Chapter 9, page 240, following eq. 9.12, the text from above the equation appears a second time.
After understanding the Jacobian derivation, the rest is the same as ordinary
graph optimization. In short, all pose vertices and pose edges constitute a graph
optimization, which is essentially a least-squares problem. The optimization variable
is the pose of each vertex, and the edges come from the pose observation constraints.
Let E be the set of all edges, then the overall objective function is
Chapter 9, page 240, following eq. 9.12, the text from above the equation appears a second time.