Pith. sign in

REVIEW 4 major objections 4 minor 35 references

Information-Aware Model Predictive Control for Satellite Inspection

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A model predictive controller for autonomous satellite inspection embeds Kalman covariance in its objective, producing trajectories that actively shrink target-feature uncertainty while respecting thrust and safety limits.

desk verdict A solid, transparent application of dual-control-style MPC to satellite inspection, but the 'sensing geometry' claim is mostly distance-minimization under the isotropic measurement model. read the letter →

arxiv 2608.07765 v2 pith:C3MXALQD submitted 2026-08-07 eess.SY cs.SY

Sarah E. Clees , Sean Phillips , Christopher Petersen This is my paper · ORCID
classification eess.SYcs.SY
keywords satelliteinspectionmodelpredictivecontrolcovariancesteeringdualKalmanfilterHill-Clohessy-Wiltshireequationsmeasurementproximityoperations
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that an inspection spacecraft can use its maneuvers more effectively by putting measurement quality into the controller itself. The proposed model predictive control (MPC) law carries the Kalman-filter covariance of target features through the prediction horizon, so each chosen trajectory is valued in part by how much it shrinks that covariance. The key coupling is a measurement covariance that grows with agent-to-target distance: closer flybys produce more informative measurements. In numerical simulations of a cube target in low Earth orbit, the controller drives the agent to the minimum safe distance, holds a bounded orbit there, and reduces the estimated uncertainty of all eight corner points while respecting thrust and safety limits. If the claim holds, autonomous inspection can balance information gain, safety, and fuel without a separate sensing-planning layer.

What carries the argument

The load-bearing object is the pair $(V,F)$: a measurement covariance $V=(\zeta\|Hx\|^2+\varepsilon)I$ that makes sensing quality a scalar function of agent-target distance, and a covariance map $F(\Gamma,\Psi(\alpha))$ that reproduces the Kalman filter's a priori and a posteriori updates in Joseph form. Embedding $F$ as a constraint chain $\Gamma_{k+1}=F(\Gamma_k,\Psi_{k+1}(\alpha_{k+1}))$ inside the MPC problem ties the trajectory to estimation accuracy: the optimizer can only reduce the terminal covariance trace by choosing states that make future measurement covariances small. This machinery turns information awareness into a deterministic nonlinear program with state, control, and covariance decision variables, solved in a receding-horizon fashion.

What would settle it

Re-run the same MPC problem replacing $C$ and $V$ with a line-of-sight-dependent measurement model in which a corner is measured only when the agent lies on the visible side of the cube, keeping all other parameters unchanged; if the controller no longer actively reduces the terminal covariance trace or remains feasible, the distance-only coupling is the source of the reported behavior.

Watch

Extended reading notes

Core claim

The central claim is that estimation covariance can be treated as a controlled state in an MPC framework for spacecraft inspection. By enforcing the Kalman filter's covariance propagation as constraints and penalizing the trace of the covariance in the running and terminal cost, the optimizer produces trajectories that actively reduce uncertainty about target features. With direct measurements of eight fixed cube corners and an isotropic, distance-dependent measurement covariance $V=(\zeta\|Hx\|^2+\varepsilon)I$, the information gain of a trajectory is determined by how close the agent gets to the target. The resulting controller drives the agent inward to the collision-avoidance radius, maintains a natural-motion circumnavigation near that boundary, and lowers the trace of each point's covariance from 3 km$^2$ toward the terminal bound $\Gamma_f = 5\times10^{-6}$ km$^2$, settling at a nonzero level set by process noise. The paper also shows that feasibility is initial-condition dependent: starts far from the target along the NMC semi-major axis can violate the thrust or terminal-distance constraints within the finite horizon.

Load-bearing premise

The argument depends on Assumption 4 and Eq. (14): measurement noise is isotropic and depends only on scalar distance, with direct, always-available measurements of all eight corners; if real sensing is angle-dependent, occluded, or feature-dependent, a distance-minimizing trajectory need not reduce actual estimation error.

