Position-Superposed Observer

• Position-superposed observer is a family of methods that structurally embed position data into estimation dynamics, ensuring enhanced observability and stability.
• The approach involves techniques like state augmentation, automorphism-based embedding, and hierarchical composition to linearize nonlinear measurements and improve sensor fusion.
• It also encompasses quantum-reference frame applications where an observer’s spatial superposition enables relational measurements without decoherence.

Searching arXiv for papers on "position-superposed observer" and closely related observer-design usages. Using arXiv search to retrieve papers on the term and its technical variants. A position-superposed observer is a technical expression that appears in distinct research traditions. In recent control and navigation work, it denotes observer architectures in which position measurements, position-related quantities, or position innovations are structurally superposed into estimation dynamics, for example through state augmentation, automorphism-based invariant errors, hierarchical composition, or unified projective measurement embeddings. In recent quantum-reference-frame work, it denotes a laboratory or detector whose center-of-mass position is coherently superposed, with measurements defined on relational observables and recorded in ordinary internal registers (Sifour et al., 5 Dec 2025, Goor et al., 2023, Lopez, 2023, Vanrietvelde, 3 Jul 2026).

1. Terminological scope

The cited literature uses the expression in several technically distinct senses. In observer design, “superposition” usually refers to augmentation, composition, or coherent fusion of position-related information within an estimator. In quantum-reference-frame work, it refers literally to coherent superposition of an observer’s spatial position.

Usage	Mechanism	Representative paper
Single-range inertial navigation	Augment position-related scalars so squared range is a linear output of an extended LTV system	(Sifour et al., 5 Dec 2025)
Position-only INS on $\mathrm{SE}_2(3)$	Superpose an automorphism onto the right-invariant error	(Goor et al., 2023)
Contracting hierarchy	Superpose a translation observer on top of an orientation observer	(Lopez, 2023)
Pose estimation on $\mathrm{SE}(3)$	Superpose position and direction measurements in $\mathbb{RP}^3$	(Hua et al., 2015)
Quantum reference frame	A lab’s center-of-mass position is coherently superposed	(Vanrietvelde, 3 Jul 2026)

This multiplicity of uses matters. The phrase does not designate a single canonical algorithm, but a family of constructions in which position enters an observer in a structurally privileged way. In some papers the emphasis is on linearization of nonlinear outputs, in others on symmetry-preserving error definitions, contraction-preserving interconnection, or relational measurement theory (Sifour et al., 5 Dec 2025, Goor et al., 2023, Hua et al., 2015).

2. Dynamic augmentation in single-range-aided inertial navigation

A particularly explicit control-theoretic meaning is given in “Cascaded Tightly-Coupled Observer Design for Single-Range-Aided Inertial Navigation” (Sifour et al., 5 Dec 2025). There, a position-superposed observer is realized by embedding position-related quantities into an extended linear time-varying observer so that the inherently nonlinear single-range measurement becomes a linear output of the augmented state. The system uses an IMU, a body-frame vector measurement such as a magnetometer, and a distance measurement from a fixed anchor point. In body-frame coordinates,

$$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$

with body-frame kinematics

$$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$

The single-range measurement

$d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$

is nonlinear in position. The paper replaces it by the squared output

$y := \tfrac12 d_y^2$

and differentiates this quadratic quantity along the body-frame dynamics. Four auxiliary coordinates are then introduced,

$$\xi_i = \tfrac12 (\bar{x}^{\mathcal B})^\top C_i \bar{x}^{\mathcal B}, \qquad i=1,\dots,4,$$

with

$$\bar{x}^{\mathcal B} := \operatorname{col}(p^{\mathcal B}, v^{\mathcal B}, g^{\mathcal B}) \in \mathbb{R}^9,$$

so that the augmented state

$$x^{\mathcal B} := \operatorname{col}(\xi_1,\xi_2,\xi_3,\xi_4,\bar{x}^{\mathcal B}) \in \mathbb{R}^{13}$$

satisfies an LTV system whose measured output is linear: $\mathrm{SE}(3)$0 This dynamic augmentation lifts the quadratic constraint into a linear time-varying output, enabling a Riccati/Luenberger-type observer for position, velocity, and gravity direction.

The translational observer is

$\mathrm{SE}(3)$1

with

$\mathrm{SE}(3)$2

and $\mathrm{SE}(3)$3 given by a Riccati equation with $\mathrm{SE}(3)$4 and uniformly positive definite $\mathrm{SE}(3)$5. From $\mathrm{SE}(3)$6 one recovers $\mathrm{SE}(3)$7, $\mathrm{SE}(3)$8, and $\mathrm{SE}(3)$9. Full attitude is then reconstructed on $\mathbb{RP}^3$0 by combining the recovered gravity direction with the magnetometer in a complementary filter: $\mathbb{RP}^3$1

$\mathbb{RP}^3$2

The result is a cascaded architecture: translational state recovery from the extended LTV observer, followed by attitude recovery from two inertial directions (Sifour et al., 5 Dec 2025).

