This book "Articulated Motion and Deformable Objects: 7th International Conference" (1) describes how a unique (non-zero) solution to this homogeneous system of linear equations arises $ MR = [… -I … R_{ij} …][… R_i … R_j]^ T = 0 $ for relative rotation decoding (where $ j = $ 1.2.3). This equation is derived from the relationships between relative rotations and absolute rotations $ R_ {i, j} = R_iR_j $ reformulated to $ R_ {i, j} R_i – R_j = 0 $,

Thus, it can be solved by anchoring one or more rotation matrices in the least square sense $ R_i $ Guarantee uniqueness: $ M ^ TM + lambda I_k = M ^ TR_i + lambda ^ 2 cdot R_k $However, I do not understand where the parameter is $ lambda $ comes out and how to get it and what is meant by "solution in the least square sense".

Any help is greatly appreciated!