Editorial extensions

If this is right

  • An inspection agent will actively pull toward the minimum safe distance because that minimizes measurement noise, then hold a constraint-boundary orbit while control effort and covariance cost balance.
  • The terminal covariance constraint, not just the running cost, guarantees that each finite horizon ends with estimation uncertainty below a chosen threshold.
  • Feasibility is not automatic: initial conditions with large NMC distances, roughly 4 km or more, or starts near the semi-major axis can violate thrust or terminal-distance limits within the fixed horizon.
  • Estimation covariance does not go to zero; it settles to a bounded level set by process noise and the distance-limited measurement noise, consistent with practical stability.
  • The value function decreases monotonically and reaches a near-steady state once the agent reaches the minimum-distance boundary, indicating a predictable closed-loop regime.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the paper's phrasing that trajectories improve observability is stronger than what the isotropic model supports; under Eq. (14) the only sensing geometry is range, so the contribution is best read as range-aware control until occlusion or field-of-view effects are added.
  • Editorial inference: the same coupling could be applied to other information metrics, such as expected information gain or mutual information, replacing $\operatorname{trace}(\Gamma)$ in the cost; the Kalman-covariance version is computationally cheaper but may undervalue viewing angle.
  • Editorial inference: the value-function plateau near the constraint boundary hints at a practical stability argument; one could test numerically whether $\operatorname{trace}(\Gamma_k)$ is a Lyapunov-like decreasing sequence outside a small neighborhood, which the paper leaves as future work.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. This paper presents a model predictive control (MPC) framework for satellite inspection in which the estimation covariance of eight target points of interest is propagated through the prediction horizon via a linear Kalman filter and penalized in the objective, alongside a terminal covariance constraint. The agent dynamics are the discretized Hill-Clohessy-Wiltshire relative-motion equations, and the measurement covariance is taken to be an isotropic function of the scalar agent-target distance. A receding-horizon simulation shows the agent moving to the minimum-distance constraint boundary and then orbiting along it, while the trace of the estimation covariance decreases to the specified terminal bound. A second study maps the feasibility and first-horizon value function over a mesh of initial natural-motion-circumnavigation parameters, identifying regions where actuation and terminal-distance constraints make the problem infeasible.

Significance. If the claims are appropriately scoped, the paper provides a useful demonstration that covariance dynamics can be embedded as constraints in a receding-horizon optimization for relative spacecraft motion, and the initial-condition feasibility mesh is a practical contribution for mission design. The formulation is clearly written, the assumptions are stated, and the authors are explicit in the conclusion about the model's limitations (isotropic, distance-only measurements; no field-of-view or occlusion). The main weakness is that the abstract and introduction claim optimization of 'sensing geometry,' whereas Eq. (14) makes the measurement covariance depend only on range, so the demonstrated behavior is essentially distance minimization subject to safety constraints. The work is reproducible in the sense that all simulation parameters are tabulated, and the numerical feasibility results are not circular.

