Benamou-Brenier and Kantorovich on sub-Riemannian manifolds with no abnormal geodesics

Basic Information

• Paper ID: 2507.20959
• Title: Benamou-Brenier and Kantorovich on sub-Riemannian manifolds with no abnormal geodesics
• Authors: Giovanna Citti, Mattia Galeotti, Andrea Pinamonti
• Classification: math.OC (Optimization and Control)
• Publication Date: October 15, 2025 (arXiv v2)
• Paper Link: https://arxiv.org/abs/2507.20959

Abstract

This paper establishes the equivalence between the Benamou-Brenier dynamic formulation and the Kantorovich static formulation of the optimal transport problem on connected, complete, boundaryless sub-Riemannian manifolds $M$ when the manifold admits no nontrivial abnormal geodesics, where the transport problem is considered between two measures with finite second moments. Furthermore, the paper proves the existence of minimizers for the Benamou-Brenier formulation and connects them to optimal transport plans.

Research Background and Motivation

Problem Background

1. Development of Optimal Transport Theory: The optimal transport problem was originally posed by Monge in the late 18th century, later relaxed by Kantorovich into a more analytically tractable version. The equivalence between the Kantorovich formulation and the Benamou-Brenier dynamic formulation has been established on Euclidean domains and Riemannian manifolds.
2. Challenges in Sub-Riemannian Geometry: Sub-Riemannian manifolds, characterized by a completely nonintegrable distribution $HM \subset TM$ and a positive definite quadratic form $g_{SR}$, possess more complex geometric structures than the Riemannian case. Notably, the sub-Riemannian framework admits two classes of geodesics: normal geodesics and abnormal geodesics.
3. Impact of Abnormal Geodesics: The existence of abnormal geodesics causes the sub-Riemannian distance to lose local convexity, which is the primary obstacle to the existence of optimal maps.

Research Motivation

• Extend classical optimal transport theory to the sub-Riemannian geometric framework
• Resolve the open problem of equivalence between Benamou-Brenier and Kantorovich formulations on sub-Riemannian manifolds
• Provide theoretical foundations for optimal transport in sub-Riemannian geometry

Core Contributions

1. Equivalence Theorem: Establishes that on complete, connected sub-Riemannian manifolds without abnormal geodesics, the infima of the Kantorovich, Benamou-Brenier, and relaxed Benamou-Brenier formulations are finite, attainable, and equal.
2. Relaxed Theoretical Framework: Constructs a relaxed Benamian-Brenier problem based on Young measures, replacing the traditional vector field integral with an integral of probability measures on $[0,1] \times HM$.
3. Measurable Selection Theorem: Constructs a measurable mapping $S: M \times M \to \text{Geod}(M)$ that selects a geodesic connecting each pair of points, representing a key technical breakthrough.
4. Superposition Principle: Establishes the Young superposition principle in the sub-Riemannian framework, proving that any transport measure can be decomposed into probability measures on generalized curves.
5. Optimality Characterization: Proves that the support of Benamou-Brenier minimizers decomposes over (generalized constant-speed) geodesics.

Methodology Details

Problem Formulation

Consider the optimal transport problem between two probability measures $\mu_0, \mu_1 \in P_2(M)$ with finite second moments on a sub-Riemannian manifold $(M, H, g_{SR})$.

Kantorovich Formulation: $J_{Kan}(\gamma) = \int_{M \times M} d^2(x,y) d\gamma(x,y)$ where $\gamma \in \Pi(\mu_0, \mu_1)$ is a feasible transport plan.

Benamou-Brenier Formulation: $J_{BB}(\mu_t, v_t) = \int_0^1 \int_M \|v_t(x)\|^2 d\mu_t(x) dt$ subject to the continuity equation: $\dot{\mu}_t + \text{Div}(\mu_t v_t) = 0$.

Relaxation Framework

Young Measure Theory

Define the Young measure space: $Y(I; X) = \{\eta \in P(I \times X) : \pi^{(t)}_\# \eta = L\}$ where $L$ is the Lebesgue measure.

Relaxed Benamou-Brenier Problem

$J^*_{BB}(\eta) = \int_{[0,1] \times HM} \|v\|^2 d\eta(t,v)$ subject to the relaxed continuity equation: $\int_{I \times HM} \partial_t \phi(t, \pi(v)) + \langle v, \nabla_H \phi(t, \pi(v)) \rangle d\eta(t,v) = 0$

Technical Innovations

1. Construction of Measurable Selection Mapping

Constructs the mapping $S: M \times M \to \text{Geod}(M)$ using Suslin set theory, with key steps including:

• Proving that the extremal trajectory set $E$ is a Suslin set
• Applying the measurable selection theorem to establish the existence of $S$
• Handling technical difficulties under the assumption of absent abnormal geodesics

2. Establishment of Relaxation Theory

Establishes via Jensen's inequality: $J_{BB}(\mu^{\eta}_t, v^{\eta}_t) \leq J^*_{BB}(\eta)$ where $(\mu^{\eta}_t, v^{\eta}_t)$ is the classical pair derived from the Young measure $\eta$.

