Words worth knowing

A named scale that gives a number physical meaning, such as metres, seconds, or volts.

A unit is an agreed reference quantity; algebra on units exposes incompatible equations and conversion errors.

The kind of quantity being described, independent of the chosen unit.

Dimensions such as length, time, and current combine algebraically; a valid physical equation must be dimensionally homogeneous.

A deliberately simplified, bounded answer used to guide a decision before exact data exist.

An estimate states assumptions, scale, uncertainty, and validity range; it may combine models and measurements without claiming exactness.

How quickly energy is transferred or converted.

Power is the time derivative of energy; electrical DC power is VI and mechanical rotational power is torque times angular speed.

Digits that honestly communicate the resolution supported by inputs and measurement.

Rounding should reflect uncertainty and instrument resolution, while guard digits may be retained during intermediate calculations.

A statement that two mathematical expressions represent the same value.

Equivalent transformations apply the same valid operation to both sides while preserving domain restrictions and units.

A named symbol whose value may change or be unknown.

A variable has a domain, meaning, units, and often a frame or time index that must remain consistent.

A trigonometric ratio connecting an angle to the opposite side of a right triangle.

For angle θ, sin θ = opposite/hypotenuse and is also the y-coordinate on the unit circle.

A trigonometric ratio connecting an angle to the adjacent side of a right triangle.

For angle θ, cos θ = adjacent/hypotenuse and is the x-coordinate on the unit circle.

An angle unit defined by arc length divided by radius.

One full turn is 2π radians; radian measure makes derivatives and angular equations take their standard form.

A quantity described by magnitude without a spatial direction.

A scalar is invariant under coordinate rotation, though its units and uncertainty still matter.

A quantity with magnitude and direction.

Vector components depend on a basis and transform between frames while geometric magnitude is preserved by pure rotation.

A named origin and set of axes used to describe positions and directions.

Frame transforms combine rotation and translation; free vectors use rotation, while points also depend on origin offset.

A matrix that changes vector coordinates between rotated axes.

A proper rotation matrix is orthogonal with determinant +1, so its transpose is its inverse and it preserves lengths.

One signed amount of a vector along a chosen axis.

Components are projections onto basis directions and reconstruct the vector only when paired with that basis.

How much one quantity changes per unit of another, commonly per time.

An instantaneous rate is a derivative; an interval rate is a finite difference and may hide variation within the interval.

The local rate at which one quantity changes with another.

The derivative is the limit of a difference quotient; sampled estimates depend on interval, noise, and numerical method.

Depending on context, either accumulating small contributions mathematically or joining subsystems through tested interfaces.

A definite integral is a limit of weighted sums; system integration instead verifies that electrical, data, timing, mechanical, and behavioral contracts work together.

Elapsed time between consecutive measurements.

The interval sets temporal resolution and appears explicitly in discrete integration, differentiation, and aliasing limits.

A gradual departure from the true or desired value over time.

Persistent bias integrated through a model produces unbounded or growing state error unless corrected by an external reference.

A quantified doubt about which values could reasonably correspond to the measurand.

Uncertainty combines stated sources and coverage assumptions; it is distinct from tolerance, error, and repeatability.

The sum of observations divided by their count.

The sample mean estimates a distribution's central value but is sensitive to outliers and does not describe spread.

A measure of how far observations typically spread around their mean.

Sample standard deviation is the square root of squared deviations divided by n−1; interpretation depends on sampling and distribution.

In measurement it means close agreement among repeats; in classification it means the fraction of positive predictions that are correct.

Measurement precision describes dispersion under stated conditions. Classifier precision is TP/(TP+FP), depends on threshold and evaluation prevalence, and is distinct from recall.

How close a result is to the accepted reference or true value.

Accuracy is qualitative unless tied to a defined error or uncertainty statement; it includes systematic as well as random effects.

A repeatable tendency for results to lie on one side of the target.

Statistical bias is expected estimate minus true parameter; sensor bias often appears as offset or scale error and accumulates when integrated.

An instrument for measuring quantities such as voltage, current, resistance, and continuity.

Different modes alter the meter's internal circuit; voltage is measured in parallel, current in series, and resistance on unpowered circuits.

Evidence that two points are joined by a sufficiently low-resistance path.

A meter sources a small test signal and beeps below an instrument-specific threshold; parallel paths can affect the result.

The smallest change an instrument or representation can distinguish or display.

Resolution follows display counts, quantization, range, and noise; it is separate from accuracy and repeatability.

A change in the system caused by connecting the measuring instrument.

Finite input impedance, burden voltage, capacitance, mass, or thermal contact alters the measurand and must be included in the model.

The circuit node assigned zero volts so other node voltages have a common meaning.

Voltage is always a difference; choosing a reference simplifies equations but does not necessarily connect it to earth.

The voltage drop introduced by a current-measuring instrument.

