Additive results of the dual Moore--Penrose generalized inverse with an application to rigid-body motion estimation

arXiv:2610.16852 · math.NA, cs.NA · Submitted 2026-10-12 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.

Ada: We're looking at the paper: "Additive results of the dual Moore--Penrose generalized inverse with an application to rigid-body motion estimation".

Emmy: Here's the gist of the paper. Additive results of the dual Moore–Penrose generalized inverse with an application to rigid-body motion estimation is a paper that derives explicit formulas for the dual Moore–Penrose generalized inverse (DMPGI) of dual complex matrix sums and applies these formulas to rigid-body motion estimation with perturbed observation points, recovering observable motion components from velocity data Derivation of DMPGI Formulas The authors first express the DMPGI of a matrix partitioned into column blocks in terms of these blocks and associated projectors <ref:2610.16852#pg5>.

Ada: They then derive two formulas for matrix sums, applying them to rigid-body motion estimation with perturbed observation points that preserves the rank-deficient observation geometry <ref:2610.16852#pg6>.

Emmy: The paper develops a representation for

Ub Vb: and yields a formula for (UbUb∗ + VbVb∗)† <ref:2610.16852#pg4>.

Ada: Under the condition UbVb∗ = O, they also obtain a formula for (Ub + Vb)† <ref:2610.16852#pg11>.

Title and authors: Ada: Let's talk about the title and authors of this paper, "Additive results of the dual Moore--Penrose generalized inverse with an application to rigid-body motion estimation." The authors are Yue Zhaoa and Daochang Zhangb.

Emmy: I look at that title and think it's pointing toward a specific mathematical structure being added to the standard dual Moore--Penrose generalized inverse.

Ada: Right, it suggests they aren't just using an existing tool; they are deriving new results for matrix sums under certain conditions, which is what the authors claim to do.

Kolya: So, what does that mean practically? Is this just theoretical work on dual complex matrices, or is there something tangible we can use for motion tracking?

Emmy: It's applied directly to rigid-body motion estimation with two observation points that have a first-order perturbation in their positions. This isn't just abstract math; it's solving a real problem in robotics where the geometry isn't perfect.

Ada: The paper shows that this specific dual inverse recovers the observable motion components from the velocity data, which is important because sometimes you can only measure certain parts of motion accurately.

Kolya: So, if we're tracking something moving in three dee space with noisy measurements, this method helps us isolate the part of the motion we *can* see <ref:2610.16852#pg2>?

Emmy: That's the structure they are describing. They build a representation for Ub Vb, and then that leads to that formula for (Ub Ub* plus Vb Vb*)†, which is key for solving those linear dual equations.

Ada: And under a specific condition where Ub Vb* is zero, they get an even simpler formula for (Ub plus Vb)†. That's a neat simplification if that condition holds true.

Kolya: I wonder about that condition; what does it imply about the relationship between those two parts of the observation, U and V?

The paper's summary: Ada: So, let's look at what the paper actually summarizes. They start by expressing the dual Moore--Penrose generalized inverse of a matrix partitioned into column blocks in terms of these blocks and their associated projectors.

Emmy: That initial step is setting up the representation for the DMPGI based on those blocks, which is how they establish their foundation for solving dual complex matrix equations with partitions.

Ada: Then they derive two formulas specifically for matrix sums, and they apply these formulas to rigid-body motion estimation where the observation points have that first-order perturbation we talked about.

Kolya: So, the summary is essentially: we have a formula for a specific type of dual matrix sum, and when you plug it into tracking motion with slightly off observation points, you get an explicit way to recover what's observable.

Emmy: That’s right; they develop this representation for Ub Vb and then derive that formula for (Ub Ub* plus Vb Vb*)†, which is the main result they present.

Ada: And they also find a formula for (Ub + Vb)† if the condition Ub Vb* equals zero, which is another useful result under specific circumstances.

Kolya: So, what does this all mean for us on the ground? Does it make tracking easier or just more mathematically rigorous?

Emmy: It makes it rigorous because it shows exactly how to handle those rank-deficient observation geometries while still extracting usable information from the velocity measurements.

The paper's improvements: Ada: The paper suggests some improvements, which is where they show how these new formulas work in practice. They apply these derived formulas to rigid-body motion estimation from velocity measurements at two observation points with perturbed positions.