3. Geometric, contracting, and projective variants

A second meaning appears in “Constructive Equivariant Observer Design for Inertial Navigation” (Goor et al., 2023). The INS state is embedded on $\mathbb{RP}^3$3, and the observer introduces an auxiliary automorphism state $\mathbb{RP}^3$4. The invariant error is defined by

$\mathbb{RP}^3$5

so position measurements are “superposed” through automorphism conjugation onto the natural right-invariant error. With suitable correction terms, the error dynamics become autonomous,

$\mathbb{RP}^3$6

and the observer, aided only by GNSS position, is almost-globally asymptotically and locally exponentially stable under persistence of excitation of the specific acceleration in the inertial frame. In this usage, position superposition is not state augmentation but structural embedding of position information into an equivariant error system (Goor et al., 2023).

A third meaning is developed in “A Contracting Hierarchical Observer for Pose-Inertial Fusion” (Lopez, 2023). There, the phrase refers to composition: a quaternion-based orientation and gyroscope-bias observer is upstream, and a translation observer for position, linear velocity, and accelerometer bias is downstream. The downstream observer is superposed on the outputs of the orientation observer, namely $\mathbb{RP}^3$7 and $\mathbb{RP}^3$8. The translation subsystem is uniformly observable in an LTV form, contracting when driven by the true orientation, and remains convergent when the true orientation is replaced by the contracting estimate. In that sense, the observer is position-superposed because the position-related subsystem is hierarchically composed on top of the attitude subsystem (Lopez, 2023).

A fourth meaning is provided by “Gradient-like observer design on the Special Euclidean group $\mathbb{RP}^3$9 with system outputs on the real projective space” (Hua et al., 2015). The observer maps both feature-point position measurements and inertial vector measurements into homogeneous rays in $$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$0 and fuses them in a single gradient-like observer. The cost function

$$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$1

treats position and direction outputs identically in projective coordinates. This is described as the “position-superposition” principle: position-related measurements enter the same cost as pure directions via homogeneous coordinates, allowing one innovation term on $$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$2 and, in an extended design, compensation of unknown constant bias in the velocity measurements (Hua et al., 2015).

Related nonlinear inertial-navigation observers use the term more operationally. “A Nonlinear Navigation Observer Using IMU and Generic Position Information” preprocesses full or partial position sensing into a unified linear output $$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$3 and superposes the resulting innovations in a Riccati-gain translational observer coupled to a complementary attitude filter on $$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$4 (Berkane et al., 2020). “Nonlinear Estimation for Position-Aided Inertial Navigation Systems” injects position-related innovations not only into position and velocity estimates but also into the attitude and gyro-bias channel, so that position aiding stabilizes the coupled translation-rotation estimation without the small-acceleration assumption (Berkane et al., 2021).

4. Observability, excitation, and implementation conditions

Across the observer-design literature, persistency of excitation and uniform observability are recurrent structural requirements. In the single-range-aided LTV observer, uniform observability of the augmented pair $$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$5 is defined by the Gramian condition

$$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$6

and a sufficient PE condition is given in terms of

$$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$7

namely

$$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$8

Under uniform observability and non-collinearity of $$p^{\mathcal B} := R^\top p^{\mathcal I}, \qquad v^{\mathcal B} := R^\top v^{\mathcal I}, \qquad g^{\mathcal B} := R^\top g^{\mathcal I},$$9 and $$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$0, the cascaded error converges to $$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$1, $$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$2 is locally exponentially stable, and the undesired equilibria in $$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$3 are unstable, yielding almost global asymptotic stability. The same paper also states practical conditions: IMU rates of $$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$4–$$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$5 Hz are suitable, UWB at tens of Hz is acceptable, and sensor biases are not estimated in the design (Sifour et al., 5 Dec 2025).

In the generic IMU-plus-position observer, observability is reduced to a condition on the preprocessed position map: $$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$6 For constant $$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$7, this reduces to $$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$8. The paper then gives explicit observability conditions for GPS, multiple UWB ranges, inertial-frame bearings from cameras, and altimeter combinations, showing, for example, that four noncoplanar anchors are sufficient in the range-only case and that an altimeter relaxes anchor or camera configuration requirements. Under the corresponding assumptions, the full observer is semi-globally exponentially stable (Berkane et al., 2020).

Persistent excitation is equally central in the position-only equivariant INS observer and in the single-direction observer with constant bias. In the former, the main condition is PE of the specific acceleration in the inertial frame, $$\dot{p}^{\mathcal B} = -[\omega]_{\times} p^{\mathcal B} + v^{\mathcal B}, \qquad \dot{v}^{\mathcal B} = -[\omega]_{\times} v^{\mathcal B} + g^{\mathcal B} + a^{\mathcal B}, \qquad \dot{g}^{\mathcal B} = -[\omega]_{\times} g^{\mathcal B}.$$9; in the latter, the direction output $d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$0 must satisfy

$d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$1

for global exponential convergence of both position and constant velocity bias estimates (Goor et al., 2023, Bras et al., 2015).