An ammeter's shunt and protection path add series impedance, potentially reducing circuit current and changing operation.

A specific explanation that predicts observable outcomes.

A useful hypothesis identifies conditions and discriminating evidence that could support or contradict it.

A comparison condition kept unchanged so an intervention's effect can be distinguished.

Controls expose background drift, confounds, and measurement artifacts by preserving all relevant conditions except the tested factor.

The factor deliberately changed or selected during an experiment.

Its levels are controlled or categorized to estimate their effect while other influential conditions are constrained or recorded.

The measured outcome expected to respond to an experimental change.

A response needs an operational definition, instrument, units, timing, and analysis rule chosen before observing results.

An uncontrolled factor that changes with the tested factor and offers another explanation.

Confounding prevents unique attribution because treatment and nuisance effects are not separately identifiable from the data.

Agreement among results repeated over a short time under the same stated conditions.

Repeatability is a precision condition with the same method, operator, instrument, location, and object; reproducibility changes conditions.

A testable statement of what a system must do or withstand.

A sound requirement names quantity, threshold, conditions, and verification method and remains distinct from a chosen implementation.

A limit the design must obey rather than optimize away.

Constraints bound feasible solutions through geometry, budget, interfaces, safety, materials, or equations; CAD constraints also encode geometric relationships.

A temporary implementation built to answer specific design questions.

Prototype fidelity should match the uncertainty being tested; form, function, interface, and production fidelity can differ.

Checking with evidence that an implementation meets a stated requirement.

Verification uses inspection, analysis, demonstration, or test with acceptance criteria and traceable conditions.

An explicit chain connecting a claim to its requirement, decision, evidence, or reference.

In projects it links requirements to tests and revisions; in metrology it links results through calibrated references with stated uncertainties.

A table that compares alternatives against explicit criteria.

Scores and weights structure judgment but require sensitivity checks because correlated criteria and arbitrary weights can dominate.

Energy transferred per unit charge between two points.

Voltage is a difference in electric potential and must name polarity or reference; it drives current only through a complete circuit model.

The rate at which electric charge crosses a boundary.

Conventional current is dq/dt; its value follows the complete circuit, source limits, and load rather than being independently pushed into every device.

How strongly a circuit element opposes current for a given voltage.

Static resistance is V/I at an operating point; only ohmic elements maintain a constant ratio across the relevant range.

The relationship among voltage, current, and resistance for an ohmic element.

V = IR models a linear resistor under its stated thermal and operating conditions; it is not a universal component law.

The rate at which a circuit element receives or delivers electrical energy.

Using passive sign convention, p = vi is absorbed power; resistor dissipation can be written I²R or V²/R.

The declared positive and negative orientation of voltage, current, or a polarized component.

Polarity establishes sign conventions and safe component orientation; reversing it changes equations and can damage polarized devices.

A set of conductor points treated as sharing one electric potential.

Ideal wires join a node with zero drop; components connect between nodes and node voltages form circuit unknowns.

A connection between two nodes containing one or more series elements and one branch current.

Branches define graph edges for Kirchhoff analysis; all elements in a non-splitting branch share current.

Current entering a node balances current leaving it.

KCL follows charge conservation when charge accumulation at the ideal node is negligible.

Signed voltage changes sum to zero around a closed circuit loop.

KVL follows conservative electric potential in lumped circuits; rapidly changing magnetic flux requires the full induced-emf model.

Elements connected in one unbranched path so the same current passes through them.

Ideal series resistances add; voltage division follows impedance, not merely component count.

Elements connected across the same two nodes and therefore sharing voltage.

Parallel conductances add; branch currents depend on each branch impedance.

A measure of stored charge per volt between conductors.

For an ideal capacitor q = Cv and i = C dv/dt; geometry and dielectric determine C.

A fundamental electrical quantity whose movement constitutes current.

Charge is conserved; current integrates to transferred charge and a capacitor stores q = Cv in the lumped model.

A timescale describing how quickly a first-order system approaches a new value.

After one τ a first-order rise completes about 63.2% of its total change, or a decay retains about 36.8%.

Temporary behavior while a system moves between operating conditions.

Transients arise from stored energy, dynamics, switching, and initial conditions before steady behavior dominates.

A change whose rate is proportional to its remaining distance from a limit.

First-order homogeneous dynamics produce e^(−t/τ); equal time intervals multiply the remaining error by a constant factor.

Small unintended current through a nominally insulating or off path.

Real capacitors, semiconductors, boards, and instruments have finite insulation resistance that changes held voltages over time.

A two-terminal device that conducts much more readily in one direction.

Its nonlinear current-voltage relation and reverse limits require an operating-point model rather than a universal fixed voltage drop.

A controllable device used to switch or amplify electrical signals.

A small terminal signal controls a larger current or voltage subject to topology, bias, ratings, and dissipation.