Emmy: In Example four point one, they compute that (Ub Ub* plus Vb Vb*)† using Theorem three point two, and the matrix inside the square brackets ends up being a specific expression involving Cb and Kb <ref:2610.16852#pg2>.

Ada: That expression turns out to be I - Ub Vb(I - Cb C b) K b Vb* (Ub) * = one/three - four/nine epsilon one which is a concrete numerical result under certain assumptions.

Kolya: That specific result, one/three minus four/nine epsilon one tells us something about the error or the structure of that quantity when you introduce that first-order perturbation term epsilon zero. What does the epsilon one represent <ref:2610.16852#pg1>?

Emmy: That term comes from the structure of how they defined their dual complex numbers, where DC equals a plus epsilon a zero and it relates to the real part of those matrices <ref:2610.16852#pg1>.

Ada: And they then analyze what happens when you look at the first component, A b one 000056T, and find that it is zero. This means the first component cannot be determined from the measurements at all.

Kolya: So, if that first component is zero, does that mean we can't determine *any* motion related to the x-axis rotation? Is it a total loss of information there?

Conclusion: Ada: To wrap up this paper on "Additive results of the dual Moore--Penrose generalized inverse with an application to rigid-body motion estimation," we see that these explicit formulas are very useful for recovering observable motion components from velocity data.

Emmy: The main implication is that this method confirms the use of dual matrix methods in kinematics and robotics, especially when dealing with rank deficiency caused by observation perturbations.

Ada: Specifically, it shows how to explicitly model and identify which rotational components become unobservable because of the rank deficiency in the observation geometry.

Kolya: From a probabilistic standpoint, this means we have a mathematical way to define the bounds on what we can actually estimate from noisy velocity data in these specific setups.

Ada: And they confirm that certain rotational components, like rotation about the x-axis, are unobservable due to those constraints.

Emmy: And they also show how the structure of the DMPGI for a product satisfies a reverse order law, where (Ub Vb)† equals Vb†Ub†.

Kolya: That reverse order law is interesting because it means we can simplify inversions when we have known product structures in our motion tracking algorithms.

Ada: So, to summarize the "Additive results of the dual Moore--Penrose generalized inverse with an application to rigid-body motion estimation," they provide explicit formulas for the DMPGI of dual complex matrix sums and show how these work on rigid-body motion estimation with perturbed observation points.

Emmy: It's a solid contribution showing how dual matrix methods help us tackle kinematic problems where standard inverses fall short.

Kolya: And that's the picture of what this paper is about; using these dual tools to pull out measurable physical information from tricky sensor data.

Department of Math & Stat, Georgia State University · College of Mathematics, Changchun Normal University

math.NA, cs.NA

Submitted: 2026-10-12

Updated: 2026-10-12

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 56/100

The gist: Additive results of the dual Moore–Penrose generalized inverse with an application to rigid-body motion estimation is a paper that derives explicit formulas for the dual Moore–Penrose generalized

Key concepts

Dual Complex Numbers (DC): These are numbers of the form a + εa0, where ε is a complex number such that ε² = 0. They extend standard complex numbers by introducing a nilpotent element, which is crucial for handling certain linear algebra problems in kinematics and robotics.
Dual Moore–Penrose Generalized Inverse (DMPGI): The DMPGI is a generalized inverse specifically defined for dual complex matrices. It is derived using four Penrose equations and provides a robust way to invert these matrices, which are essential for solving systems arising in rigid-body motion estimation.
Rigid-Body Motion Estimation: This involves determining the motion (position and orientation) of an object in 3D space based on measurements, such as velocity data from two observation points. The paper uses the DMPGI to extract specific, observable parts of this motion despite measurement perturbations.
Rank Deficiency in Observation Geometry: This occurs when the way observations are set up (the geometry) does not provide enough independent information to uniquely determine all possible motion parameters. The study shows how this deficiency leads to certain rotational components being unobservable.

