Carriage: Contexts and Applications

• Carriage is a context-specific term that denotes a carrier relation, describing physical objects or abstract states that transport loads, signals, or microbial agents across various disciplines.
• In mechanics and control, carriage represents moving entities such as railway cars, belt-driven guides, and rotating superstructures, where dynamics and control strategies are rigorously modeled.
• In computational systems, carriage underpins frameworks like CARRIAGE for recipe retrieval and metrics like carriage value, illustrating its pivotal role in data-driven and information-theoretic applications.

Searching arXiv for papers using the term "carriage" across domains. In recent arXiv literature, carriage is not a unitary technical object. It denotes, depending on domain, a railway car or other moving mechanical body, a rotating or translating subsystem in a machine, a transport objective such as cargo carriage, a biological state of microbial colonization, a consumer-network metric called carriage value, a device carriage state in IMU-assisted sensing, the acronym CARRIAGE = Cultural-Aware Recipe Retrieval Augmented Generation, and, in one recent information-geometric paper, a paper-specific horizontal information-carrying subspace rather than a classical geometric entity (Srzednicki, 2017, Spinelli et al., 2024, Stensen et al., 2022, Paul et al., 2023, Hu et al., 29 Jul 2025, Cheng et al., 3 Jul 2026). The term is therefore best understood through its local formalism rather than through any cross-domain dictionary definition.

1. Semantic range in current technical usage

The cited literature uses carriage in several distinct but structurally related ways.

Domain	Meaning of carriage	Representative source
Mechanics and control	Railway car, upper carriage, belt-driven carriage, two-wheeled carriage	(Srzednicki, 2017, Spinelli et al., 2024, Haren et al., 28 Feb 2025, Jóźwikowski et al., 2013)
Transport and routing	Cargo carriage or train-body passage	(Liebchen et al., 2019, Li et al., 2023)
Biology and epidemiology	Persistent carriage, asymptomatic carriage, colonization	(Stensen et al., 2022, Chatha et al., 2024)
Information systems	Carriage value, carriage state, CARRIAGE framework	(Paul et al., 2023, He et al., 2022, Hu et al., 29 Jul 2025)
Differential geometry in SMG	Horizontal carriage, historical horizontal carriage	(Cheng et al., 3 Jul 2026)

A plausible commonality is that the term consistently marks a carrier relation: a body that carries a load, a state that carries a microorganism, a metric that quantifies how much traffic an ISP carries per dollar, or a subspace that carries statistically meaningful variation. That commonality is interpretive rather than formal; the papers themselves define the term only within their own domain-specific constructions.

2. Mechanical and control-theoretic senses

In classical mechanics, carriage often retains its transport-vehicle meaning. In Whitney’s inverted pendulum problem, the carriage is a railway car whose prescribed motion is $f:\mathbb{R}\to\mathbb{R}$ in the line case or $f:\mathbb{R}\to\mathbb{R}^2$ in the planar case, and the pendulum state is written in coordinates attached to the carriage. The forcing enters through carriage acceleration rather than carriage position, and the paper proves existence of periodic non-falling motions when the carriage moves periodically: for straight-line motion with $f$ $T$-periodic and of $C^2$-class, and for planar motion with $f$ $T$-periodic and of $C^3$-class (Srzednicki, 2017).

A related railway sense appears in the study of railway carriage wheels equipped with disc brakes. There the carriage is the vehicle whose wheelset undergoes lock-up sliding, and the analysis focuses on wheel–rail frictional heating rather than tread-brake heating. Using a carriage of mass $50{,}000\ \mathrm{kg}$, $4$ axles, wheel nominal diameter $f:\mathbb{R}\to\mathbb{R}^2$0, lock-up sliding speed up to $f:\mathbb{R}\to\mathbb{R}^2$1, lock-up duration up to $f:\mathbb{R}\to\mathbb{R}^2$2, and adhesion coefficient $f:\mathbb{R}\to\mathbb{R}^2$3, the paper concludes that current WSP limits are not strict enough to eliminate martensite formation completely (Krénusz et al., 2021).

In heavy-machine control, upper carriage designates the rotating superstructure of a hydraulic material handler. The paper equates carriage with the rotating top structure carrying the cabin and arm, and uses near-synonyms including cabin joint, slew joint, upper carriage turn hydraulic motor, and turn hydraulic motor. This carriage is the hardest actuated joint to control because of delays, dead zones, nonlinearities, and a binary braking system; the proposed controller therefore learns a pressure-aware model of the upper carriage turn hydraulic motor and uses it inside an RL control loop (Spinelli et al., 2024).

