Intent-Addressable Physical Infrastructure (IAPI) is an architectural model in which human intent serves as the query, physical resources become resolvable endpoints, and infrastructure performs resource resolution and physical guidance or actuation.

Canonical definition

1Introduction

Digital systems hide location and implementation detail: a web user need not know the server that satisfies a request, and a GPS user names a destination rather than solving the route. Physical retrieval still requires the human to reconstruct the mapping between need and location while the environment, though it may hold rich information about the resource, stays operationally passive at the moment of need.

IAPI treats this pattern as an architectural choice rather than an unavoidable property of physical work. Where technically and operationally appropriate, infrastructure can perform the translation from resource intent to physical location and return a response that makes the resolved resource directly actionable.

Physical environments often require people to translate WHAT they need into WHERE it is. IAPI asks whether the environment itself can perform that resolution and return actionable GUIDANCE.

Architectural problem

The problem is not object identification, location tracking, natural-language understanding, or physical actuation considered separately. It is their closed-loop composition around a higher-level user abstraction: resource intent. This paper is architectural and conceptual. It proposes terminology, a reference model, conformance criteria, capability levels, and research questions. It is not a standards document, and it does not assert that every environment should become intent-addressable.

2Definition and Terminology

The IAPI use of intent follows a broader computing tradition in which users specify outcomes and systems determine how to realize them; RFC 9315 defines intent in networking as declarative goals without implementation detail [4]. Dey defined context as information characterizing an entity's situation and described context-aware systems that use it to provide relevant services [3]. IAPI claims neither as novel: context is one optional input to the specific mapping from resource intent to physical resource and address. 'I need a 10 mL syringe' is intent-addressable because it names the resource, not a coordinate. 'Illuminate rack 3, shelf 2, bin 7' is already an addressed command and bypasses the abstraction.

3Reference Interaction Model

WHAT → WHERE → GUIDANCE

Human-facing loop

WHAT is the resource intent, communicated as speech, text, selection, gesture, or workflow event; its defining property is that no physical address is supplied. WHERE is the infrastructure resolution problem: which resource satisfies the intent, and what is its current actionable address, using semantic interpretation, inventory state, role, recency, permissions, or environmental state. GUIDANCE is the physical closure of the loop: exact-location illumination, directional indication, display, audio, compartment activation, access control, or machine positioning. The person specifies the WHAT; the infrastructure determines the WHERE.

Intent → Resource Resolution → Physical Actuation

Machine-facing loop

The machine-facing model does not prescribe an algorithm. Resolution may be symbolic, statistical, neural, rule-based, probabilistic, hybrid, centralized, distributed, or deterministic, provided the complete system meets the application's requirements.

Formal abstraction and end-to-end reliability

Let I be an intent statement, C a bounded context state, R the candidate physical resources, A the machine-resolvable addresses, S system state, and M the allowable guidance or actuation modes. An IAPI implementation is three coupled mappings:

• F: (I, C) → R* performs semantic and contextual resource resolution; the star indicates F may yield a ranked or constrained candidate set before selection.
• G: (R*, S) → A maps the selected resource to a current physical address.
• H: (A, M) → P converts that address into a physical response P.

End-to-end correctness is conditional across stages. With E_F correct resolution, E_G correct and current addressing given E_F, and E_H correct actuation given both:

P(loop success) = P(E_F) × P(E_G | E_F) × P(E_H | E_F, E_G)

This is the chain rule of conditional probability; it does not assume independence. If each conditional stage succeeds 95% of the time, the complete loop succeeds at about 0.857. A system can report 95% intent accuracy, 95% address correctness, and 95% actuation fidelity and still fail roughly one interaction in seven. IAPI evaluation must therefore measure the complete WHAT → WHERE → GUIDANCE loop rather than treat any subsystem metric as a proxy for user-level success.

4Addressability as the Architectural Primitive

The term addressable is deliberate but analogical. RFC 3986 defines a URI as a compact sequence that identifies a resource and separates identification from the mechanism of access [1]. IAPI proposes no new URI scheme; it borrows the idea that resource identity can be separated from the mechanism used to locate or act on the resource.

