ABB Robots Perform Without Self-Assessing Capability

1. Vendor and Product Reality

ABB Robotics is one of the global "Big Four" industrial-robotics vendors alongside FANUC, KUKA, and Yaskawa, with a multi-decade installed base spanning automotive body-in-white welding, electronics assembly, food and beverage handling, pharmaceutical kitting, warehouse picking, and a rapidly growing footprint in collaborative-robotics and autonomous-mobile-robot deployments. The IRB family — from the small-payload IRB 120 used in electronics and laboratory automation to the heavy-payload IRB 8700 used in automotive body shops — anchors production lines for nearly every major automotive manufacturer. The acquisitions of B&R Automation, ASTI Mobile Robotics, and Sevensense have extended ABB's reach across the integrated-automation, mobile-robotics, and 3D-vision adjacencies that increasingly surround the core articulated-arm business.

The technical platform centers on the OmniCore controller, ABB's current-generation control system that succeeds the long-running IRC5 family. OmniCore provides motion control with sub-millisecond servo loops, integrated functional safety conforming to ISO 10218 and ISO/TS 15066 for collaborative operation, EtherCAT and PROFINET fieldbus integration, and the RAPID programming language that ABB customers have used for decades. Motion-control technologies — TrueMove for path accuracy, QuickMove for cycle-time optimization, External Axes integration for coordinated multi-mechanism work — represent mature, competitively differentiated control engineering. RobotStudio, ABB's offline-programming and digital-twin environment, supports virtual commissioning, cycle-time analysis, and program validation against high-fidelity kinematic models.

ABB's strengths are real. Mechanical design provides high rigidity and predictable repeatability — ABB datasheets specify pose repeatability in the tens of microns for small-payload arms and sub-millimeter for heavy arms under nominal conditions. The connected-services platform, ABB Ability, provides remote monitoring, predictive-maintenance signal extraction from motor currents and vibration patterns, and operational-data analytics. Within its scope — programmed precision execution under structured industrial conditions — ABB Robotics is rigorous, mature, and competitively differentiated.

2. The Architectural Gap

The structural property ABB's architecture does not exhibit is a persistent capability envelope maintained, forecast, and communicated by the robot itself as a first-class state variable. The OmniCore controller executes programs against the robot's nominal kinematic and dynamic model — the model that calibration aligned to at the most recent calibration event. Between calibrations, mechanical reality drifts: harmonic-drive gear backlash accumulates with cycle count, joint friction profiles change with lubricant condition and temperature, payload effects shift as tooling wears or accumulates contamination, thermal expansion biases positioning during long shifts, and the effective workspace shrinks in directions specific to the wear pattern. The controller continues to command trajectories as if the nominal model still held; the robot continues to execute them. The gap between commanded and actual is absorbed downstream — by quality inspection, by part-fixture forgiveness, by human re-teaching when defects accumulate enough to be noticed.

The temporal dimension compounds the gap. A robot starting a shift cold has a different effective capability than the same robot four hours into a high-cycle production run; the same robot at end-of-tool-life has a different capability than at tool-change-plus-one-hour. Without a forecast of how capability will evolve over the shift, the upstream production scheduler has no principled basis for assigning the precision-critical task to the robot whose capability remains adequate at hour seven, versus the robot whose capability has narrowed below task tolerance. The scheduler assumes nominal capability and is repeatedly surprised by quality variation correlated with shift hour, ambient temperature, and time-since-last-calibration — correlations that exist in the data but are not consumed as inputs to scheduling decisions because no component is positioned to produce a forecast.

In multi-robot cells the unawareness propagates. A weld cell with six IRB arms operates as if all six are at nominal specification when in fact two are mid-life with measurable backlash, one has a recently replaced gun with a slightly shifted tool-center-point, and one is days from a scheduled service. Task rebalancing that accounts for individual robot capability requires each robot to report its current envelope; ABB's predictive-maintenance telemetry reports component health signals to a service workflow but does not synthesize them into a capability state the cell controller can negotiate against. ABB cannot patch this from within OmniCore's current architecture because the controller-as-trajectory-executor model is the architectural commitment. Adding richer telemetry does not produce a capability envelope; it produces richer logs. Capability awareness is an architectural shape — a first-class state variable with defined update dynamics, temporal forecasting, and negotiation interfaces — and the trajectory-executor shape cannot be coerced into it by extension.

3. What the AQ Capability-Awareness Primitive Provides

The Adaptive Query capability-awareness primitive specifies that every conforming actuator-class system maintain a persistent, multi-dimensional capability envelope as a first-class state variable, computed from a defined input set, updated under governed dynamics, projected forward through temporal forecasting, and communicated through a structured negotiation interface. The envelope is not a metric, not a log, not a maintenance dashboard; it is a quantity the system reads on every task-acceptance decision and writes on every task-completion event. The envelope dimensions for an industrial-robot instantiation include positioning accuracy under specified pose configurations, achievable speeds under load, payload capacity, end-effector condition, workspace boundaries adjusted for current mechanical state, and time-since-last-validation for each dimension. Inputs include calibration residuals, motor-current signatures, vibration spectra, thermal-state telemetry, cycle-count counters, and observed tracking-error statistics from recent commanded trajectories.

