wormhole connections) uses independent $\mathbb{C}^k$6 and $\mathbb{C}^k$7 and $\mathbb{C}^k$8.

Spectral Gap and Bra–Ket Wormhole Phase Transition

The spectral gapping property, crucial for gravitational phase transitions, is defined as $\mathbb{C}^k$9. Here, $$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$0 exhibits a gapped ground state while $$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$1 is gapless, yielding a sharp "wormhole" phase transition in the large volume ($$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$2) limit. Correlators in the connected phase show exponential decay, while disconnected correlators decay sub-exponentially.

O(k) Models and Large-k Limit

For $$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$3 drawn from an O$$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$4-invariant measure $$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$5, the large-$$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$6 limit ensures that both $$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$7 and $$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$8 self-average to simple forms, establishing universal spectral properties and guaranteeing the wormhole phase transition (Jung et al., 12 Dec 2025).

Off-shell Wormholes and Long-Range Offsets

A key prediction is the existence of nonzero long-distance correlator offsets, $$|\Psi\rangle = \sum_{i_0,\ldots,i_{L-1}=1}^k \mathrm{Tr}[A_{i_0}A_{i_1}\ldots A_{i_{L-1}}]|i_0\ldots i_{L-1}\rangle$$9, originating from off-shell wormhole effects—topologies lacking semiclassical saddle-points but contributing via random matrix fluctuations.

2. cRMPS in Fusion Plasmas: Continuous RMP Suppression

In fusion plasma physics, cRMPS refers to "continuous resonant magnetic perturbation suppression"—the emergence of wide, robust operational windows for edge-localized mode (ELM) suppression due to the overlap of resonant magnetic islands at the plasma edge (Hu et al., 2019).

Two-Fluid MHD Model and Penetration Thresholds

The TM1 code solves nonlinear two-fluid, resistive MHD equations:

• Ion and electron continuity, momentum, generalized Ohm’s law with Hall terms, Faraday’s law, and pressure transport.
• Penetration of a resonant $A_i$0 harmonic at rational $A_i$1 requires:

$A_i$2

Since $A_i$3, lowering line-averaged density $A_i$4 or increasing toroidal mode number lowers $A_i$5 and facilitates penetration over a broader $A_i$6 window.

Island Overlap and the Chirikov Parameter

When multiple islands (e.g., $A_i$7 and $A_i$8) satisfy the penetration criterion, their widths $A_i$9 may overlap if $N\times N$0, where $N\times N$1 is the spacing between the surfaces. This overlap yields a quasi-continuous chain of magnetic perturbations at the pedestal top, enabling extended ELM suppression windows (up to $N\times N$2 for sufficiently low $N\times N$3) (Hu et al., 2019).

Mode Number Scalings and Reactor Relevance

For $N\times N$4, the closer packing of $N\times N$5 surfaces means that even at high density, moderate coil fields ($N\times N$6 Gauss) drive island overlaps, yielding continuous $N\times N$7 pedestal pressure clamping over reactor-relevant $N\times N$8 and $N\times N$9.

3. cRMPS in Robotics: Continuous Riemannian Motion Policies

In robotics and control, continuous Riemannian Motion Policies (cRMPs) are defined as tuples $L$0, where $L$1 is a second-order policy generating continuous-time hyper-acceleration laws $L$2, and $L$3 is a Riemannian metric encoding directional importance or inertia (Ratliff et al., 2018).

Core Operators: Pullback, Pushforward, and Fusion

• Pullback: For a smooth map $L$4 with Jacobian $L$5, the pullback of an RMP $L$6 from $L$7 to $L$8 yields $L$9, $L/\epsilon$0, where $L/\epsilon$1 denotes the pseudoinverse.
• Pushforward: For $L/\epsilon$2 in $L/\epsilon$3, pushforward yields $L/\epsilon$4 in $L/\epsilon$5.
• Addition: Two RMPs $L/\epsilon$6 fuse as $L/\epsilon$7 with $L/\epsilon$8.

Associativity and optimality derive from minimizing joint squared-error costs; these operators guarantee geometric and information-theoretic consistency.