A voltage-controlled transistor often used as an efficient electronic switch.

Gate-source voltage controls channel resistance; safe design checks threshold versus full enhancement, charge, VDS, current, and thermal limits.

A diode that gives inductive current a safe path when a switch turns off.

It clamps the inductive voltage by recirculating current, with polarity reverse-biased during normal energization.

The control terminal whose voltage establishes a MOSFET's channel state.

The insulated gate draws little DC current but has capacitance and charge that require transient drive current during switching.

A condition where increasing a command no longer produces the assumed proportional output.

In control it is an actuator limit; in a BJT it names a junction-bias region. Context must be stated because the models differ.

A table of expected electrical loads and the margins assigned to their shared supply.

A defensible budget separates voltage rails, steady and credible peak current, duty, conversion loss, startup, wiring drop, and thermal derating.

A circuit that maintains a target output voltage across an allowed input and load range.

Linear regulators dissipate voltage difference as heat; switching converters transfer energy through switched storage elements with finite efficiency and ripple.

The fraction of input energy or power delivered as useful output.

Efficiency is output divided by input for the same boundary and interval; the remainder becomes loss, often heat, sound, or deformation.

The minimum input-to-output voltage difference a regulator needs to stay regulated.

Below dropout, output follows input and load according to the regulator topology rather than remaining at its set value.

The gap between predicted or measured temperature stress and an allowed limit.

It depends on dissipation, ambient, airflow, package thermal resistance, duration, mounting, and component derating.

A shared conductor or network distributing a named supply voltage.

A rail has source impedance, current limits, return paths, transients, and tolerance rather than being an ideal voltage everywhere.

The quantity of charge a battery can deliver under specified test conditions.

Ampere-hours integrate current over time; usable capacity varies with cutoff, rate, temperature, age, and chemistry and does not alone state energy or peak current.

An energy unit equal to one watt sustained for one hour.

For an approximately constant nominal battery voltage, energy is estimated as voltage times ampere-hours, with discharge and conversion limits stated separately.

An equivalent resistance representing voltage loss and heating inside a real source.

Terminal voltage falls approximately by I·R_internal under a simple local model; resistance varies with charge, temperature, age, and frequency.

A threshold at which circuitry disconnects charge or discharge to protect the battery or system.

Cutoff behavior includes threshold, delay, hysteresis, current, temperature, and recovery conditions and may differ between charger, pack, and load.

Circuitry that applies a chemistry-specific charging profile within voltage and current limits.

Charging may include preconditioning, constant-current, constant-voltage, termination, temperature checks, and power-path behavior defined by the device.

Electronics that interrupt a battery path when selected unsafe conditions are detected.

Typical functions include overcharge, over-discharge, overcurrent, and short-circuit response, each with specified thresholds and limitations.

A voltage range interpreted as a digital low or high.

Compatibility requires sender output guarantees, receiver thresholds, supply/reference agreement, and absolute-maximum limits under load.

A boundary used to classify a continuous value into different states.

Threshold selection trades false positives and false negatives and may include hysteresis, calibration, or an undefined region.

An output that actively pulls low but needs an external path to become high.

Multiple compliant devices can share a line because none drives high; a pull-up and bus capacitance determine high level and rise time.

A resistor that gives a signal a default high level when nothing drives it low.

Its resistance trades low-state current against RC rise time, leakage tolerance, and noise susceptibility.

The total capacitance seen by a shared signal line.

Device inputs, traces, cables, and probes add capacitance; with a pull-up it produces a finite rise time that constrains data rate.

A shared electrical reference that lets connected circuits interpret signal voltages consistently.

Return-current paths have impedance and topology; common reference does not imply zero noise or permission to join isolated systems.

A machine that converts DC electrical input into rotational torque using commutation.

A simple model combines winding resistance and inductance, torque proportional to current, and back-EMF proportional to speed.

Four-switch circuit that can apply either voltage polarity to a load.

Switch states provide forward, reverse, coast, or brake behavior; dead time prevents same-leg shoot-through.

Motor current when the shaft is not turning at a stated voltage and condition.

With back-EMF near zero, current is limited mainly by winding, driver, wiring, and source resistance; heating is severe.

Voltage generated by a spinning motor that opposes the applied drive.

In a linear motor model e = k_eω; it reduces armature current as speed rises and vanishes near zero speed.

Rapidly switching an output so its on-time fraction controls an average effect.

Duty cycle, frequency, load dynamics, switching loss, and filtering determine current ripple and delivered behavior.

Slowing a motor by providing an electrical path that opposes its motion.

Bridge braking can short motor terminals through a low-resistance path so generated current dissipates energy; regenerative braking returns energy when supported.

A push or pull that changes motion or loads a structure.

Force is a vector and the net external force equals mass times acceleration in an inertial frame.

A sketch of one isolated body with every external force and moment acting on it.