Terminology

Accuracy — 3× in this paper · explained in 1 episode
Calibration — 2× in this paper · explained in 9 episodes
Parameterization — 2× in this paper · explained in 2 episodes
Sufficiency — 1× in this paper · explained in 2 episodes

Summary

Additive results of the dual Moore–Penrose generalized inverse with an application to rigid-body motion estimation is a paper that derives explicit formulas for the dual Moore–Penrose generalized inverse (DMPGI) of dual complex matrix sums and applies these formulas to rigid-body motion estimation with perturbed observation points, recovering observable motion components from velocity data

Derivation of DMPGI Formulas

The authors first express the DMPGI of a matrix partitioned into column blocks in terms of these blocks and associated projectors <ref:2610.16852#pg5>. They then derive two formulas for matrix sums, applying them to rigid-body motion estimation with perturbed observation points that preserves the rank-deficient observation geometry <ref:2610.16852#pg6>. The paper develops a representation for [Ub Vb] and yields a formula for (UbUb∗ + VbVb∗)† <ref:2610.16852#pg4>. Under the condition UbVb∗ = O, they also obtain a formula for (Ub + Vb)† <ref:2610.16852#pg11>.

Properties of Dual Complex Matrices

The preliminaries section introduces notation and establishes basic properties of dual complex matrices and the DMPGI <ref:2610.16852#pg3>. They define dual complex numbers as DC = a + εa0, where ε 2 = 0, and recall the definition of the DMPGI via four Penrose equations <ref:2610.16852#pg3>. The existence condition for the DMPGI is given by (Im − GG†)G0(In − G†G) = O <ref:2610.16852#pg3>. Lemma 2.6 shows that the DMPGI of Xb admits representations in terms of the DMPGIs of XbXb∗ and Xb∗Xb <ref:2610.16852#pg6>.

Application to Rigid-Body Motion Estimation

The paper applies the derived formulas to rigid-body motion estimation from velocity measurements at two observation points with perturbed positions and explains why the angular velocity component along the line joining these points remains undetermined <ref:2610.16852#pg6>. In Example 4.1, they compute (UbUb∗ + VbVb∗)† using Theorem 3.2 <ref:2610.16852#pg13>. The resulting matrix inside the square brackets in Theorem 3.2 is therefore I − Ub†Vb(I − Cb†Cb)KbVb∗(Ub†)∗ = 1/3 − 4/9ε001 <ref:2610.16852#pg14>.

Unobservable Motion Component

The analysis of the rigid-body motion estimation shows that the first component cannot be determined from the measurements, since Ab1 000056T = O <ref:2610.16852#pg18>. This implies that every motion of the form 1 + α23456T, α ∈ R, produces the same measurements <ref:2610.16852#pg19>. Rotation about the x-axis is therefore unobservable in this configuration. The zero first component of bξrec does not imply that the actual angular velocity about the x-axis vanishes.

Summary of Key Results

The paper provides explicit formulas for the DMPGI of dual complex matrix sums and demonstrates their utility in recovering observable motion components from velocity data <ref:2610.16852#pg2>. The derived representation for (UbUb∗ + VbVb∗)† is given by a complex expression involving matrices Cb and Kb <ref:2610.16852#pg10>. Under the condition UbVb∗ = O, the DMPGI of (Ub + Vb)† is also explicitly formulated <ref:2610.16852#pg11>. The application to rigid-body motion estimation reveals that certain rotational components are unobservable due to rank deficiency in the observation geometry. This work confirms the use of dual matrix methods in kinematics and robotics [3, 5, 8]. The paper concludes by showing how the DMPGI of a product satisfies a reverse order law (UbVb)† = Vb†Ub†.

References

[1] J. Angeles, Is there a characteristic length of a rigid-body displacement? Mech. Mach. Theory, 41 (2006) 884–896 <ref:2610.16852#pg18>

[2] N. Castro-González, M.F. Martínez-Serrano, J. Robles, Expressions for the Moore-Penrose inverse of block matrices involving the Schur complement, Linear Algebra Appl. 471 (2015) 353–368 <ref:2610.16852#pg3>

[3] H.H. Cheng, S. Thompson, Dual iterative displacement analysis of spatial mechanisms using the C H programming language, Mech. Mach. Theory, 32 (1997) 193–207 <ref:2610.16852#pg2>

[4] W.K. Clifford, Preliminary sketch of biquaternions, Proc. London Math. Soc., s1-4 (1871) 381–395 <ref:2610.16852#pg6>

[5] D. Condurache, A. Burlacu, Dual tensors based solutions for rigid body motion parameterization, Mech. Mach. Theory, 74 (2014) 390–412

[6] D. Condurache, A. Burlacu, Orthogonal dual tensor method for solving the AX = XB sensor calibration problem, Mech. Mach. Theory, 104 (2016) 382–404 <ref:2610.16852#pg6>

