What is it?
This wiki contains a summary of the article accepted as a tutorial for IEEE Robotics and Automation Magazine on the 4th June 2012.
Rigid bodies are essential primitives in the modelling of robotic devices, tasks and perception, starting with the basic geometric relations such as relative position, orientation, pose, translational velocity, rotational velocity, and twist. This wiki elaborates on the background and the software for the semantics underlying rigid body relationships. This wiki is based on the research of the KU Leuven robotics group, in this case mainly conducted by Tinne De Laet, to explain semantics of all coordinate-invariant properties and operations, and, more importantly, to document all the choices that are made in coordinate representations of these geometric relations. This resulted in a set of concrete suggestions for standardizing terminology and notation, and software with a fully unambiguous software interface, including automatic checks for semantic correctness of all geometric operations on rigid-body coordinate representations.
The geometric relations semantics software prevents commonly made errors in geometric rigid-body relations calculations like:
- Logic errors in geometric relation calculations: A lot of logic errors can occur during geometric relation calculations. For instance (there is no need to understand the details just have a look at the difference in syntax), the inverse of is , while the inverse of the translational velocity is . When using the semantic representation proposed in this paper, the semantics of the inverse geometric relation can be automatically derived from the forward geometric relation, preventing logic errors. A second example emerges when composing the relations involving three rigid bodies: in order to get the geometric relation of with respect to body one can compose the geometric relation between and third body with the geometric relation between body and the body (and not the geometric relation between the body and the body for instance). Such a logic constraint can be checked easily by including, for instance, the body and reference body in the semantic representation of the geometric relations.
- Composition of twists with different velocity reference point: Composing twists requires a common velocity reference point (i.e. the twists have to express the translational velocity of the same point on the body). By including the velocity reference point of the twist in the semantic representation, this constraint can be checked explicitly.
- Composition of geometric relations expressed in different coordinate frames: Composing geometric relations using coordinate representations like position vectors, translational and rotational velocity vectors, and 6D vector twists, requires that the coordinates are expressed in the same coordinate frame. By including the coordinate frame in the coordinate semantic representation of the geometric relations, this constraint can be checked explicitly.
- Composition of poses and orientation coordinate representations in wrong order: The rotation matrix and homogeneous transformation matrix coordinate representations can be composed using simple multiplication. Since matrix multiplication is however not commutative, a common error is to use a wrong multiplication order in the composition. The correct multiplication order can however be directly derived when including the bodies, frames, and points in the coordinate semantic representation of the geometric relations.
- Integration of twists when velocity reference point and coordinate frame do not belong to same frame: A twist can only be integrated when it expresses the translational velocity of the origin of the coordinate frame the twist is expressed in. When including the velocity reference point and the coordinate frame in the coordinate semantic representation of the twist, this constraint can be explicitly checked.