3. Application of Superposition Principle

For any transport measure $\eta$, there exists a decomposition $\tilde{\eta} \in P(G_H(I,M))$ such that: $\eta = \int_G \nu d\tilde{\eta}(\nu)$

Experimental Setup

Theoretical Verification Framework

This is primarily a theoretical work verified through rigorous mathematical proofs rather than numerical experiments. The verification framework includes:

1. Verification of Assumptions:
   • Completeness and connectedness of the manifold
   • Boundaryless condition
   • Absence of nontrivial abnormal geodesics
2. Constructive Proofs: Establishes existence of various mappings and decompositions through explicit construction

Technical Tools

• Suslin set theory and measurable selection theorem
• Young measure theory and calculus of variations
• Hamiltonian theory in sub-Riemannian geometry
• Measure theory and functional analysis

Main Results

Equivalence Theorem (Theorem 1.1)

Main Conclusion: Let $M$ be a connected, complete, boundaryless sub-Riemannian manifold without nontrivial abnormal geodesics, and let $\mu_0, \mu_1 \in P_2(M)$. Then: $C_{Kan}(\mu_0, \mu_1) = C_{BB}(\mu_0, \mu_1) = C^*_{BB}(\mu_0, \mu_1)$ and all infima are attainable.

Relaxed Equivalence (Theorem 3.6)

Proves the equivalence between the classical Benamou-Brenier formulation and its relaxed version: $C_{BB}(\mu_0, \mu_1) = C^*_{BB}(\mu_0, \mu_1)$

Optimality Characterization (Theorem 3.11)

For Benamou-Brenier minimizers $(μ_t, v_t)$, their Young decomposition $\tilde{\eta}$ is supported on geodesics.

Technical Lemmas

• Lemma 3.7: The mapping $\gamma \mapsto F_\#(L \otimes \gamma)$ maps Kantorovich plans to feasible Young transport measures
• Lemma 3.8: $J_{Kan}(\gamma) = J^*_{BB}(F_\#(L \otimes \gamma))$

Related Work

Classical Optimal Transport

• Establishment of Kantorovich relaxation theory
• Brenier's breakthrough work in the Euclidean case
• McCann's generalization to Riemannian manifolds

Sub-Riemannian Optimal Transport

• Existence and uniqueness results on Heisenberg groups
• Special cases of 2-generated distributions
• Impact of abnormal geodesics on existence of optimal maps

Young Measure Theory

• Applications of control relaxation techniques
• Foundational work by Bernard et al. on Young measures
• Development of superposition principles

Conclusions and Discussion

Main Conclusions

1. Under appropriate assumptions, optimal transport problems on sub-Riemannian manifolds possess favorable properties similar to the Riemannian case
2. The relaxation method provides effective tools for handling the complexity of sub-Riemannian geometry
3. The absence of abnormal geodesics is a key condition for ensuring equivalence

Limitations

1. Abnormal Geodesics Restriction: The method cannot be directly extended to manifolds with nontrivial abnormal geodesics
2. Technical Assumptions: Requires completeness, connectedness, and boundarylessness of the manifold
3. Constructive Dependence: Key results depend on the existence of measurable selection mappings

Future Directions

1. Study optimal transport in the presence of abnormal geodesics
2. Develop more general relaxation theory frameworks
3. Explore numerical algorithms and computational methods
4. Apply to specific sub-Riemannian manifolds (e.g., Carnot groups)

In-Depth Evaluation

Strengths

1. Theoretical Innovation

• First to establish complete equivalence between Kantorovich and Benamou-Brenier formulations in the sub-Riemannian geometric framework
• Innovatively employs Young measure theory to address technical difficulties in sub-Riemannian geometry
• Cleverly utilizes Suslin set theory to resolve measurable selection problems

2. Technical Rigor

• Proofs are complete and rigorous with careful handling of technical details
• Clear understanding of the necessity of the abnormal geodesics absence assumption
• Technical results in appendices provide solid support for main theorems

3. Theoretical Completeness

• Not only proves equivalence but also establishes existence of minimizers
• Provides explicit connections between optimal transport plans and dynamic formulations
• The theoretical framework exhibits good internal consistency

Weaknesses

1. Limited Applicability

• The assumption of absent abnormal geodesics is quite restrictive, excluding many important sub-Riemannian manifolds
• Generalization of the method to more general settings faces fundamental technical obstacles

2. Practical Considerations

• As a purely theoretical work, lacks numerical verification and application examples
• Limited guidance for concrete computation and algorithm implementation

3. Technical Complexity

• Proofs involve multiple advanced mathematical branches with high barriers to understanding
• Key measurable selection results rely on abstract set-theoretic techniques

Impact Assessment

1. Academic Value

• Establishes important theoretical foundations for optimal transport in sub-Riemannian geometry
• Advances development in the intersection of geometric analysis and optimal transport
• Provides important theoretical tools for subsequent research

2. Methodological Contribution

• Application of Young measure methods in sub-Riemannian geometry serves as a model
• Use of relaxation techniques offers new perspectives for handling non-smooth geometry

3. Potential Applications

• Provides theoretical support for optimal transport on manifolds in machine learning
• Has promising applications in geometric methods for image processing and computer vision

Applicable Scenarios

1. Theoretical Research: Geometric analysis, optimal transport theory, sub-Riemannian geometry
2. Applied Mathematics: Partial differential equations, calculus of variations, measure theory
3. Computational Geometry: Manifold learning, theoretical foundations for geometric deep learning

References

The paper cites important literature in the field, including:

• Ambrosio-Gigli's classical tutorial on optimal transport
• Bernard's foundational work on Young measures and superposition principles
• Figalli-Rifford's pioneering research on optimal transport on sub-Riemannian manifolds
• Agrachev et al.'s comprehensive work on sub-Riemannian geometry

Overall Assessment: This is a high-quality theoretical mathematics paper that makes important contributions at the intersection of sub-Riemannian geometry and optimal transport. While the technical assumptions are strong, it provides complete and profound theoretical results within its scope of applicability. It holds significant value for advancing related theoretical developments.