Contacts are replaced by reaction components consistent with their constraints, enabling force and moment balance.

The turning effect of a force about a chosen point or axis.

Moment is the cross product r×F; only perpendicular lever arm contributes, and sign follows the stated convention.

A condition with zero net force and zero net moment.

Static equilibrium additionally has no motion; dynamic equilibrium can have constant velocity while accelerations remain zero.

The weighted-average location of an object's mass.

Translational motion responds as if net force acts on total mass at this point; gravity may be modeled there in a uniform field.

Rotational effort produced by a force acting through a lever arm.

Torque is r×F and rotational power is τω; available torque depends on speed, current, gearing, and thermal limits.

The ratio that trades rotational speed for torque between input and output.

For reduction G = ω_in/ω_out; ideal torque multiplies by G while real output includes efficiency and load-dependent losses.

How quickly an angle changes with time.

Angular speed is magnitude of angular velocity; rad/s connects naturally to tangential speed and rotational power.

The rate at which a force or torque does mechanical work.

Translational power is F·v and rotational power is τ·ω using aligned components and consistent instantaneous values.

Clearance that allows an input to reverse slightly before the output follows.

Gear tooth, screw, or joint clearance creates a dead zone and direction-dependent position error, especially during reversals.

Internal force distributed over a cross-sectional area.

Average axial stress is F/A; local stress varies near holes, notches, contacts, and bending.

Deformation measured relative to an original dimension.

Axial engineering strain is change in length divided by original length and is dimensionless.

A material measure of axial stiffness in its linear elastic range.

It is the slope of stress versus strain under specified conditions; geometry separately controls part stiffness.

The stress at which substantial permanent deformation begins by a specified criterion.

Yield values depend on material condition, direction, temperature, test method, and definition such as 0.2% offset.

A ratio reserving strength or capacity beyond a calculated demand.

It compares a defined failure limit with a compatible calculated stress or load and must reflect uncertainty and consequence.

Tension intentionally created in a fastener during tightening.

Preload compresses joint members and can resist separation or slip; tightening torque estimates it only indirectly through uncertain friction.

The compressive force holding mating joint surfaces together.

It arises from fastener preload and supports frictional shear transfer until separation or slip changes the load path.

A load acting perpendicular to a shaft's axis.

Bearing radial capacity and life depend on load magnitude, direction, speed, lubrication, fit, and combined axial load.

A load acting along a shaft's axis.

Axial reaction requires a thrust-capable bearing, shoulder, or other defined load path rather than ordinary radial support alone.

A machine element that supports constrained relative motion while carrying load.

Selection considers radial/axial load, speed, life, fit, alignment, lubrication, sealing, and installation path.

A sliding bearing surface that supports a shaft without rolling elements.

Clearance, material pair, pressure-velocity limits, lubrication, wear, and alignment govern performance.

The named target size used to identify or design a feature.

Actual permissible sizes are defined by tolerances around or relative to the nominal value.

The allowed range of variation in a dimension or property.

Tolerance specifies conformance limits chosen for function and manufacturing; it is distinct from measured uncertainty and actual process spread.

A fit where the hole remains larger than the shaft across allowed sizes.

Minimum clearance is smallest hole minus largest shaft; positive clearance permits assembly and motion but affects play.

A fit where the shaft is larger than the hole before assembly.

Interference creates contact pressure and retention but raises assembly force and material stress; temperature and material matter.

A defined reference feature from which geometry or measurements are established.

Datums create a repeatable coordinate system and constrain degrees of freedom for inspection, assembly, and CAD design.

A named value that configures a design or model while variables change within it.

Parameters may be specified, measured, or estimated; identifiability and operating range determine whether fitted values are meaningful.

The combined effect of multiple allowed dimensional variations on an assembly result.

Worst-case sums guarantee bounds under stated independence-free limits; statistical stacks make distribution assumptions and different risk claims.

The width of material removed by a cutting process.

Tool diameter, beam, material, speed, focus, and wear influence kerf; compensation offsets the toolpath relative to desired geometry.

The functional relationships a model is meant to preserve when dimensions change.

Constraints, datums, parameters, and feature order encode why geometry exists so edits regenerate predictably.

The study of motion without first solving the forces that cause it.

Kinematics relates position, orientation, velocity, and acceleration through mechanism geometry and time.

One independent coordinate needed to describe a system's configuration.

Constraints remove degrees of freedom; mechanism mobility depends on independent constraints, not just joint count.

The set of positions or poses a mechanism can reach.

Workspace depends on link geometry, joint limits, collisions, singularities, and whether orientation requirements are included.

A matrix relating small joint changes or rates to task-space changes or rates.

For robot kinematics, V = J(q)q̇; rank and conditioning reveal lost motion directions and force amplification.

A configuration where a mechanism loses an instantaneous motion direction or requires extreme joint rates.

