Geometry

package com.arcrobotics.ftclib.geometry

FTCLib provides access to geometry classes taken from WPILib. Since we like copy-pasting straight from WPILib instead of linking to the original material, that's what we're gonna do.

Translation

Translation in 2 dimensions is represented by FTCLib'sTranslation2d class. This class has an x and y component, representing the point (x,y)(x,y) or the vector [xy]\begin{bmatrix} x\\ y \end{bmatrix} on a 2-dimensional coordinate system.

You can get the distance to another Translation2d object by using the getDistance(Translation2d other), which returns the distance to another Translation2d by using the Pythagorean theorem.

Rotation

Rotation in 2 dimensions is represented by FTCLib’s Rotation2d class. This class has an angle component, which represents the robot’s rotation relative to an axis on a 2-dimensional coordinate system. Positive rotations are counterclockwise.

Pose

Pose is a combination of both translation and rotation and is represented by the Pose2d class. It can be used to describe the pose of your robot in the field coordinate system, or the pose of objects, such as vision targets, relative to your robot in the robot coordinate system. Pose2d can also represent the vector [xyθ]\begin{bmatrix} x\\ y\\ \theta \end{bmatrix} .

Vector

A vector in 2 dimensions is represented by the Vector2d class. It holds an xx and a yy value similarly to a Translation2d. These components representing the point (x,y)(x,y) or as the matrix[xy]\begin{bmatrix} x\\ y \end{bmatrix}.

Unlike a Translation2d, there are a few different methods and features.

Transform and Twist

FTCLib provides 2 classes, Transform2d, which represents a transformation to a pose, and Twist2d which represents a movement along an arc. Transform2d and Twist2d all have xx , yy and θ\theta components.

Transform2d represents a relative transformation. It has an translation and a rotation component. Transforming a Pose2d by a Transform2d rotates the translation component of the transform by the rotation of the pose, and then adds the rotated translation component and the rotation component to the pose. In other words, Pose2d.plus(Transform2d) returns [xpypθp]+[cosθpsinθp0sinθpcosθp0001][xtytθt]\begin{bmatrix} x_{p} \\ y_{p} \\ \theta_{p} \end{bmatrix} + \begin{bmatrix} \cos{\theta_{p}} & -\sin{\theta_{p}} & 0 \\ \sin{\theta_{p}} & \cos{\theta_{p}} & 0 \\ 0 & 0 & 1 \end{bmatrix} \begin{bmatrix} x_{t} \\ y_{t} \\ \theta_{t} \end{bmatrix} .

Twist2d represents a change in distance along an arc. Usually, this class is used to represent the movement of a drivetrain, where the x component is the forward distance driven, the y component is the distance driven to the side (left positive), and the θ\theta component is the change in heading.

Both classes can be used to estimate robot location. Twist2d is used in some of the FTCLib odometry classes to update the robot’s pose based on movement, while Transform2d can be used to estimate the robot’s global position from vision data.

Last updated