Intent → Resource Identity → Physical Address → Physical Response

The system first resolves meaning to resource identity, then maps that identity to the resource's current address. A resource can therefore move without changing the user-level intent. Three domains are connected: the semantic domain (what does the user mean?), the resource domain (which physical resource satisfies that intent?), and the physical domain (where is it now, and how can the environment guide or act toward it?).

A physical address may be absolute, relative, hierarchical, topological, or state-dependent: site → unit → room → rack → shelf → bin in a hospital; line → station → fixture → slot in manufacturing; zone → aisle → bay → level → position in warehousing. IAPI requires no universal syntax, only that the selected resource resolves to an actionable endpoint with sufficient accuracy and freshness for the application.

5Minimum IAPI Conformance Criteria

To keep IAPI from becoming a label for any smart environment, this paper proposes minimum architectural criteria. They are conceptual, not a standard, and they let earlier domain-specific systems be analyzed as partial antecedents rather than excluded.

An open vocabulary is not required; a bounded one may be preferable in safety-critical or resource-constrained deployments. Voice, machine learning, computer vision, robotics, cloud connectivity, and adaptation are optional implementation choices. Some historical systems satisfy several of these steps, and in narrow workflows potentially all of them. That does not invalidate the category; it shows intent-addressable behavior existed in application-specific forms before the general abstraction was articulated. The distinction is generalization, explicit resource-address abstraction, and applicability beyond a single order, formulary, cabinet, or warehouse workflow.

6Historical Antecedents and Adjacent Technologies

Automated dispensing cabinets. A 1992 study of the Pyxis MedStation described an automated storage and distribution system that dispensed items to authorized users and recorded transactions [13]; modern cabinets add precise guidance, such as Omnicell's LED Guiding Lights that direct users to the drawer and bin [15]. Their design center is a bounded dispensing workflow tied to one appliance, its inventory, authorization, and formulary. IAPI abstracts the interaction away from the cabinet into a general relationship among intent, identity, mutable address, and response.

Pick-by-light. Shelf-mounted light modules indicate the location from which a picker should retrieve an item, and studies find they can reduce picking time or workload relative to paper [16]. The initiation direction differs: the order is generated by a warehouse workflow and the worker executes a pushed task rather than originating a request. The actuation primitive, turning a logical requirement into exact-location illumination, is plainly part of the IAPI lineage.

Voice-directed picking. Honeywell's Vocollect deployments prompt workers by voice and accept spoken confirmation [18], and field research describes voice-directed order selection as a computerized headset workflow [17]. Voice is not new; in conventional systems it is the delivery interface for a system-originated task sequence, whereas in IAPI it may be one means by which the human originates intent.

Ubiquitous and context-aware computing. Weiser's 1991 vision proposed computation woven into everyday environments [2], and Dey formalized context-aware applications [3]. IAPI narrows that lineage to one interaction: resolving a resource need to a physical endpoint and closing the loop.

7Relationship to Physical AI and Intent-Based Systems

Research increasingly treats the integration of AI and cyber-physical systems as a distinct systems problem involving uncertainty, resource constraints, safety, and physical coupling [8][9]; a 2026 IEEE TPAMI survey frames Physical AI as a progression from perception through reasoning and modeling to embodied interaction [10]. IDC's August 2026 analysis extends Physical AI beyond robots into buildings, vehicles, and infrastructure, describes a common loop of sense → analyze → act → adapt, argues that the intelligence resides in the loop rather than any device, and names hospital logistics among near-term value areas [11]. An IAPI implementation instantiates Physical AI as responsive infrastructure rather than a mobile or humanoid form factor.

Read against that loop, sense is receiving or inferring resource intent and environment state (WHAT); analyze is resolving intent to a resource and current address (WHERE); act is producing guidance or actuation at the resolved endpoint (GUIDANCE); and adapt is an optional layer that updates context, priors, state, or policy.