In precision mechatronics, a belt-driven carriage is a 1 degree-of-freedom linear carriage on a guide actuated by a BLAC motor through a toothed pulley and timing belt. The paper treats its dominant nonuniformity as position-dependent dynamics and learns parameter-varying feedforward control; on this benchmark, the proposed LPV feedforward reduces RMS tracking error from $f:\mathbb{R}\to\mathbb{R}^2$4 to $f:\mathbb{R}\to\mathbb{R}^2$5, a $f:\mathbb{R}\to\mathbb{R}^2$6 improvement, and lowers the final-trial maximum absolute error by $f:\mathbb{R}\to\mathbb{R}^2$7 (Haren et al., 28 Feb 2025).

A more abstract mechanical usage appears in the two-wheeled carriage example in vakonomic versus nonholonomic dynamics. The configuration space is $f:\mathbb{R}\to\mathbb{R}^2$8, the carriage is subject to rolling constraints, and the paper derives the exact condition under which every nonholonomic trajectory is also vakonomic in the shifted-center case: $f:\mathbb{R}\to\mathbb{R}^2$9 This makes carriage a constrained rigid-body system on a Lie group rather than a transport vehicle in the ordinary engineering sense (Jóźwikowski et al., 2013).

Finally, in cooperative train control, a high-speed train is modeled as a sequence of multiple carriages. The paper indexes trains by $f$0 and carriages within train $f$1 by $f$2, distinguishes head carriages from other carriages, and designs a distributed observer-controller pair for every carriage. Intra-train regulation acts at the carriage level, while inter-train cooperative cruise is enforced through each train’s head carriage using information from the preceding train’s tail carriage (Zhang et al., 2022).

3. Carriage as transport objective and as transmission interface

In active-matter navigation, cargo carriage is an application domain rather than a subsystem name. The microswimmer paper formulates route planning as a generalized Fermat principle and studies trajectories minimizing travelling time, drag-related energy dissipation, or self-propulsion fuel use. Its explicit motivation is microswimmer applications such as cargo carriage and drug targeting to cancer cells, and one of its main lessons is that the optimal path need not be geometrically shortest: in vortex or shear fields, the swimmer may temporarily move away from the target or even down the adverse flow to arrive sooner (Liebchen et al., 2019).

In mmWave train communications, the train carriage is instead a propagation barrier. The high-speed-train paper states that mmWave signals encounter severe losses when passing through the carriage and therefore places an intelligent refracting surface on the train window to refract signals into the carriage. The “Without IRS” baseline is explicitly the case where onboard users can only receive signals going through the carriage, and the proposed IRS-assisted scheme improves over that baseline by about 23.7% at $f$3 users per frame (Li et al., 2023).

These two usages are technically different but conceptually adjacent. In one, carriage is the payload-delivery task; in the other, the carriage is the physical shell through which electromagnetic energy must pass. Both make carriage central to path or channel design rather than peripheral terminology.

4. Biological carriage: colonization, persistence, and hidden transmission

In microbiology and epidemiology, carriage denotes host colonization rather than transport. The youth-population study defines persistent nasal carriage of Staphylococcus aureus using two nasal swab cultures, with persistent carriers being those for whom both nasal cultures were positive. It finds that both persistent carriage and spa-type were transmitted in the social network, and reports that the probability of carriage increased by 3.7% and 5.0% for each additional S. aureus positive friend in univariable regression and multivariable autocorrelation analysis, respectively (Stensen et al., 2022).

The epidemic-inference paper generalizes this sense to healthcare-associated infections with high rates of asymptotic carriage. It explicitly notes that a “Colonized-Uncolonized model” would be more precise than “Susceptible Infected” for these pathogens. In its partially observed model, $f$4 is the latent colonization state, while $f$5 records observable infection through admission screening or later symptom onset. Carriage is therefore the hidden transmission state, not necessarily symptomatic disease, and this latent-carrier structure is the reason the likelihood becomes computationally intractable (Chatha et al., 2024).

A common misconception, addressed directly by both papers, is to equate carriage with overt infection. The literature here does not do so. In the S. aureus network study, carriage is a colonization phenotype measured by swabs; in the HAI model, the transmitting state is best interpreted as colonized/carrier, while observed infection is only an incomplete and delayed projection of that state.

