Models & learning
Pose estimation
Pose estimation is the process of inferring the position and orientation of a body, object, camera, hand, or robot relative to a specified coordinate frame. In three-dimensional robotics this is often called 6D or 6-DoF pose estimation because the result has three translational and three rotational degrees of freedom, even when orientation is stored with more than three numbers.
Also known as: 6D pose estimation, 6-DoF pose estimation
Updated
A pose combines position and orientation
ISO 8373 defines pose as the combination of position and orientation in space. The estimate is always relative to a frame: an object pose in the camera frame and the same object pose in the robot-base frame are different numerical values connected by a transform.
Orientation may be represented by Euler angles, a rotation matrix, quaternion or another parameterisation. Calling a result “6D” describes the body's six degrees of freedom, not necessarily the length of the stored vector.
The target determines the estimation problem
Robot-pose estimation may infer a mobile base or floating humanoid. Hand and body pose estimation infer articulated joints. Object-pose estimation often estimates one rigid transform from RGB, depth or point clouds. Category-level methods may also handle unseen instances or object scale.
The BOP benchmark formalises rigid-object pose estimation from RGB-D and accounts for ambiguous poses caused by object symmetries. An apparently large rotation error can describe an equivalent physical pose for a symmetric bottle or tool.
Estimated pose is not ground truth
Occlusion, motion blur, calibration error and missing depth create uncertainty. A single best transform can hide several plausible hypotheses. Downstream manipulation should not treat every estimate as exact.
Datasets should state the target, parent frame, representation, units, timestamp, estimator or measurement system, calibration, confidence and symmetry convention. Motion-capture or simulator poses should be distinguished from model predictions, and any smoothing or interpolation should remain part of the provenance.
Sources
Related terms
Hardware & control
Coordinate frame
A coordinate frame is a defined origin and set of oriented axes used to express positions, orientations, motions, forces, or other spatial quantities. A value has no complete geometric meaning until its frame and convention are known. Transformations relate measurements expressed in frames such as world, robot base, camera, end effector, object, or sensor.
Data & collection
Depth data
Depth data records the distance associated with image locations or sensor rays, usually as a depth image in which each pixel stores a metric value relative to a camera. The exact geometry, units, invalid-value convention, and coordinate frame depend on the sensor and encoding. RGB-D data pairs depth with colour imagery; a point cloud is a separate 3D representation derived from or aligned with such measurements.
Hardware & control
Robot kinematics
Robot kinematics describes the geometric relationship between a robot’s joint configuration and the position, orientation and velocity of its links or end-effector, without modelling the forces that cause the motion. Forward kinematics computes pose from joint values; inverse kinematics searches for joint values that achieve a requested pose.
Data & collection
Robot state
Robot state is the set of variables used to describe a robot at a particular time, such as joint positions and velocities, base pose, end-effector pose, gripper state, actuator measurements, or estimated motion. In control theory, a complete state contains enough information to predict future evolution given an action; in robot datasets, “state” often means only the measured or estimated subset that was logged.
Data & collection
Motion capture
Motion capture is the measurement and reconstruction of a person’s or object’s movement over time, commonly as joint positions, orientations or a fitted body model. Optical markers, cameras and inertial sensors can supply the measurements. Humanoid robotics uses the resulting motion sequences for analysis, imitation and retargeting to a robot body.