The Jacobian loses rank; inverse kinematics and force transmission become ill-conditioned near the singularity.

A coordinate frame attached to a joint for describing its axis and connected links.

Consistent joint frames define transform order, positive rotation, and link geometry in a kinematic chain.

Friction that prevents relative sliding up to a limit.

Its magnitude adapts as needed up to approximately μ_sN; it is not always equal to that maximum.

Friction acting between surfaces that slide relative to each other.

A simple Coulomb model uses magnitude μ_kN opposite sliding, though real friction varies with material, speed, temperature, and lubrication.

The contact force component perpendicular to a surface.

It is determined by the full force and acceleration balance; it need not equal weight on slopes or accelerating systems.

Input power converted into an unintended form rather than useful output.

Loss depends on the chosen system boundary and may appear as heat, sound, vibration, slip, fluid drag, or electrical dissipation.

A quantity that carries information by varying with time, space, or another variable.

A signal has units, domain, bandwidth, sampling, reference, and uncertainty; its representation is not the physical phenomenon itself.

A bounded part of the world whose outputs respond to inputs and internal state.

A dynamic system maps input and initial state to evolving state and output under a model and operating assumptions.

The smallest remembered information needed to predict future model behavior given future inputs.

State variables summarize stored energy or memory and evolve through differential or difference equations.

How much output changes for a change in input under specified conditions.

Static gain is Δoutput/Δinput near an operating point; dynamic gain may vary with frequency and state.

A system's output after its input changes abruptly from one constant level to another.

Rise, overshoot, settling, delay, and final gain expose dynamics within the tested amplitude and operating condition.

Using measured output to adjust the command based on error.

Negative feedback can reject disturbances and reduce sensitivity but also creates dynamics, noise paths, and possible instability.

The desired value supplied to a controller.

The reference may be filtered or trajectory-generated so demanded rate and acceleration remain feasible.

The signed difference between desired and measured output.

With convention e = r−y, sign determines corrective action; sensor bias means computed error may differ from true error.

A feedback rule combining present, accumulated, and changing error.

u = Kp e + Ki∫e dt + Kd de/dt, implemented with sampling, filtering, limits, and anti-windup.

Excess accumulated controller integral while the actuator cannot deliver the requested output.

Saturation breaks the assumed loop, allowing integral state to grow and cause overshoot or slow recovery; anti-windup bounds or back-calculates it.

The property that bounded disturbances or initial errors do not make behavior diverge uncontrollably.

Definitions include asymptotic, Lyapunov, and bounded-input bounded-output stability; each requires a specified model and operating region.

A dimensionless measure of damping relative to critical damping in a second-order model.

ζ<1 is underdamped, ζ=1 critically damped, and ζ>1 overdamped for the canonical linear model.

How far a response passes beyond its final or target value.

Percent overshoot is usually peak excess divided by step magnitude under a specified step-response convention.

Time until a response enters and remains within a stated band around its final value.

The band, reference value, disturbance, and observation window must be declared, commonly 2% or 5% for model comparisons.

An output's periodic response occurring later in cycle than its input.

Phase is frequency-dependent; delay and stored-energy dynamics consume feedback phase margin and can promote oscillation.

A measure of how much modeled loop change remains before instability.

Gain and phase margins are frequency-domain properties of an open-loop model and do not cover every nonlinearity or uncertainty.

A filter replacing each sample with the average of a recent window.

It reduces uncorrelated high-frequency noise but adds delay and has a frequency response with sidelobes and zeros.

A filter that passes slow changes more strongly than rapid changes.

Its cutoff and phase response trade noise reduction against delay and attenuation of real dynamics.

The difference between an observed value and a model's prediction.

Residual structure versus time, input, or state reveals bias, missing dynamics, correlation, or changed conditions.

Unwanted variation that obscures the quantity of interest.

Noise is characterized by distribution, spectrum, correlation, and dependence on operating conditions; it is not synonymous with all error.

Estimating a model from controlled input-output data.

Model structure, excitation, parameter estimation, validation data, and operating region jointly determine useful predictive scope.

Testing whether a model predicts data not used to fit it well enough for its intended decision.

Validation inspects residuals and task metrics across declared conditions and compares against alternative models or baselines.

Input variation that reveals a system's response and parameters.

Persistent excitation covers the frequencies and amplitudes needed to distinguish model effects while remaining safe and within scope.

Using a model beyond the range of data that supported it.

Extrapolation depends strongly on model structure and can fail abruptly at saturation, thresholds, or changed regimes.

A time-indexed desired path including position and often velocity and acceleration.

A trajectory parameterizes motion so dynamic, actuator, collision, and timing constraints can be checked along the whole path.

The rate and direction of change of position.

Velocity is a vector derivative expressed in a frame; speed is its magnitude.

The rate at which velocity changes.

Acceleration is a vector and includes changes in direction even when speed stays constant.