major comments (4)
  1. [Target State and Measurement Model; Eq. (14), Assumption 4] The measurement covariance in Eq. (14) depends only on the scalar distance ||Hx_i|| and is identical for all eight points of interest, so the controller optimizes range, not sensing geometry. The abstract and introduction claim that the framework 'optimize[s] sensing geometry' and that the measurement model 'depends on the relative geometry,' but the model is isotropic and distance-only. The paper's own Results text (after Fig. 6) states that 'information gain [depends] primarily on the relative distance between the agent and the target rather than viewing geometry,' and the Conclusion defers field-of-view and self-occlusion to future work. Please revise the claims throughout to describe the contribution as range-aware MPC, or extend the model to include per-point line-of-sight and occlusion so that the 'points of interest' are actually distinguished.
  2. [Results, Example Inspection Trajectory] The reported trajectory approaches the target until the r_min constraint binds and then rides the constraint boundary, which is exactly what a distance-minimizing controller under the same constraints would do, since V in Eq. (14) increases monotonically with ||Hx||. No baseline controller is provided; without comparing against, e.g., a zero-control natural-motion circumnavigation, a fixed-radius orbit at r_min, or a greedy distance-minimizing policy, the numerical results cannot establish that the MPC formulation yields any benefit over a non-information-aware policy. Please add a baseline comparison that quantifies final covariance, control effort, and constraint satisfaction.
  3. [Approach, Eqs. (20) and (30)] The covariance reduction is built into the design objective rather than being an emergent property: the running cost penalizes trace(Γ_k) and the terminal constraint forces trace(Γ_N) ≤ Γ_f. The paper should therefore present the covariance decrease as evidence that the constrained optimization is feasible and satisfies its design objective, not as independent evidence that the trajectories are information-aware. The mesh analysis, which examines feasibility under actuation and terminal-distance limits, is the more distinctive result and should be foregrounded as such.
  4. [Target State and Measurement Model, Eqs. (12)-(14)] The measurement model treats all eight points of interest as a single lumped target: V in Eq. (14) is identical for every point and depends only on the distance from the agent to the target center, not on the distance to each corner or on which face of the cube is visible. With the cube side length d = 1 m and r_min = 50 m, the variation between feature distances is negligible, so the simulations do not demonstrate inspection of distinct features. If the contribution is meant to address inspection of specific points, the measurement covariance and output matrix must depend on the individual features; otherwise the scope should be stated as range-based sensing of a point target.
minor comments (4)
  1. [Approach, Eq. (29)] Equation (29) defines F only in words; for reproducibility, write the combined a priori/a posteriori update explicitly, e.g., Γ_{k+1} = (I - K_k C)(A_θ Γ_k A_θ^T + W)(I - K_k C)^T + K_k Ψ_{k+1} K_k^T with K_k as in Eq. (17).
  2. [Initial Condition Feasibility Mesh Analysis, after Fig. 10] The sentence 'All initial conditions were feasible in the terminal covariance constraint or the minimum distance constraint' should use 'and' rather than 'or,' since Fig. 10 shows that the minimum distance and terminal covariance constraints are feasible for all tested initial conditions, while the control magnitude and terminal distance constraints are the ones that fail for some cases.
  3. [References, [23]] Reference [23] lists the year as 2005, but IEEE Transactions on Automatic Control Volume 70 corresponds to 2025; please correct the year.
  4. [Results, Fig. 7 discussion] The statement that the value function 'decreases monotonically over time' is stronger than what receding-horizon MPC generally guarantees; if this is an empirical observation, please label it as such or provide a justification.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the covariance-reduction behavior is an explicit objective of the MPC formulation, not a hidden fitted input, and the feasibility and mesh analyses are self-contained.

full rationale

The paper does not derive a physical prediction from an input that already contains it. The claimed covariance reduction is literally encoded in the MPC objective: Equation (20) includes a running and terminal penalty on trace(Gamma_k), and Equation (30) enforces trace(Gamma_N) <= Gamma_f. The observation that the optimizer drives the agent to the minimum-distance constraint is a direct consequence of the chosen cost and of Assumption 4, Equation (14), which makes measurement covariance an explicit function of distance. The paper itself acknowledges this: it states that the isotropic model is 'intentionally adopted as a simplified approximation of sensing performance to isolate the coupling between trajectory distance and covariance evolution,' and the conclusion lists field-of-view constraints and self-occlusion as future work, rather than claiming those effects are already captured. The feasibility mesh analysis and constraint-activity results are independent of any fitted data and are not circular. The paper's self-citations appear only in contextual references to prior inspection and observability work; no load-bearing uniqueness theorem or ansatz is imported from the authors' own prior papers. Overall, the derivation chain is self-contained: given the dynamics, measurement model, cost, and constraints, the reported trajectories and value functions follow from solving the stated optimization problem.

Assumptions & free parameters 13 free parameters · 8 assumptions · 0 invented entities

The central result rests on a small set of hand-chosen simulation parameters and several simplifying domain assumptions about the target, sensor, and solver. The measurement model in Eq. (14) is the most consequential ad hoc choice: it reduces information gain to a function of distance, so the reported covariance reduction is largely a consequence of the optimization objective. No new physical entities are introduced.

