Evidence-led research and interpretation
Classification is a retrieval index, not a technology verdict
The IPC and CPC are formal systems for organizing subject matter in patent records; neither maps one-to-one onto the commercial label robotics and autonomous machines. A searcher should begin with the exact invention features, then inspect the current scheme titles, notes, definitions and linked documents before retaining symbols. The boundary here covers physical sensing, motion planning, actuation, safety interlocks and machine autonomy. A useful classification guide discloses the date and version of the official scheme, not only a shorthand code from an undated third-party table.
Five technical entry points
Potential search entry points include perception and sensor fusion; motion planning and trajectory constraints; robot manipulation and end-effectors; industrial or field operating context; human-robot safety and intervention. Search each separately before combining results. When a document touches several functions, record which claim or passage supports each association. The IPC main group may be too broad for the problem, while a narrow CPC subgroup may omit relevant historical patent families or applications indexed under another office's practice. Do not infer that a symbol belongs to this sector until its official scope note is checked.
Worked feature-to-symbol exercise
Consider a warehouse mobile robot that uses a camera and range sensor to plan an obstacle-avoiding path. First underline the invention's claimed physical or digital features; then distinguish those features from its intended commercial setting. Compare candidate IPC and CPC entries by definition, inspect a small positive sample of documents and an intentionally negative sample, and record false-positive reasons. Retain separate columns for claim feature, classification version, symbol candidate, scope note, test document and reviewer decision. This illustrates a method; no unverified code is proposed as a definitive classification.
Boundary conditions and nearby art
In robotics and autonomous machines, adjacent sectors may share keywords while describing different technical contributions. A generic software planning claim can coexist with a much narrower claimed mechanical actuator. Therefore a query must combine relevant synonyms with technical mechanisms and relevant classification paths. Expand the query using citation trails and related applications only after checking whether each expansion introduces off-topic material. The goal is recall with reviewable precision, not a perfectly precise classification-only filter.
Classification test: Perception and sensor fusion
**Classification decision and source test.** Start with camera calibration, range-sensor timing and coordinate frames as observable claim features, not as an assumed class name. Build one query from the apparatus or method features and a second from functional synonyms. The essential distinction is a claim that merely names a sensor versus a claim that specifies how uncertain measurements are combined; the latter may yield different official IPC/CPC paths even where the commercial vocabulary overlaps. Inspect the hierarchy and scope notes in the dated [WIPO IPC publication](https://ipcpub.wipo.int/) and [official CPC scheme](https://www.cooperativepatentclassification.org/cpcSchemeAndDefinitions) before accepting any symbol. Record a candidate symbol only after examining a relevant publication and its claim language.
**Negative-control experiment.** Select an example whose title contains terms associated with perception and sensor fusion but whose claims solve a different engineering problem. Compare it with a positive example specifying camera calibration, range-sensor timing and coordinate frames. Note whether the candidate symbol retrieves the negative case, whether it misses positive cases, and which adjacent class should be inspected next. Importantly, an obstacle detector may work in a mapped warehouse but fail on reflective surfaces. This is a retrieval-design procedure, not a verified assignment of a particular class or a claim that the sample is exhaustive. Record classification version, sample identifiers and exclusion reasons for reproducibility.
Classification test: Motion planning and trajectory constraints
**Classification decision and source test.** Start with path optimization, collision envelopes and real-time replanning as observable claim features, not as an assumed class name. Build one query from the apparatus or method features and a second from functional synonyms. The essential distinction is a trajectory solver from a separate actuator controller; the latter may yield different official IPC/CPC paths even where the commercial vocabulary overlaps. Inspect the hierarchy and scope notes in the dated [WIPO IPC publication](https://ipcpub.wipo.int/) and [official CPC scheme](https://www.cooperativepatentclassification.org/cpcSchemeAndDefinitions) before accepting any symbol. Record a candidate symbol only after examining a relevant publication and its claim language.
**Negative-control experiment.** Select an example whose title contains terms associated with motion planning and trajectory constraints but whose claims solve a different engineering problem. Compare it with a positive example specifying path optimization, collision envelopes and real-time replanning. Note whether the candidate symbol retrieves the negative case, whether it misses positive cases, and which adjacent class should be inspected next. Importantly, a simulation result cannot establish reliable operation under latency or payload changes. This is a retrieval-design procedure, not a verified assignment of a particular class or a claim that the sample is exhaustive. Record classification version, sample identifiers and exclusion reasons for reproducibility.
Classification test: Robot manipulation and end-effectors
**Classification decision and source test.** Start with grasp geometry, force feedback and compliance as observable claim features, not as an assumed class name. Build one query from the apparatus or method features and a second from functional synonyms. The essential distinction is novel contact control versus standard gripper packaging; the latter may yield different official IPC/CPC paths even where the commercial vocabulary overlaps. Inspect the hierarchy and scope notes in the dated [WIPO IPC publication](https://ipcpub.wipo.int/) and [official CPC scheme](https://www.cooperativepatentclassification.org/cpcSchemeAndDefinitions) before accepting any symbol. Record a candidate symbol only after examining a relevant publication and its claim language.