A move that accelerates, cruises at bounded speed, then decelerates.

It becomes triangular when travel is too short to reach the speed limit; position is the integral of its velocity segments.

The difference between a commanded trajectory and actual motion.

Tracking error reflects plant dynamics, disturbances, sensing, delay, saturation, and controller design and is time-dependent.

A bound on position, speed, acceleration, jerk, torque, or another motion quantity.

Limits come from geometry, actuator capability, traction, payload, comfort, cables, and safety, and need enforcement plus fault behavior.

Combining complementary measurements into one estimate while retaining uncertainty and disagreement.

Fusion requires compatible quantities, frames, timestamps, calibration, noise models, and correlation assumptions.

A measure of squared spread around a mean.

Variance is expected squared deviation and has squared units; inverse variance weights independent unbiased estimates by modeled precision.

Comparing an instrument with references to characterize how readings relate to known values.

Calibration estimates offset, scale, nonlinearity, and uncertainty under stated conditions; adjustment is a separate action.

A named origin and axes in which coordinates are expressed.

Fusion requires transforms and covariance handling between frames plus explicit handedness and timestamp.

The difference between a new measurement and what the current estimate predicted that sensor would see.

Its covariance supports consistency checks and gating; persistent structure signals bias, timing, frame, or model problems.

Errors that tend to vary together rather than independently.

Shared environment, calibration, power, algorithms, or ancestry creates covariance; ignoring it makes fused uncertainty overconfident.

The intended time between successive discrete controller updates.

It shapes numerical approximation, delay, bandwidth, processor utilization, and gain interpretation; actual periods include jitter.

Turning a continuous model or controller into updates at separate times or values.

The chosen integration, hold, sampling, and quantization alter dynamics and can destabilize an otherwise stable continuous design.

A numerical update that projects the current slope across one time step.

x[k+1]=x[k]+Δt f(x[k],u[k]); accuracy and numerical stability depend on step size relative to dynamics.

Variation of actual event timing around its intended schedule.

Jitter changes sample intervals, latency, PWM edges, and numerical gains; measure timestamps rather than assuming configuration equals execution.

The latest time by which a computation or action must complete.

Real-time correctness includes value and timing; relative deadlines are compared using monotonic, rollover-safe time arithmetic.

Mapping a continuous range or fine values into discrete levels.

Quantization introduces bounded rounding error under ideal conditions and can create limit cycles, clipping, or structured error in feedback.

A behavior model with named states and explicit rules for changing between them.

Next state is a function of current state and inputs; outputs depend on state alone or state plus input.

A defined change from one state to another when a condition or event occurs.

A transition specifies source, guard/event, destination, outputs, and any entry or exit action.

A noteworthy occurrence that may trigger program behavior.

Events may be edges, messages, timeouts, or state changes and should be queued or handled with explicit loss and ordering policy.

Work that prevents a task from handling other obligations until it finishes.

Blocking I/O or delay increases response latency and can violate scheduling deadlines unless isolated or bounded.

A grouped value whose named fields describe one coherent item.

Records preserve field meaning and units; packed layouts trade readability and portability for predictable memory use.

A collection where items normally leave in the order they arrived.

A bounded queue requires explicit full, empty, concurrency, priority, and loss behavior.

Fixed storage whose read or write index wraps from the end back to the start.

Modulo indexing bounds memory; full-state representation determines whether new data overwrite old data or are rejected.

The maximum number of items a bounded structure can hold under its representation.

Capacity differs from current length and must include payload, alignment, metadata, and reserved memory constraints.

An attempt to add data beyond available bounded storage.

Valid policies include reject, back-pressure, drop-oldest, drop-newest, or escalation, selected from loss and ordering requirements.

A configurable microcontroller pin for observing or driving logic-level signals.

Mode, alternate function, thresholds, source/sink current, voltage limits, and boot state define safe use.

A voltage boundary used by a digital input to recognize low or high.

Datasheets specify guaranteed low and high ranges with an undefined region; thresholds can vary with supply and temperature.

Current flowing out of an output pin into a load.

Per-pin and aggregate source limits, output voltage droop, package dissipation, and absolute maxima constrain operation.

Current flowing from a load into an output pin driven low.

Sink capability has guaranteed voltage and aggregate limits that may differ from source capability.

A peripheral that counts clock events to measure time or schedule actions.

Clock source, prescaler, counter width, compare registers, and rollover determine period and resolution.

Time required for one complete repeating cycle.

Period is the reciprocal of frequency for a periodic signal, with timing variation described separately as jitter.

The number of repeating cycles per unit time.

Frequency is reciprocal period; usable PWM frequency also depends on timer clock, count range, and required duty resolution.

The fraction of each period that a periodic digital signal is active.

Duty D = t_on/T; average load behavior depends on amplitude, dynamics, switching topology, and filtering.