free parameters (13)
  • Measurement covariance scaling zeta = 1e-5
    Chosen by hand in Eq. (14); sets how strongly measurement noise grows with agent-target distance.
  • Minimum measurement variance epsilon = 1e-8 km^2
    Floor on measurement noise in Eq. (14); chosen by hand.
  • Process noise covariance W = 1e-6 * I24 km^2
    Chosen in Table 1; sets the steady-state estimation covariance floor reached at the safety boundary.
  • Covariance cost weight S = 10
    Chosen in Table 1; balances covariance reduction against control effort in Eq. (20).
  • Control cost weight R = I3
    Chosen in Table 1; weights control effort in Eq. (20).
  • Terminal covariance bound Gamma_f = 5e-6 km^2
    Chosen in Table 1; target inspection quality enforced by Eq. (30).
  • Minimum distance r_min = 5e-2 km
    Chosen in Table 1; safety constraint in Eq. (23) that the trajectory hugs after 480 s.
  • Terminal max distance r_max = 1.5 km
    Chosen in Table 1; terminal bound in Eq. (24).
  • Max control u_max = 1 N
    Chosen in Table 1; thruster limit in Eq. (25).
  • Prediction horizon N = 8 time steps
    Chosen in Table 1; affects feasibility and computational load.
  • Discretization step Delta t = 60 s
    Chosen in Table 1; discretizes HCW dynamics.
  • Initial estimation covariance P(t0) = I24 km^2
    Chosen in Table 1; large starting uncertainty used to demonstrate reduction.
  • Initial agent state x(t0) = [0, 1, 0, 5.657e-4, 0, 1.1314e-4] km, km/s
    Chosen as an NMC orbit to provide a feasible warm start.
assumptions (8)
  • domain assumption Hill-Clohessy-Wiltshire linear relative dynamics are valid for the inspection trajectory.
    Invoked in Eqs. (5)-(10); assumes circular target orbit and small relative distance.
  • domain assumption The target is a non-rotating cube with eight fixed corners in the LVLH frame.
    Assumptions 1 and 2 in Target State and Measurement Model; makes the target state matrix identity with q=24.
  • domain assumption Measurements are direct noisy observations of the position of each corner, with identity output matrix.
    Assumption 3; removes feature extraction and bearing-only sensing from the model.
  • ad hoc to paper Measurement noise is isotropic and depends only on scalar agent-target distance.
    Eq. (14); central sensing simplification that makes the information objective equivalent to minimizing distance.
  • standard math Process and measurement noise are zero-mean Gaussian with known covariance matrices.
    Kalman filter model in Eqs. (2)-(3) and Section Kalman Filter.
  • domain assumption All covariance and noise matrices stay diagonal, so Kalman gains and propagated covariances remain diagonal.
    Section on storage vector; used to reduce decision variables to 24 diagonal entries.
  • domain assumption The nonconvex SQP solver with warm start returns a feasible trajectory for each reported feasible initial condition.
    Results rely on fmincon flags and the NMC zero-control warm start; no global optimality or convergence guarantee.
  • domain assumption Discrete-time LTI approximation of HCW is accurate at Delta t=60 s.
    Eqs. (9)-(10) and Table 1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Information-Aware Model Predictive Control for Satellite Inspection." pith.science (2026). https://pith.science/paper/C3MXALQD

@misc{pith2026260807765,
  author       = {Pith},
  title        = {Pith review of: Information-Aware Model Predictive Control for Satellite Inspection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C3MXALQD}},
  note         = {Machine review of arXiv:2608.07765}
}
read the original abstract

Autonomous spacecraft inspection requires trajectories that satisfy safety and control constraints while enabling the collection of informative measurements about a target spacecraft. Traditional guidance and control methods typically decouple estimation from control, resulting in trajectories that do not explicitly optimize sensing geometry. This work presents a model predictive control (MPC) framework that incorporates estimation covariance in the control objective using a formulation inspired by dual control and covariance steering. The estimation covariance evolves according to a linear Kalman filter, and the measurement model depends on the relative geometry between the agent spacecraft and the target. By embedding the covariance dynamics within the MPC problem, the resulting trajectories account for measurement quality, actively reduce uncertainty, and improve observability in the estimated features of the target. The problem is formulated using relative motion dynamics via the Hill-Clohessy-Wiltshire equations with constraints on control input, relative distance, and terminal maximum covariance. Numerical simulations demonstrate that the proposed framework generates feasible inspection trajectories that actively reduce estimation covariance of points of interest on a target while satisfying input and safety constraints of the agent. A mesh analysis of initial conditions further illustrates how feasibility and the value function of the trajectory depend on the initial conditions and constraint activity.

