The read-only inverse property of the XRRigidTransform interface returns another XRRigidTransform object which is the inverse of its owning transform. That is, you can always get the inverse of any XRRigidTransform using its inverse property, instead of having to explicitly generate it.

An XRRigidTransform which contains the inverse of the XRRigidTransform on which it's accessed.

Applying the inverse of a transform to any object previously transformed by the parent XRRigidTransform always undoes the transformation, resulting in the object returning to its previous pose. In other words, its position and orientation both return to their prior configurations.

In this example, the model view matrix for an object is computed by taking the view matrix and multiplying it by the object's pose matrix.

const modelViewMatrix = mat4.create();

for (const view of pose.view) {
  const viewport = glLayer.getViewport(view);
  gl.viewport(viewport.x, viewport.y, viewport.width, viewport.height);

  // …

  mat4.multiply(modelViewMatrix, view.transform.inverse.matrix, objectMatrix);
  gl.uniformMatrix4fv(
    programInfo.uniformLocations.modelViewMatrix,
    false,
    modelViewMatrix,
  );

  // …
}

This outline of a renderer's core code shows how the pose's view gets represented by taking its transform's inverse's matrix as the model view matrix used to transform objects based on the viewer's position and orientation. The inverse's matrix is multiplied by the object's matrix to get the model view matrix, which is then passed into the shader program by setting a uniform to contain that information.