Applications and Unification

• Obstacle and joint-limit avoidance: Task-specific RMPs with adaptive metrics are pulled back and combined.
• Dynamic primitives and operational-space control: Standard approaches are recovered as special cases of unified cRMP addition in suitable manifolds.
• MPC blending: Time-varying or task-local quadratic policies derived from MPC may be encoded and stably fused with reactive sub-policies.

4. Mathematical Structures and Operator Properties

cRMPS mathematical structure is determined by context:

• In quantum gravity, the object is a path-ordered exponential over a field of operator-valued functions, and the state normalization and correlators reduce to traces over exponential transfer Hamiltonians.
• In plasma physics, cRMPS hinges on the spatial overlap of dynamically-evolving magnetic island widths governed by coupled MHD and transport equations.
• In robotics, cRMPS tuples reside in function spaces over the tangent bundle and are manipulated via matrix (co)tangent transformations consistent with the geometry of task/configuration spaces.

Core operator properties—symmetry, positive-definiteness, associativity, order-agnostic fusion—are strictly preserved in all implementations, with rigorous closed-form expressions for composition and transformation.

5. Scaling Laws, Regimes, and Physical Implications

Each cRMPS context features characteristic scaling regimes:

• Quantum gravity: Off-shell correlator offsets and higher-point contributions scale as $L/\epsilon$9, and presence/absence of phase transitions is set by spectral gap inequalities.
• Fusion plasmas: Island width overlaps and suppression window widths scale with $$A_i \to I_N + \epsilon Q \otimes I + \sqrt{\epsilon} R_i \otimes \psi_i^\dagger(x)$$0, $$A_i \to I_N + \epsilon Q \otimes I + \sqrt{\epsilon} R_i \otimes \psi_i^\dagger(x)$$1, $$A_i \to I_N + \epsilon Q \otimes I + \sqrt{\epsilon} R_i \otimes \psi_i^\dagger(x)$$2, and field amplitude, with the Chirikov parameter defining the transition into continuous suppression regimes.
• Robotics: The magnitude and conditioning of the summed metric $$A_i \to I_N + \epsilon Q \otimes I + \sqrt{\epsilon} R_i \otimes \psi_i^\dagger(x)$$3 govern dynamic responsiveness and numerical stability, and successful blending depends on positive-definite fusion of component policies.

Physical implications include nonperturbative gravitational phase completion, robust control of instabilities in next-generation fusion devices, and stable, geometrically consistent multi-objective robot behavior synthesis.

6. Limitations, Validation, and Open Directions

• Quantum gravity: While cRMPS provides a robust framework for studying all-orders topology and off-shell effects, extension to real-time de Sitter geometries requires handling non-Hermitian $$A_i \to I_N + \epsilon Q \otimes I + \sqrt{\epsilon} R_i \otimes \psi_i^\dagger(x)$$4, a matter deferred to future studies (Jung et al., 12 Dec 2025).
• Fusion plasmas: The TM1-based cRMPS approach relies on cylindrical geometry, simplified viscosity models, and neglects sheath/SOL and full kinetic corrections, but quantitatively matches experimentally observed suppression windows and pedestal clamping (Hu et al., 2019).
• Robotics: The cRMP machinery guarantees optimality only under sufficient smoothness, rank, and regularization; practical scenarios necessitate active damping, metric learning, or local projections for constraint handling.

A plausible implication across all cases is that the continuum generalization of discrete mathematical tools—be they random tensor networks, magnetic island ladder logic, or quadratic policy blending—enables tractable, rigorous analysis and synthesis in regimes where discrete frameworks become unwieldy.

---

Key References:

• Quantum gravity and random matrix product state cRMPS: "Random matrix product state models of gravitationally prepared states" (Jung et al., 12 Dec 2025)
• ELM suppression and continuous RMPS in fusion plasmas: "Wide operational windows of edge-localized mode suppression by resonant magnetic perturbations in the DIII-D tokamak" (Hu et al., 2019)
• Continuous Riemannian motion policies in robotics: "Riemannian Motion Policies" (Ratliff et al., 2018)