5. Computational and information-system uses

One computational use is the acronym CARRIAGE, expanded as Cultural-Aware Recipe Retrieval Augmented Generation. This is a training-free, plug-and-play RAG framework for cross-cultural recipe adaptation whose four main mechanisms are Query Rewriting, Diversity-aware Re-ranking, Dynamic Context Organization, and Contrastive Context Injection. The paper’s key claim is that standard RAG underuses contextual diversity in creative generation, whereas CARRIAGE improves diversity while maintaining a diversity-quality tradeoff that the authors describe as Pareto efficient relative to the main closed-book baselines (Hu et al., 29 Jul 2025).

A second information-systems use is carriage value, defined as “the Mbps of a user's traffic that an ISP carries for one dollar,” effectively $f$6 when $f$7 is download speed and $f$8 is monthly price. At an address, the paper reports the best carriage value among available plans; at census block group level it uses the median across sampled addresses. Using this metric, the study finds that carriage value can vary as much as 600% within a city for DSL/fiber providers, and that cable-based ISPs offer up to 30% greater carriage value when competing with fiber-based ISPs in a block group (Paul et al., 2023).

A third computational use is device carriage state in BLE proximity detection. Here carriage state means how a receiver device is being carried—such as hand or pocket—and it is represented through IMU-derived features from gravity, linear acceleration, and angular velocity. The PRID framework treats carriage as both signal and confounder: IMU helps account for RSSI changes due to carriage, but bias arises if carriage state becomes spuriously correlated with the proximity label during data collection. PRID addresses this through carriage state regularization, a kernel-mean-matching reweighting scheme applied during classifier training (He et al., 2022).

These three usages share a technical pattern: carriage is not the primary prediction target but a structured intermediary. It names a diversity-oriented generation framework, a value metric aggregating speed and price, or a latent contextual factor that must be represented without being allowed to dominate the wrong objective.

6. Paper-specific mathematical terminology: horizontal carriage in SMG

A particularly nonstandard use occurs in “Statistically Meaningful Geometry (SMG) Beyond the Euclidean Paradigm, with Application to Generative AI” (Cheng et al., 3 Jul 2026). The paper states that “carriage” is not a standard differential-geometric term with an independent classical definition. Instead, it is a descriptive name for a horizontal information-carrying subspace/direction structure inside the fiber-bundle formalism

$f$9

Within that framework, the vertical distribution is

$T$0

the horizontal distribution is

$T$1

and the paper identifies

$T$2

The cleanest geometric characterization given in the paper is that carriage is the horizontal component that remains after the Ehresmann connection removes vertical content: $T$3

The phrase historical horizontal carriage is used in the sequential-adaptation setting. There it refers to the previously learned horizontal subspace, represented by $T$4 and its orthogonal projector $T$5. The downstream update is then filtered by

$T$6

which the paper interprets as projecting onto the orthogonal complement of the historical horizontal carriage. The claimed consequence is exact preservation of prior-task observable behavior, formalized by

$T$7

The terminological caution is essential. In classical differential geometry one would ordinarily speak of a horizontal distribution, horizontal lift, or parallel transport, not of a horizontal carriage. The paper uses the term metaphorically to denote the subspace that carries statistically meaningful learning motion and, in the continual-learning setting, the remembered horizontal knowledge subspace from prior tasks.

7. Conceptual summary

Across these papers, carriage is consistently tied to a relation of support, transport, containment, or conveyance, but the object being carried changes sharply by field. In mechanics it may be the body that carries a pendulum, cabin, arm, wheelset, or train passengers; in active matter it is the cargo-delivery task; in communications it is the train body that attenuates mmWave propagation; in microbiology it is the host’s colonization state; in information systems it is an affordability ratio or a user’s device-carrying configuration; and in SMG it is the horizontal subspace that carries statistically meaningful variation (Liebchen et al., 2019, Li et al., 2023, Stensen et al., 2022, Paul et al., 2023, He et al., 2022, Cheng et al., 3 Jul 2026).

The resulting encyclopedia point is not that these meanings collapse into a single formal definition. Rather, the current arXiv record shows carriage as a strongly context-bound technical term whose meaning is determined by the surrounding model: ODE forcing in the Whitney problem, wheel–rail thermomechanics in railway safety, hierarchical carriage indexing in cooperative train control, colonization and observation processes in epidemiology, structured retrieval and prompt history in recipe adaptation, or horizontal-versus-vertical decomposition in information geometry.