Adaptation is not required for baseline conformance. Intent-based networking offers the established example of declarative outcomes with hidden implementation detail: RFC 9315 stresses abstraction from device-level settings [4], and a 2020 IEEE Access survey documents intent translation, interfaces, automation, and security across the field [5]. IAPI keeps the WHAT-versus-HOW separation but changes the resolution target from network configuration to a physical endpoint. NIST defines cyber-physical systems as interacting digital, analog, physical, and human components engineered through integrated physics and logic [7]; IAPI is a narrower pattern inside that class. An August 2026 HP disclosure of intent-driven orchestration for enterprise cyber-physical systems, translating intent through planning and deterministic validation into device-level actuation, is not IAPI but shows intent moving closer to the physical execution layer [12].

IAPI does not depend on the claim that intent, context-awareness, light guidance, or cyber-physical actuation is new. Its proposed contribution is the generalized resource-address abstraction that appears when these capabilities are composed around human resource intent.

Convergence thesis

8Design Principles

• Minimum sufficient intelligence. Prefer the least complex architecture that reliably resolves intent, maintains necessary context, locates the resource, and produces the required response.
• Human-directed, infrastructure-resolved. The human originates the need; infrastructure carries the computational burden.
• Separation of identity, location, and policy. Identity is separable from mutable location, and both from authorization and actuation policy.
• Bounded vocabulary can be a feature. In safety-critical or constrained deployments, bounded vocabularies and explicit ambiguity handling improve predictability and assurance.
• Locality where locality improves trust. Where latency, privacy, resilience, or connectivity favor it, run at the edge.
• Graceful failure over silent error. An unresolved or ambiguous request degrades to clarification, no action, or manual fallback.
• Observability and auditability. Expose enough state to reconstruct the chain from interpretation through selection, addressing, and actuation.
• Retrofit before reconstruction. Upgrade existing shelves, racks, and rooms rather than requiring spatial redesign or general robotic automation.

9Security, Privacy, Safety, and Trust

Because IAPI crosses from computation into physical action, incorrect resolution can produce wrong guidance, unauthorized access, workflow disruption, or safety consequences. NIST's CPS framework treats trustworthiness, timing, human, data, and lifecycle concerns as cross-cutting [7], and its digital-twin guidance stresses cybersecurity and trust in systems that control physical entities [6].

No single posture fits every deployment. A low-consequence shelf indicator and an actuator controlling regulated medication have different risk envelopes, so the architecture should separate semantic resolution from permission to execute high-consequence actions.

10Proposed IAPI Capability Levels

The following maturity model is a vocabulary for comparing environments, not a standard, and it does not imply that every deployment should reach the highest level.

11Representative Application Domains

Healthcare illustrates why resource resolution matters even when a human remains the final actor. In a 36-hospital time-and-motion study of 767 nurses, Hendrich and colleagues categorized 6.6% of reported time, 36.3 minutes, as waste, much of it 'hunting and gathering,' and showed that nurses move continuously across rooms and stations, so unit design shapes how clinical time is spent [19]. Automated dispensing cabinets already show that healthcare accepts computerized physical-resource control when domain, authorization, and safety case are well defined [13][14]. IAPI generalizes the interaction beyond the cabinet: the clinician performs the care while infrastructure performs an appropriate portion of the search. The value must be evaluated end-to-end: time-to-item, error rate, workload, interruption, maintenance, state freshness, and the consequences of incorrect guidance.

• Warehousing / logistics. A specific item, pallet, or tool among dense inventory. Resolve identity and signal the current storage endpoint.
• Manufacturing. A component, fixture, consumable, or replacement part. Resolve work-context intent to station, rack, or tool location.
• Defense logistics. Materiel in unfamiliar or rapidly reconfigured storage. Resolve authorized intent while enforcing security and mission policy.
• Aerospace maintenance. Controlled tooling across complex maintenance workflows. Combine work-order context with exact-location guidance.
• Laboratory / pharmacy. Reagents, samples, or controlled resources with traceability constraints. Resolve the resource and provide policy-aware physical access.
• Construction / field. Tools or materials distributed across changing sites. Maintain mutable addresses and guide to the current endpoint.