Figures

Figures reproduced from arXiv: 2608.07765 by the authors.

Figure 1
Figure 1. Inertial and LVLH frame representations of motion due to the HCW equa [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. MPC closed-loop diagram. The resulting finite-horizon optimization problem is formulated as follows: min ⃗α,β,⃗γ ⃗ JN = N X−1 k=0 h β⃗⊤ k Rβ⃗ k + S trace(Γk) i + S trace(ΓN ), (20) subject to ⃗α0 = ⃗x(ti), (21) ⃗αk+1 = Ad⃗αk + Bdβ⃗ k, k = 0, . . . , N − 1, (22) ||H⃗αk|| ≥ rmin, k = 0, . . . , N, (23) ||H⃗αN || ≤ rmax, (24) ∥β⃗ k∥∞ ≤ umax, k = 0, . . . , N − 1, (25) Ψ0 = V(⃗x(ti)), (26) Ψk+1 = (ζ||H⃗αk+1||2 + ε)Ip, k… view at source ↗
Figure 3
Figure 3. Agent trajectory in the LVLH frame with the minimum and maximum [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Control input of the agent as seen in the LVLH frame of the target. The [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Measurement covariance over time. The measurement covariance is [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Estimation covariance. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Value function of the MPC solution [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Value function of the first prediction horizon as a function of the agent’s initial condition. [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: fmincon flag output. Black indicates infeasible solutions that converged, and white indicates feasible solutions that converged. Gray indicates suboptimal solu￾tions that terminated prior to convergence due to the SQP iteration limit. Feasibility of the gray points dep…
Figure 10
Figure 10. Figure 10: Constraint feasibility for inequality constraints. Black indicates infeasible [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]

Review history (3 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 4 linked inside Pith

  1. [1]

    Safe and Constrained Rendezvous, Prox- imity Operations, and Docking,

    C. D. Petersen, R. J. Caverly, S. Phillips, and A. Weiss, “Safe and Constrained Rendezvous, Prox- imity Operations, and Docking,”American Control Conference (ACC), San Diego, CA, May 2023, 10.23919/ACC55779.2023.10155826

    arXiv 2023

  2. [2]

    Basics of Satellite Communications,

    E. K. Musisi, “Basics of Satellite Communications,”ITU/ITSO Workshop on Satellite Communications, Aug. 2016

    2016

  3. [3]

    Dual Control Theory I,

    A. A. Feldbaum, “Dual Control Theory I,”Avtomat. i Telemekh, 1960

    1960

  4. [4]

    Dual Control Theory Problems,

    A. A. Feldbaum, “Dual Control Theory Problems,”IF AC Proceedings, V ol. 1, 1963

    1963

  5. [5]

    Stochastic Model Predictive Control with Active Uncertainty Learning: A Survey on Dual Control,

    A. Mesbah, “Stochastic Model Predictive Control with Active Uncertainty Learning: A Survey on Dual Control,”Annual Reviews in Control, 2018, 10.1016/j.arcontrol.2017.11.001

    doi:10.1016/j.arcontrol.2017.11.001 2018

  6. [6]

    Deep Reinforcement Learning for Multi- Agent Autonomous Satellite Inspection,

    H. H. Lei, M. Shubert, N. Damron, K. Lang, and S. Phillips, “Deep Reinforcement Learning for Multi- Agent Autonomous Satellite Inspection,”AAS Guidance, Navigation, and Control Conference, Breck- enridge, CO, Feb. 2022, 10.1007/978-3-031-51928-4 76

    doi:10.1007/978-3-031-51928-4 2022

  7. [7]

    Trajectory Synthesis for the Coordinated Inspection of a Spacecraft with Safety Guarantees,

    M. Hibbard, M. Cubuktepe, M. Shubert, K. Lang, U. Topcu, and S. Phillips, “Trajectory Synthesis for the Coordinated Inspection of a Spacecraft with Safety Guarantees,”AIAA Journal of Guidance, Control, and Dynamics, 2023

    2023

  8. [8]