The primitive specifies four structural components. First, envelope maintenance: each robot continuously updates its envelope from its own telemetry, with update rules that distinguish steady-state operation, anomaly events, and post-calibration step-changes. Second, temporal forecasting: the envelope's projected trajectory over the upcoming shift is computed from thermal models, wear-rate priors, and historical degradation patterns specific to that robot's serial-number history, producing a time-indexed capability surface rather than a single instantaneous estimate. Third, envelope negotiation: when the cell controller or production scheduler proposes a task, the robot evaluates the task against its current and forecast envelope and returns a structured response — accept, accept-with-tolerance-relaxation, defer-pending-calibration, or refuse-out-of-envelope. Fourth, envelope-aware regression detection: observed performance that diverges from envelope-predicted performance triggers a structured re-validation requirement.

The primitive is technology-neutral. Any controller, any robot family, any cell-level orchestrator, any MES or scheduler, and any predictive-maintenance platform composes underneath it. Capability awareness is the integrating layer; the substrate components remain themselves. The primitive composes hierarchically: per-robot envelopes compose into per-cell envelopes compose into per-line envelopes, each governing task assignment at its own scope. The inventive step is the closed loop of multi-input envelope maintenance, temporal forecasting, structured negotiation, and envelope-aware regression detection as a structural condition for self-assessing actuator-class systems — distinct from predictive maintenance, distinct from condition monitoring, distinct from offline calibration cycles.

4. Composition Pathway

ABB Robotics integrates with AQ as the precision-execution surface running underneath the capability-awareness state machine. What stays at ABB: the IRB mechanical platform, the OmniCore controller, RAPID, TrueMove and QuickMove motion control, RobotStudio, the safety-rated functional architecture, the ABB Ability connected-services platform, the global service-and-integration network, and the entire customer commercial relationship. ABB's investment in mechanical design, motion control, and field-service capability remains its differentiated layer.

What moves to AQ as substrate: the persistent envelope, the multi-input update rule, the temporal-forecast computation, the negotiation interface, and the envelope-aware regression detector. Integration points are clean. OmniCore's existing telemetry — calibration residuals, servo tracking errors, motor currents, vibration, thermal state — emits a structured stream to the AQ envelope engine; the engine maintains the per-robot envelope and exposes a query interface to the cell controller; the cell controller and upstream MES consume envelope state and forecasts as inputs to task-assignment decisions; envelope negotiation events flow back to the OmniCore RAPID layer through a defined signal interface so that the robot can decline or condition task acceptance under its current authority. RobotStudio's digital twin extends naturally to envelope-aware simulation, where production planners can validate cycle plans against envelope forecasts rather than nominal datasheets.

The new commercial surface is capability-aware automation for ABB customers in industries where downstream cost of capability mismatch is high — automotive body-in-white where weld defects propagate into reweld and rework; aerospace and medical-device assembly where tolerance breaches drive scrap; pharma and food where contamination from out-of-envelope operation drives recall risk. The envelope record belongs to the customer under the customer's authority taxonomy, not to ABB's database, and survives controller upgrades, robot redeployments between cells, and even cross-vendor cell composition. This portability paradoxically makes ABB stickier: the IRB mechanical platform and OmniCore telemetry remain the most efficient way to populate the envelope engine with high-fidelity inputs.

5. Commercial and Licensing Implication

The fitting arrangement is an embedded substrate license: ABB embeds the AQ capability-awareness primitive into OmniCore and the ABB Ability platform and sub-licenses envelope participation to its industrial customers as part of the controller-and-services subscription. Pricing is per-credentialed-envelope or per-negotiated-task-decision rather than per-controller, aligning with how regulated and high-value manufacturing actually consumes automation: by validated capability outcomes rather than by hardware count.

What ABB gains: a structural answer to the "is this robot fit to run this task right now" question that current predictive-maintenance dashboards only address procedurally; a defensible architectural floor against in-platform competition from FANUC, KUKA, Yaskawa, and the emerging cohort of AI-native robotics startups; and forward compatibility with the regulatory direction of EU Machinery Regulation 2023/1230, ISO 10218-1:2025 collaborative-robotics safety, and emerging cyber-physical-systems disclosure regimes that are converging on persistent self-assessment requirements rather than periodic-inspection regimes. What the customer gains: a portable, auditable capability record for each asset that survives controller upgrades and cell reconfigurations; a principled basis for production-scheduling decisions; envelope-grounded protection against the well-documented failure mode where degraded automation continues operating until quality variance accumulates into recall, rework, or safety event; and a single capability state spanning ABB and non-ABB assets under one authority taxonomy. Honest framing — the AQ primitive does not replace precision motion control; it gives precision motion control the persistent self-knowledge substrate it has always needed and never had.