Vendor and Product Reality
Uber Freight operates a large digital freight brokerage, with a marketplace that matches shippers, including manufacturers, consumer-goods companies, and third-party logistics providers, against a carrier base that ranges from single-truck owner-operators to large asset-based fleets. The Uber Freight Transportation Management product, expanded after the acquisition of Transplace, layers managed-services capability on top of the spot marketplace and gives Uber Freight a planning view across shipper networks rather than transaction by transaction. Autonomous capacity has been integrated through structured partnerships rather than fleet ownership: Aurora Innovation has publicly operated autonomous freight on Texas lanes including the Dallas to Houston corridor, and Uber Freight has announced integration relationships with autonomous-driving developers such as Waabi to make autonomous capacity available on its network. These are described here at the architecture level as publicly announced integration relationships; the specific status, scope, and certification posture of any individual partnership is that partner's to state and changes over time.
The operational pattern is consistent across partners and is structurally important: autonomous tractors handle the highway middle-mile while human drivers perform the first and last legs at transfer hubs, producing a hybrid driver-plus-autonomous load that is matched, dispatched, billed, and settled through Uber Freight's marketplace as a single commercial unit. The marketplace must therefore reason about a load that is decomposed into at least three legs, pickup yard to origin hub, hub to hub on the autonomous lane, destination hub to consignee, where each leg has different carrier constraints and different failure modes. A load committed to an autonomous middle-mile under one set of weather and configuration assumptions is not the same load if those assumptions change after dispatch.
Uber Freight controls the marketplace economics, the shipper relationship, and the commercial commitment to deliver the load on time at the contracted rate; the AV partner controls the autonomous stack and its operational eligibility for a specific lane at a specific time. The dispatch decision sits between these and is currently expressed in ad-hoc integration logic that varies per partner.
The Architectural Gap
The freight-matching engine treats a dispatch as a binding tender to a carrier: the carrier accepts, the load is committed, and the marketplace assumes responsibility for fulfillment. Hybrid driver-plus-autonomous loads break this model because the commitment is no longer to a single carrier but to a composite of human carriers and an autonomous-driving developer whose eligibility for the middle-mile leg is conditioned on factors the marketplace does not natively model. Autonomous-driving systems in general define an operational design domain, the bounded set of conditions, including roadway, geography, weather, and speed, under which the system is designed to operate, and each developer maintains its own ODD and its own release process. This is a widely stated architectural property of the field rather than a claim about any one company. What no marketplace-plus-partner integration inherently supplies is a uniform, credentialed schema through which the matching engine can reason about whether a particular load, at a particular pickup time, is currently eligible for autonomous fulfillment, with a graduated fall-back to human capacity if eligibility drops.
The hub-handoff structure compounds the gap. A committed load depends on a human driver being available at the destination hub at the time the autonomous tractor arrives, and that human-side capacity must itself be matched and committed before the autonomous middle-mile is sensible to dispatch. The marketplace currently models these as two separate commitments stitched together in operational tooling, which means a failure on the human-leg side after autonomous dispatch produces stranded freight and contractual exposure. There is no native primitive that says: commit the autonomous middle-mile only if both hub-leg capacities are jointly committable, and otherwise defer the autonomous commitment until the composite is feasible.
Post-dispatch verification is similarly missing from a pure marketplace-plus-partner integration. Once an autonomous middle-mile is committed, the marketplace has limited structured visibility into whether the leg executed as committed, whether the vehicle remained on the intended lane, whether handoffs occurred at the agreed hubs, and whether human-oversight escalation events occurred and under what conditions. Settlement and shipper reporting therefore tend to rely on partner-supplied summaries rather than a marketplace-verifiable actuation record.
What The AQ Primitive Provides
As disclosed in the filing, governed actuation treats a proposed physical actuation, here, a dispatch, as an input to a composite admissibility evaluator that produces one of a plurality of outcomes rather than a binary admit-or-reject. The disclosed outcomes are admit, gate, defer, solicit, reject, and escalate. Mapped onto dispatch: admit commits the load to the dispatched composite under the credentialed conditions in force at the moment of commit; gate commits at reduced weight or subject to additional constraints; defer holds the load with explicit re-evaluation triggers tied to changed weather, updated ODD declarations, hub-side capacity, or software-configuration updates; solicit emits a governed discovery query requesting the additional credentialed observations needed to resolve uncertainty before committing; reject declines the autonomous fulfillment path with a structured, lineage-recorded reason and routes the load to human carriage; and escalate refers the decision to a higher-authority endpoint. The evaluator reasons over authority-credentialed observations weighted by an evidential-weighting mechanism, so a first-party ODD declaration and a third-party telematics inference are not treated homogeneously.
Beyond the admit-or-reject axis, the filing discloses a graduated-actuation mode selector that chooses how an admitted actuation executes, spanning at least disabled, simulated, advisory, consultative, shadowed, partial, constrained, stage-gated, deferred, full, and emergency-accelerated modes, with a continuous mapping from composite-admissibility determination to mode. For dispatch this supports graceful degradation: as confidence in autonomous eligibility falls, the selector can move from full commitment toward stage-gated, partial, or advisory execution instead of forcing a binary all-or-nothing outcome, and can de-escalate a commitment already in progress when a newly arriving observation indicates reduced admissibility. The partial mode maps naturally onto composite freight, committing the autonomous middle-mile while leaving a destination-hub leg as a deferred sub-commitment.