The IPMSM position observer gives yet another engineering realization of position superposition: a probing signal is superposed at standstill to restore observability. The injected voltage

$d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$2

enforces PE when rotor standstill would otherwise make the system unobservable. The observer is globally exponentially convergent under PE, assuming the saliency is not too large, and the experiments use injection amplitudes of $d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$3–$d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$4 V and $d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$5 rad/s (Ortega et al., 2019).

5. Quantum-reference-frame and detector-theoretic usage

In the quantum-reference-frame literature, a position-superposed observer has a literal operational meaning. “Specifying the operational meaning of quantum reference frames” defines it as a laboratory whose absolute center-of-mass position is coherently superposed, for example

$d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$6

The lab’s internal degrees of freedom play the standard role of measurement pointers and memory registers. Measurements performed by such an observer are measurements of relational observables, with relative position

$d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$7

and a controlled-translation unitary

$d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$8

used to pass to the lab’s frame. A key result is that outcomes obtained by a position-superposed observer may be broadcast to a well-localized observer without decohering the superposition, provided the measurement instrument acts only on relational degrees of freedom and is covariant under translations of the lab. The paper explicitly distinguishes this setting from Wigner’s-friend scenarios: the superposition is over positions, not over outcomes (Vanrietvelde, 3 Jul 2026).

A detector-theoretic specialization is given by “Particle detector in a position-superposed black hole spacetime” (Walleghem et al., 13 Apr 2026). There, an Unruh–DeWitt detector is considered in a $d_y(t) = \|p^{\mathcal I}(t) - p_i^{\mathcal I}\|$9d BTZ spacetime containing a black hole in a superposition of locations; after a quantum-reference-frame transformation, the same setup can be viewed as a detector in a superposition of locations in a single classical black hole spacetime. The response probability decomposes as

$y := \tfrac12 d_y^2$0

where the cross term $y := \tfrac12 d_y^2$1 is the nonclassical contribution absent for a classical mixture of positions. The paper further shows that the position-superposed BTZ case yields smooth dependence of the interference response on the detector positions, unlike the mass-superposed case, where extra spectral singularities appear (Walleghem et al., 13 Apr 2026).

A related network-theoretic usage appears in “Genuine quantum networks: superposed tasks and addressing” (Miguel-Ramiro et al., 2020). In that framework, a position-superposed observer can be represented as

$y := \tfrac12 d_y^2$2

with superposed addressing implemented by a quantum control register and local programmable operations. The observer’s measurement is enacted only on the payload at branch $y := \tfrac12 d_y^2$3, while other branches perform the corresponding task on dummy states with matched outcome statistics, preserving coherence across locations (Miguel-Ramiro et al., 2020).

6. Conceptual distinctions, misconceptions, and limitations

A recurring source of confusion is the word “superposed.” In the control and observer-design papers, it does not denote quantum superposition. It refers instead to dynamic augmentation of position-related scalars, superposition of position measurements in a common cost or output map, or hierarchical composition of a position subsystem with an orientation subsystem. In the quantum-reference-frame papers, by contrast, the superposition is literal and concerns the observer’s spatial position (Sifour et al., 5 Dec 2025, Lopez, 2023, Vanrietvelde, 3 Jul 2026).

The practical limitations are similarly domain-specific. Single-anchor single-range aiding offers minimal infrastructure and a tightly coupled full-state design, but it requires sufficient motion excitation, non-collinearity of inertial vectors, and does not estimate sensor biases (Sifour et al., 5 Dec 2025). The position-only equivariant INS observer requires PE of the specific acceleration in the inertial frame (Goor et al., 2023). The contracting hierarchical observer assumes position and orientation measurements and constant IMU biases (Lopez, 2023). The generic IMU-plus-position observer requires a uniformly observable $y := \tfrac12 d_y^2$4 built from the available sensing geometry (Berkane et al., 2020).

Quantum versions impose a different constraint: no operation may leak which-path information about the observer’s absolute position. In the operational account of quantum reference frames, broadcasting works only because the recorded outcome is definite in the internal register and the measurement-and-broadcast map is relational and translation-covariant. In superposed addressing and detector-response settings, coherence is preserved only if branch information is not encoded in auxiliary records or environmental couplings (Vanrietvelde, 3 Jul 2026, Miguel-Ramiro et al., 2020, Walleghem et al., 13 Apr 2026).

This suggests that “position-superposed observer” is best understood as a family resemblance term rather than a single formal object. What unifies the cited usages is not one state space or one proof technique, but a common structural move: position, or position-derived information, is embedded so that it influences estimation or measurement in a more global way than a purely local correction term would permit. In navigation and control, that move serves observability, stability, and tight coupling. In quantum-reference-frame theory, it serves the operational description of measurements performed from a delocalized spatial perspective.