A circuit that represents an input voltage as a digital code.

Range, reference, resolution, sample-and-hold, source impedance, noise, linearity, and conversion timing govern the result.

The voltage scale against which an ADC interprets input.

Reference error and noise become gain error in converted results; input must remain within allowed range and source conditions.

The number of measurements taken per second.

Rate, aperture timing, analog bandwidth, and anti-alias filtering together determine which signal content can be represented.

Fast signal variation appearing as a different slower pattern after sampling.

Frequency components above half the sampling rate fold into the sampled band unless limited before conversion.

Flattening when a signal exceeds the representable or allowed range.

Samples pin at a minimum or maximum code, losing amplitude and waveform information beyond the limit.

A mechanism that temporarily redirects processor execution when an event needs service.

Interrupt priority, masking, latency, nesting, shared data, and service duration affect real-time behavior.

Elapsed time from an initiating event to the relevant response.

Latency includes sensing, queuing, computation, communication, scheduling, and actuation; distributions and worst credible values matter.

Code run in response to an interrupt.

An ISR should bound execution, acknowledge its source, avoid unsafe blocking, and coordinate shared state with appropriate atomicity.

Code that accesses shared state and must not be interleaved with conflicting access.

Protection may disable interrupts briefly, use locks, or use atomic operations; duration must remain bounded.

The fraction of available processor time demanded by work.

For periodic tasks, nominal utilization sums execution time divided by period, but schedulability also depends on deadlines, blocking, and priority.

A bug where the outcome depends on uncontrolled ordering of concurrent events.

Non-atomic shared-state access can lose updates or observe inconsistent combinations; synchronization must protect the invariant.

An agreed contract for exchanging messages.

It defines physical assumptions, framing, fields, units, ordering, timing, error detection, retries, and version behavior.

A bounded group of transmitted symbols interpreted as one communication unit.

Framing uses length, delimiters, fixed fields, timing, or encoding and must recover after corruption or byte loss.

The number of transmitted symbols per second.

Baud equals bit rate only when each symbol carries one bit; asynchronous UART peers also need matching framing and tolerable clock error.

A value computed from message data to detect likely corruption.

Coverage depends on algorithm and error model; checksums detect but generally do not correct errors or authenticate senders.

A declared maximum wait before treating an expected event as late or absent.

Timeout selection accounts for normal latency and jitter while bounding stale behavior; recovery and evidence must be explicit.

An operation whose repeated application has the same intended effect as applying it once.

Idempotency simplifies retries when acknowledgments can be lost; sequence numbers may still be needed for ordering and freshness.

The ability to obtain a consistent result using documented methods under changed people, equipment, or setting.

Reproducibility needs versions, conditions, raw evidence, uncertainty, and deviations; repeatability is the narrower same-condition case.

A check that a condition believed necessary is currently true.

Assertions expose violated programmer assumptions near their source and should record enough context without replacing external input validation.

A log whose named fields follow a consistent schema rather than only prose.

Timestamps, event names, units, IDs, state, and version permit filtering and correlation across subsystems.

Narrowing a symptom to the subsystem, interface, or cause that can explain it.

Change one condition from a known-good baseline and use discriminating measurements rather than replacing parts randomly.

Recorded evidence of why a processor restarted.

Hardware status may distinguish power-on, watchdog, brownout, software, or external reset, though flags and retention behavior are device-specific.

A defect that changes or disappears when attempts to observe it alter timing or state.

Logging, breakpoints, probes, optimization, and memory layout can perturb races, electrical loading, and deadlines.

The division of robot responsibilities into components and explicit interfaces.

Architecture assigns sensing, estimation, planning, control, supervision, timing, ownership, and fault containment across boundaries.

A defined boundary through which components exchange power, data, forces, or services.

An interface specifies connector or call, direction, units, timing, valid range, ownership, startup, and faults.

An independent mechanism that triggers a bounded response when expected activity stops.

A watchdog must be refreshed only by demonstrated healthy progress and needs a defined reset, safe-output, evidence, and recovery path.

How recently an actuator command was produced or validated.

Freshness uses monotonic timestamps, sequence numbers, and expiry policy to prevent stale but syntactically valid commands from persisting.

A component that monitors subsystem health and coordinates operating or fault modes.

It evaluates cross-system conditions and requests degraded or safe states while local components retain immediate safety limits.

An ordered set of rigid links connected by joints.

Successive joint transforms map configurations to end-effector pose; open and closed chains have different constraint structure.

Computing an end pose from known joint values and link geometry.

Ordered homogeneous transforms compose from base to tool; the result is generally unique for a fully specified chain configuration.

Finding joint values that produce a desired end pose.

IK can have zero, one, many, or continuous solutions subject to joint limits, singularities, and numerical convergence.

An object's location and orientation relative to a named frame.