The filing further discloses a reversibility-aware commitment-point evaluator that classifies each proposed actuation into reversibility classes, including reversible, partially reversible, time-bounded reversible, and irreversible, and identifies the commitment point beyond which the actuation cannot be unwound. Applied to dispatch, this distinguishes commitments that can still be reversed, such as a deferred autonomous leg or a not-yet-tendered destination-hub leg, from commitments that cannot, such as a load already in transit on the autonomous lane, and prefers reversible paths and late commitment points among admissible candidates. Every admissibility outcome, mode selection, reversibility classification, and downstream verification result is written to a lineage field, producing a deterministic provenance record that downstream settlement, shipper reporting, and regulatory inquiry can all draw from.
Composition Pathway
The governed-actuation step composes with the prior primitives in the chain in a manner that maps onto a marketplace architecture. Authority-credentialed observation provides the signed inputs the actuation gate reasons over: the developer's current ODD declaration, a weather provider's authenticated forecast, the relevant lane permissions, and the human-carrier Hours-of-Service representation for the hub legs. Evidential weighting then normalizes these credentialed observations into confidence-weighted views, so a first-party ODD declaration is weighted differently from a third-party telematics inference, and the gate composes them without collapsing the difference.
Composite admissibility tests the joint feasibility that hybrid loads require. A hybrid commitment is admissible only if the autonomous-leg credentials, the origin-hub human capacity, and the destination-hub human capacity are all individually admissible and jointly compatible at the dispatch timestamp. The composite admissibility evaluator disclosed in the filing captures exactly this kind of multi-field structural test, and without it a dispatch gate would be vulnerable to commitments that look acceptable component by component but fail jointly. Lineage-recorded provenance closes the audit surface, capturing the inputs, the evaluator's decision, the mode selected, and the verification result in a single deterministic record.
For a marketplace integrating multiple autonomous-driving developers, the composition pathway means middle-miles from different developers can flow through the same actuation gate, with credentialed configurations that differ per partner but obey a uniform schema. The marketplace gains a single governed dispatch surface across a heterogeneous autonomous-supply base rather than a bespoke integration per partner.
Commercial and Licensing Implication
Uber Freight's commercial position is that of a marketplace and managed-services platform whose competitive advantage compounds with the structured visibility it can offer shippers and carriers. Autonomous capacity multiplies that visibility problem unless it is integrated through a uniform governance surface; without one, every new autonomous-driving partner adds another bespoke integration and another category of dispatch failure mode that operations teams must manage manually. Applying the governed-actuation step to a transportation-management stack converts the dispatch decision from ad-hoc partner integration logic into a structured commitment surface that shippers, carriers, regulators, and insurers can reason about in the same terms.
The commercial implications are concrete. Shipper contracts increasingly specify on-time-in-full performance with structured exception reporting, which the primitive supplies natively through its post-actuation verification record. Carrier settlement on hybrid loads depends on disambiguating responsibility across the autonomous middle-mile and the human hub legs, which the actuation gate's mode selection and lineage record capture by construction. Insurance underwriting on hybrid freight requires structured per-load documentation of the conditions under which each leg was committed, which an actuation gate of this kind produces by construction. The step is competitively meaningful because it sits at the marketplace boundary a broker already controls and converts that boundary into a governance substrate over a fragmenting autonomous-trucking supply landscape.
Enablement and Embodiments
A skilled implementer can build the approach described here. Represent each proposed dispatch as an actuation request carrying the load composite (pickup leg, autonomous middle-mile, delivery leg), the candidate credentialed observations (developer ODD declaration, authenticated weather forecast, lane permissions, hub-side Hours-of-Service capacity), and a timestamp. Feed these to a composite admissibility evaluator that computes a composite evidential weight from authority, staleness, modality, and dispositional factors and emits one of admit, gate, defer, solicit, reject, or escalate. Route the outcome through a graduated-actuation mode selector implementing a configurable mapping from admissibility to mode (disabled, simulated, advisory, consultative, shadowed, partial, constrained, stage-gated, deferred, full, emergency-accelerated), and through a reversibility-aware commitment-point evaluator that classifies the actuation, locates its commitment point, and prefers reversible paths and late commitment among admissible candidates. Write every determination to a lineage field with the contributing inputs and decision.
The approach is not limited to trucking. Contemplated embodiments include, without limitation: any brokered or dispatched physical-fulfillment marketplace (drayage, last-mile delivery, rail intermodal, maritime port operations); centralized, distributed, and hybrid evaluator topologies; per-actuator, per-authority, and per-deployment-domain configuration of thresholds, modes, and reversibility ontology; single-developer and multi-developer autonomous-supply bases; and human-only, autonomous-only, and hybrid composite loads. The credentialed-observation inputs, the enumerated admissibility outcomes, the enumerated actuation modes, and the reversibility classes are each extensible under governance policy without departing from the disclosed architecture.
Disclosure Scope
The inventive subject matter described in this article, governed actuation gated by composite admissibility over credentialed observations, graduated actuation modes, reversibility-aware commitment-point evaluation, and lineage-recorded actuation provenance, is disclosed in U.S. Provisional Application No. 64/049,409. This article is a dated public description of that disclosure as of its publication date.
References to Uber Freight, Aurora Innovation, Waabi, and any other named company, product, partnership, or corridor are external market context describing the state of the autonomous-trucking field, not claims of the filing. Those descriptions are stated at the architecture level as public fact and may change over time; the specific status, scope, certification posture, and operational details of any named party are that party's to state. Nothing here asserts a defect, incident, capability, certification, contract, or figure of any named company beyond what is publicly and neutrally described, and no comparison is intended as disparagement. The scope of the invention is defined solely by U.S. Provisional Application No. 64/049,409 and any application claiming priority to it.