[7] C. Cui, L. Qi, A genuine extension of the Moore–Penrose inverse to dual matrices, J. Comput. Appl. Math., 454 (2025) 116185 <ref:2610.16852#pg3>

[8] D. de Falco, E. Pennestrì, F.E. Udwadia, On generalized inverses of dual matrices, Mech. Mach Theory, 123 (2018) 89–106 <ref:2610.16852#pg2>

[9] G.H. Golub, C.F. Van Loan, Matrix Computations, 4th ed., Johns Hopkins University Press, Baltimore, 2013 <ref:2610.16852#pg3>

[10] Y.L. Gu, J. Luh, Dual-number transformation and its applications to robotics, IEEE Journal on Robotics and Automation, 3 (1987) 615–623 <ref:2610.16852#pg2>

[11] Y. Jin, X. Wang, The application of the dual number methods to Scara kinematics, International Conference on Mechanic Automation and Control Engineering, IEEE 2010, 3871–3874 <ref:2610.16852#pg6>

[12] E. Pennestrì, P.P. Valentini, Linear dual algebra algorithms and their application to kinematics, Multibody Dynamics: Computational Methods and Applications. Dordrecht: Springer Netherlands, (2009) 207–229 <ref:2610.16852#pg2>

[13] E. Pennestrì, P.P. Valentini, D. de Falco, The Moore–Penrose dual generalized inverse matrix with applications to kinematic synthesis of linkages, ASME J. Mech. Design 140 (2018) 102303.1-102303.7 <ref:2610.16852#pg6>

[14] F.E. Udwadia, Dual generalized inverses and their use in solving systems of linear dual equations, Mech. Mach. Theory 156 (2021) 104158 <ref:2610.16852#pg3>

[15] F.E. Udwadia, E. Pennestri, D. de Falco, Do all dual matrices have dual Moore–Penrose generalized inverses? Mech. Mach. Theory, 151 (2020) 103878 <ref:2610.16852#pg3>

[16] H. Wang, Characterizations and properties of the MPDGI and DMPGI, Mech. Mach. Theory, 158 (2021), 104212 <ref:2610.16852#pg6>

[17] T. Wei, W. Ding, Y. Wei, Singular value decomposition of dual matrices and its application to traveling wave identification in the brain, SIAM J. Matrix Anal. Appl. 45 (2024) 634-660 <ref:2610.16852#pg2>

[18] J. M. Fitzpatrick, J. B. West, C. R. Maurer, Jr., Predicting error in rigid-body point-based registration, IEEE Trans. Med. Imaging 17 (1998) 694–702 <ref:2610.16852#pg6>

[19] J. B. West, J. M. Fitzpatrick, S. A. Toms, C. R. Maurer, Jr., R. J. Maciunas, Fiducial point placement and the accuracy of point-based, rigid body registration, Neurosurgery 48 (2001) 810–817 <ref:2610.16852#pg6>

[20] X.H.

Improvements for AI systems

Dual Moore-Penrose generalized inverse computation for rigid-body motion estimation allows AI systems to recover observable motion components from velocity data, even when observation geometry is rank deficient due to first-order perturbations in observation points. This enables precise determination of angular and translational velocities from noisy or incomplete sensor measurements in robotics and computer vision tasks.
The derived formula for the DMPGI of a matrix sum, (UbUbstar + VbVbstar)†, provides an explicit method for calculating the dual inverse of combined system matrices in complex dual number settings. This capability can be used to solve linear dual equations arising in mechanism synthesis and kinematic analysis where standard generalized inverses are insufficient.
The framework supports the recovery of unobservable motion components by showing that the angular velocity component along the line joining these points remains undetermined when using perturbed observation points, as demonstrated in Section 5. This informs AI systems to explicitly model and identify which degrees of freedom are inherently unobservable under specific measurement configurations.
The structure of the DMPGI for products, (UbVb)† = Vb†Ub†, can be leveraged to simplify complex system inversions in motion tracking algorithms involving dual number representations. This leads to computationally efficient solutions for systems where the product structure is known, accelerating real-time processing of kinematic data.
The explicit formula for the DMPGI when standard parts have full column or row rank provides a direct computational path for high-dimensional problems in rigid-body estimation. AI systems can utilize this formula to rapidly compute motion estimates from velocity measurements at two observation points with perturbed positions, ensuring that the recovered components adhere to physical constraints defined by the dual matrix algebra.