JournalResearchFieldsDatabaseWorkspace
← Fields

Robotics & Embodied Intelligence

A robotics record remains physical end to end: sensor location and frame determine an estimate, the estimate changes a control decision, an actuator changes the body or contact state, and a new measurement closes the loop.

body / contactsense → plan → act → correct

Body

Physical arrangement

  • link
  • joint
  • foot
  • arm
  • gripper

State

Measured condition

  • pose
  • contact
  • force
  • material
  • occupancy

Control

Decision and actuation

  • estimate
  • planner
  • controller
  • command

Coordination

System-scale operation

  • task
  • fleet
  • global controller
  • evidence
ObjectStructureRecordForm
Robot morphologyBody, links, joints, legs, arms, end effectorConnection, degree of freedom, placement and variantMechanical assembly
Mechanics and contactContact state, force components, pose and rotation estimatesReference frame, sensor source, state before and afterContact / force plate
Control architectureProcessor, controller, storage, sensor and actuatorSignal path, component responsibility, feedback interfaceControl block diagram
Motion processCommand, estimate, plan, force allocation and actuationOrdered blocks, update condition and controlled quantityMotion flow
PerceptionPoint cloud, classifier, model, material or map objectData acquisition, feature or primitive, classification and usePerception pipeline
Multi-robot coordinationMobile members, fixed members, global controller and task objectsMembership, shared resource, assignment and conflictFleet topology
Simulation and trainingEnvironment, labeled and unlabeled data, model and scoreDomain conditions, transfer step, update and validationSimulation / training loop
Motion evidenceRequested, measured and target trajectories over timeAxis, threshold, event time and comparison conditionEvidence curve
Sensor systemJoint sensors, IMU, force, camera or ranging deviceMounting location, frame, sample and derived estimateSensor placement map
ManipulationBase, arm, end effector, target object and force feedbackApproach, engagement, confirmation and releaseManipulation sequence

body → joint → end effector

Mechanical record

Connection and movement relationship among physical components.

sensor → estimate → command → feedback

Closed-loop record

The measured value, control decision, physical action, and resulting update.

state + trigger + transition

State record

Standing, walking, manipulating, degraded, or other named operating states.

requested vs measured

Performance record

Comparable curves, thresholds, event markers, and test conditions.

Measure A

Joint position and link geometry produce a kinematic motion estimate in a stated frame.

Measure B

An IMU or other independent sensor produces a second body-motion estimate.

Compare

The two estimates are aligned and their difference is tested against a threshold.

Reallocate

A detected discrepancy changes contact state, force priority, or force allocation.

Close

Actuator commands alter the body; subsequent force and motion measurements show the response.

The path ties a computational trigger to contact mechanics and an observable change in the robot.

Command ≠ physical actionPose estimate depends on reference framePerception object needs a producer and consumerSimulation evidence ≠ hardware evidence

What makes a control diagram a physical closed loop?

The path must reach named sensors and actuators. A controller output is followed through joint, gripper, wheel, or contact action and then back to a measured state.

  • sensor mounting
  • controlled quantity
  • actuator interface
  • feedback sample

How is slip or contact failure distinguished from ordinary motion?

Two estimates can be formed from independent sources—for example joint kinematics and an IMU. Their difference is compared with a threshold that triggers a contact-state or force-allocation update.

  • two estimate sources
  • shared coordinate frame
  • difference measure
  • triggered control update

What does a force or motion plot need to show?

Requested, measured, and target quantities share an axis and time base; event markers identify contact, threshold crossing, or controller change. Test conditions identify the robot and motion state.

  • physical quantity and units
  • event time
  • threshold
  • requested / measured / target traces

The relevant mechanism is the complete physical loop: mounted sensing, framed estimation, a defined discrepancy, changed actuation, and subsequent measurement. A controller output alone ends too early.

  1. The same numeric estimate has different meaning in different coordinate frames or mounting locations.

  2. Independent estimates make a slip or contact inference distinguishable from ordinary motion.

  3. Requested, measured, and target traces connect the threshold event to the controller transition and physical response.

  • Sensor locations, axes, and sampling relations are known.
  • The compared estimates are expressed in a common frame and time base.
  • Actuator commands and post-command measurements can be associated with the same event.