**Negative-control experiment.** Select an example whose title contains terms associated with robot manipulation and end-effectors but whose claims solve a different engineering problem. Compare it with a positive example specifying grasp geometry, force feedback and compliance. Note whether the candidate symbol retrieves the negative case, whether it misses positive cases, and which adjacent class should be inspected next. Importantly, an arm-level performance claim may rely on an unchanged end-effector. This is a retrieval-design procedure, not a verified assignment of a particular class or a claim that the sample is exhaustive. Record classification version, sample identifiers and exclusion reasons for reproducibility.
Classification test: Industrial or field operating context
**Classification decision and source test.** Start with dust, vibration, lighting, fleet coordination and task constraints as observable claim features, not as an assumed class name. Build one query from the apparatus or method features and a second from functional synonyms. The essential distinction is a transport robot algorithm from an entire warehouse system; the latter may yield different official IPC/CPC paths even where the commercial vocabulary overlaps. Inspect the hierarchy and scope notes in the dated [WIPO IPC publication](https://ipcpub.wipo.int/) and [official CPC scheme](https://www.cooperativepatentclassification.org/cpcSchemeAndDefinitions) before accepting any symbol. Record a candidate symbol only after examining a relevant publication and its claim language.
**Negative-control experiment.** Select an example whose title contains terms associated with industrial or field operating context but whose claims solve a different engineering problem. Compare it with a positive example specifying dust, vibration, lighting, fleet coordination and task constraints. Note whether the candidate symbol retrieves the negative case, whether it misses positive cases, and which adjacent class should be inspected next. Importantly, a deployment claim may reflect site infrastructure rather than robot design. This is a retrieval-design procedure, not a verified assignment of a particular class or a claim that the sample is exhaustive. Record classification version, sample identifiers and exclusion reasons for reproducibility.
Classification test: Human-robot safety and intervention
**Classification decision and source test.** Start with safe stopping, operator override and contact-force limitation as observable claim features, not as an assumed class name. Build one query from the apparatus or method features and a second from functional synonyms. The essential distinction is a physical interlock from a monitoring display; the latter may yield different official IPC/CPC paths even where the commercial vocabulary overlaps. Inspect the hierarchy and scope notes in the dated [WIPO IPC publication](https://ipcpub.wipo.int/) and [official CPC scheme](https://www.cooperativepatentclassification.org/cpcSchemeAndDefinitions) before accepting any symbol. Record a candidate symbol only after examining a relevant publication and its claim language.
**Negative-control experiment.** Select an example whose title contains terms associated with human-robot safety and intervention but whose claims solve a different engineering problem. Compare it with a positive example specifying safe stopping, operator override and contact-force limitation. Note whether the candidate symbol retrieves the negative case, whether it misses positive cases, and which adjacent class should be inspected next. Importantly, an apparent safety improvement may depend on training or a separate protective device. This is a retrieval-design procedure, not a verified assignment of a particular class or a claim that the sample is exhaustive. Record classification version, sample identifiers and exclusion reasons for reproducibility.
Search reproducibility and version drift
Archive the exact scheme release, document date range, patent corpus, query syntax, sorting rule and export timestamp. The official CPC scheme may be updated and previously assigned classifications may be reclassified. A comparison across years should therefore use one documented crosswalk or disclose the residual mismatch. Compare navigation and manipulation claim features separately; do not treat every mention of artificial intelligence as a robot invention. Patent family deduplication prevents one invention appearing as several unrelated signals, but is not a substitute for a claim-level review.
Evidence interpretation and legal limits
A patent's presence under a relevant symbol does not imply that any particular product infringes it, that it is enforceable, or that its applicant has significant market sales. Read the claims, filing history and available legal-status documents before forming even provisional conclusions. This public guide provides a search architecture rather than legal advice. A search query can support a research inventory; freedom-to-operate and validity opinions require separate jurisdiction-specific professional review.
Review protocol and expected audit trail
For a robotics and autonomous machines classification project, record a positive sample, an adverse sample and at least two variants of the feature vocabulary. Explain why symbols were kept, rejected or deferred. Return a table that identifies the source passage for each classification decision, a named reviewer role and a date for revision. If a new patent or official classification entry changes coverage, amend the table without silently revising earlier data. This is an information-quality procedure rather than evidence of an actual completed mapping.
Questions for future research
A specialist can extend this reference by asking whether a new technical architecture changes the mapping, whether relevant non-patent literature uses different vocabulary, and whether a shift in classification reflects official restructuring rather than innovation growth. Answer each question with named sources and documented exclusions. Keep robotics and autonomous machines as an explanatory sector label while treating IPC/CPC entries as individual versioned search instruments.
Primary sources, version and scope
Verify relevant claims directly against: WIPO IPC 2026.01; Official CPC scheme 2026.08. The referenced publications establish their own stated methods or definitions, not the outcome of an unperformed research assignment. Where a specific claim relies on a patent record, office manual or institutional policy, inspect that original document before advancing the claim. Checked for this planning build on 15 November 2026.
| Research unit | Evidence needed | Review limitation |
|---|---|---|
| perception and sensor fusion | Versioned primary document and identifiable passage | Unpublished records, database limits and interpretive uncertainty |
| motion planning and trajectory constraints | Versioned primary document and identifiable passage | Unpublished records, database limits and interpretive uncertainty |
| robot manipulation and end-effectors | Versioned primary document and identifiable passage | Unpublished records, database limits and interpretive uncertainty |
| industrial or field operating context | Versioned primary document and identifiable passage | Unpublished records, database limits and interpretive uncertainty |





