Robotics
Kinematic calibration
Definition
Kinematic calibration estimates geometric robot-model parameters, such as link dimensions, joint zero offsets, or axis alignment, from measurements of the robot's motion or pose. The corrected parameters reduce systematic errors in forward kinematics and quantities derived from it.
Also known as: Robot kinematic calibration
Updated
Estimate geometry from measured discrepancies
A nominal robot model may not match the assembled machine because of manufacturing tolerance, joint-zero error, wear, deformation, or an approximate axis location. Kinematic calibration chooses model parameters so predicted poses or velocities better match independent measurements.
The procedure needs an error model, informative robot configurations, reference measurements, and an estimator or optimiser. Parameters can be identified offline from a calibration routine or online while the robot operates. The corrected values then feed forward kinematics, Jacobians, odometry, planning, or control.
Leg geometry affects humanoid odometry
Legged robots often infer body velocity from joint encoders and feet assumed to be stationary in contact. An incorrect link length changes the kinematic velocity prediction and can create integrated pose drift.
Yang and colleagues compare leg-odometry velocity with an independent body-velocity measurement and update selected kinematic parameters inside a filter or sliding-window estimator. They analyse observability and report simulation and hardware results on a Unitree A1. Their results support the method and setup studied; they do not imply that every leg dimension is observable during arbitrary motion.
Calibration targets must be distinguished
Kinematic calibration estimates robot geometry. It differs from hand-eye calibration, which estimates a transform between sensor and robot frames, and from dynamic system identification, which can target mass, inertia, friction, or actuator parameters. A combined calibration may include several of these, but their parameters and evidence should remain identifiable.
Parameter correlations can make different errors explain the same measurements. Contact slip, flexible deformation, timing offsets, or a biased external reference can then be absorbed into the wrong geometric parameter. Validation should use poses or motions not used for fitting. Calibration compensates a model; it does not repair loose hardware or guarantee accuracy outside the excited configurations, loads, and temperatures.
Sources
Related terms
Forward kinematics
Forward kinematics calculates the position and orientation of a robot link or end-effector from the robot geometry and joint positions. It maps a robot configuration to a pose.
System identification
System identification estimates a model of a physical system from measured inputs and outputs. In robotics, it can recover parameters such as inertia and friction or learn a more general model of how actions change the system state.
Hand-eye calibration
Hand-eye calibration estimates the fixed rigid-body transformation between a robot-mounted sensor and the robot hand or end effector. The result lets sensor measurements and commanded robot motion be expressed in consistent coordinate frames.
Robot Jacobian
A robot Jacobian is a configuration-dependent matrix that maps joint velocities to a chosen task velocity, often an end-effector twist. It describes the local relationship between joint motion and task motion.