A rigid-body pose is an element of SE(3), commonly represented by rotation and translation with conventions stated.

A matrix combining rotation and translation between coordinate frames.

A 4×4 SE(3) transform composes frame changes by matrix multiplication; order and direction are essential.

Estimating where a robot is within a map or reference frame.

Localization combines motion and observation models with uncertainty; observability, data association, and map quality bound performance.

Estimating motion by accumulating incremental measurements from the robot.

Wheel or inertial increments integrate into pose and therefore accumulate scale, bias, slip, timing, and geometry errors.

A sensor producing counts or codes related to mechanical position or motion.

Resolution, edge decoding, index, missed counts, mounting, direction, and calibration determine usable information.

A mobile base steered by commanding left and right wheel speeds independently.

Center speed is the mean wheel speed and yaw rate is their difference divided by track width under no-slip planar assumptions.

Growing pose error caused by accumulating imperfect motion estimates.

Heading error couples into position and often grows faster than simple distance-scale error without absolute corrections.

The effective lateral distance between left and right wheel contact paths.

Differential-drive yaw estimation uses effective track width, which may differ from a ruler measurement because of compliance and scrub.

Turning sensor data into task-relevant descriptions of the robot and its surroundings.

Perception includes calibration, preprocessing, feature extraction, inference, uncertainty, and evaluation against labeled conditions.

A selected measurable pattern used by an estimator or classifier.

Features may be engineered or learned and must be robust to nuisance variation while retaining task information.

The fraction of actual positive cases that a detector finds.

Recall = TP/(TP+FN) and must be reported with class definition, threshold, and test distribution.

A table comparing predicted classes with actual reference classes.

Counts of true/false positives and negatives support class-specific metrics and expose imbalance hidden by accuracy.

Finding a route from start to goal that respects a model of obstacles and costs.

Planning searches configuration or state space; feasibility, optimality, dynamics, uncertainty, and replanning depend on the chosen abstraction.

A spatial representation assigning traversal penalty or prohibition to regions.

Costs encode obstacles, clearance, terrain, uncertainty, or preferences and strongly shape the planned route.

An estimate of remaining path cost used to guide a search.

An admissible heuristic never overestimates true remaining cost; consistency supports efficient A* graph search.

A graph-search algorithm prioritizing traveled cost plus estimated remaining cost.

It expands nodes by f(n)=g(n)+h(n); completeness and optimality depend on graph and heuristic conditions.

Expanding map obstacles to account for robot size and desired clearance.

For a rigid footprint, inflation approximates a configuration-space obstacle via the footprint shape and chosen safety margin.

Computing a new route when the map, goal, state, or execution evidence changes.

Triggers balance responsiveness against computation and route thrashing; the robot needs a safe behavior while a new plan is unavailable.

A testable agreement describing what connected components provide and require.

It covers electrical, data, timing, units, frames, mechanical datums, startup, versioning, and fault response.

An allocation of allowable delay across stages of an end-to-end response.

Sensing, transport, queues, computation, scheduling, and actuation contributions combine, including jitter and worst credible cases.

A periodic indication that a component is still communicating or making specified progress.

Heartbeat period, sequence, payload, timeout, and independence determine which failures it can reveal.

A test with predefined pass criteria deciding whether a requirement is met for a stated configuration.

The procedure records setup, revision, instruments, conditions, actual results, deviations, and evidence linked to the requirement.

Probability that a system performs its required function for a stated time under stated conditions.

Reliability needs a defined success event, population or model, duration, environment, and dependence assumptions.

A defect or abnormal condition that can lead to an incorrect internal state.

Faults may be transient, intermittent, permanent, design, manufacturing, environmental, or operational and require defined detection scope.

An incorrect internal system condition caused by a fault.

An error may remain latent, be detected and contained, or propagate across interfaces until service fails.

A system's delivered service departing from its required behavior.

Failure is observed at a defined system boundary and can result from one or more propagated faults and errors.

One shared cause defeating multiple supposedly redundant channels.

Shared power, environment, connector, calibration, software, or design assumptions invalidate naive independence calculations.

A deliberate reduced-capability operating state used after a fault or loss of confidence.

It specifies retained functions, tighter limits, annunciation, evidence, entry conditions, and recovery or shutdown criteria.

A short durable account of a choice, alternatives, evidence, and consequences.

It captures context, status, assumptions, selected option, rejected alternatives, revision, and links to calculations or tests.

A documented environmental, hardware, software, load, and procedural circumstance under which evidence was collected.

Conditions bound applicability and include versions, instruments, calibration, timing, configuration, and deviations.

The interval from the smallest to largest value actually observed in a dataset.

Range summarizes extremes in the sample but is sensitive to sample count and is not a guaranteed population bound.

A record of where an artifact or claim came from and how it changed.

Provenance links source identity, version, transformations, authorship, timestamps, licenses, and raw evidence to derived outputs.