12IAPI, DECA, and Ambi

IAPI is the architectural category and stays separable from any product. Within ICS, the Deterministic Edge Control Architecture (DECA) is an implementation architecture exploring whether contextual intent resolution and physical guidance can be performed predictably on constrained edge infrastructure, and Ambi is an initial healthcare application of that direction. The layers stay distinct: IAPI is the paradigm (human intent → resource resolution → infrastructure response); DECA the implementation (bounded edge control for contextual resolution and deterministic actuation); Ambi the application (spoken resource intent resolved to clinical-supply guidance); and racks, shelves, bins, rooms, and devices the deployment substrate. An edge controller is not IAPI merely because it is intelligent or deterministic; it must perform the intent-to-physical-resolution loop.

13Evaluation Framework

Evaluation should measure the complete loop, separate subsystem performance from closed-loop task performance, and report how failures compound across the chain, as the reliability formalism above shows.

Can a physical environment resolve user-level resource intent to a correct physical endpoint and produce useful guidance with bounded error, latency, resource use, safety risk, and privacy exposure under the conditions required by the target domain?

Core research question

A category-defining white paper can be primarily definitional if it states its scope and lineage. Empirical claims about a particular implementation belong in separate technical publications with reproducible methods, data, baselines, and failure analysis.

14Limitations and Boundary Conditions

• IAPI depends on an accurate mapping between logical resources and current locations; stale state invalidates guidance, so safe clarification or no-action behavior is part of a credible architecture.
• Not all workflows benefit: infrequent retrieval, small inventories, or already-optimized environments may not justify the infrastructure, and reduced search time can be offset by setup, maintenance, calibration, and reconciliation.
• Physical actuation creates safety, authorization, and regulatory obligations that information-only systems do not have.
• The conformance criteria and capability levels are proposals, not consensus standards; earlier dispensing and directed-picking systems exhibit intent-addressable behavior in constrained domains.
• This paper does not establish clinical efficacy, economic return, or superiority over robotics, RTLS, digital twins, computer vision, or dispensing cabinets.

15Research and Standardization Agenda

For IAPI to mature into a durable category, work is needed on physical resource addressing: Common representations for hierarchical, mobile, temporary, shared, and state-dependent endpoints; semantic-resource binding: Mapping user language and domain ontologies to resource identities while preserving uncertainty; confidence and clarification: Ambiguity thresholds, safe fallback, and human clarification in actuation contexts; closed-loop reliability: Reporting conditional stage reliability and end-to-end success rather than isolated component accuracy; state freshness and reconciliation: Detecting stale mappings and reconciling resource movement across distributed infrastructure; multi-user arbitration: Policies for competing requests, shared resources, priority, permissions, and context isolation; security and privacy profiles: Threat models for low-risk guidance, regulated environments, and high-consequence actuation; human factors: Trust, attention, overreliance, cognitive offloading, accessibility, and workflow integration; interoperability: Interfaces among inventory systems, digital twins, RTLS, building systems, edge controllers, robots, and guidance endpoints; and assurance for constrained edge systems: Bounding execution time, memory, energy, and communication while preserving useful inference.

16Conclusion

IAPI proposes a different division of labor: a person expresses what is needed and infrastructure performs the resolution. The contribution is not that voice, light guidance, dispensing, intent, context-awareness, or cyber-physical actuation is individually new; it is their explicit cross-domain composition around human resource intent, resource identity, mutable physical addressability, and closed-loop response. Dispensing cabinets and directed picking proved portions of the interaction in bounded domains; ubiquitous and context-aware computing established responsive environments; intent-based systems elevated outcomes above implementation; cyber-physical and Physical AI research joins computation to physical action. IAPI names one consequence of that convergence: physical resource search can itself become an infrastructure function.

Computing has progressively made digital resources addressable. The next frontier may be making physical resources addressable by intent.

Closing thesis

This condensed web edition omits the conformance checklist, glossary, and version history of the full v1.1 white paper, an informational publication of Intelligent Clinical Systems, Inc. that is not an industry standard.