    Exposure-Based Multi-Agent Inspection of a Tumbling Target using Deep Reinforcement Learning,

    J. Aurand, S. Cutlip, H. Lei, K. Lang, and S. Phillips, “Exposure-Based Multi-Agent Inspection of a Tumbling Target using Deep Reinforcement Learning,”AAS Guidance, Navigation, and Control Con- ference, Breckenridge, CO, Feb. 2023, arXiv:2302.14188v2

    arXiv 2023

Show all 35 references
  1. [9]

    Assessing Autonomous Inspection Regimes: Active Versus Passive Satellite Inspection,

    J. Aurand, C. Pang, S. Mokhtar, H. Lei, S. Cutlip, and S. Phillips, “Assessing Autonomous Inspection Regimes: Active Versus Passive Satellite Inspection,”AIAA SciTech F orum, Orlando, FL, Jan. 2025, 10.2514/6.2025-0755

    2025 doi

  2. [10]

    Optimal NMC Trajectory Design for On-Orbit, Multi- Agent Inspection Missions,

    M. R. Mercier, D. H. Curtis, and S. R. Phillips, “Optimal NMC Trajectory Design for On-Orbit, Multi- Agent Inspection Missions,”AIAA SciTech F orum, Orlando, FL, Jan. 2024, 10.2514/6.2024-1867

    2024 doi

  3. [11]

    Enabling Proliferated Space Sensor Awareness Constellations Through Topological Observability,

    C. W. Hays, T. Henderson, S. Phillips, and A. A. Soderlund, “Enabling Proliferated Space Sensor Awareness Constellations Through Topological Observability,”Proceedings of the 2023 AIAA SciTech F orum, National Harbor, MD, 2023

    2023

  4. [12]

    A Mathematical Theory of Communication,

    C. E. Shannon, “A Mathematical Theory of Communication,”The Bell System Technical Journal, 1948

    1948

  5. [13]

    Probabilistic Robotics,

    S. Thrun, “Probabilistic Robotics,”Communications of the ACM, 2002. https://dl.acm.org/doi/fullHtml/10.1145/504729.504754

    2002

  6. [14]

    Information Based Adaptive Robotic Exploration,

    F. Bourgault, A. A. Makarenko, S. B. Williams, B. Grocholsky, and H. F. Durrant-Whyte, “Information Based Adaptive Robotic Exploration,”IEEE/RSJ International Conference on Intelligent Robots and Systems, Lausanne, Switzerland, Sept. 2002, 10.1109/IRDS.2002.1041446

    2002 arXiv

  7. [15]

    Simon,Optimal State Estimation

    D. Simon,Optimal State Estimation. John Wiley & Sons, 2006

    2006

  8. [16]

    T. P. Setterfield,On-Orbit Inspection of a Rotating Object Using a Moving Observer. PhD thesis, Mas- sachusetts Institute of Technology, 2017. https://dspace.mit.edu/entities/publication/8c71690b-873b- 4407-bf02-8932d2149be3

    2017

  9. [17]

    Measurement-limited Multi-Agent, Relative Pose Estimation for On-Orbit Inspection,

    M. Mercier, D. Curtis, and C. Taylor, “Measurement-limited Multi-Agent, Relative Pose Estimation for On-Orbit Inspection,”IEEE International Conference on Robotics and Automation, Yokohama, Japan, May 2024, 10.1109/ICRA57147.2024.10611394

    2024

  10. [18]

    Simultaneous Localization and Mapping For Satellite Rendezvous and Proximity Operations Using Random Finite Sets,

    L. Schlenker, M. Moretto, D. Gaylor, and R. Linares, “Simultaneous Localization and Mapping For Satellite Rendezvous and Proximity Operations Using Random Finite Sets,”AAS/AIAA Space Flight Mechanics Meeting, Ka‘anapali, HI, Jan. 2019

    2019

  11. [19]

    Line-Based Monocular SLAM for Spacecraft RPO,

    I. G. Velentzas, M. Dor, and P. Tsiotras, “Line-Based Monocular SLAM for Spacecraft RPO,”Space Imaging Workshop, Atlanta, GA, Oct. 2024

    2024

  12. [20]

    Decoupled, Consistent Node Removal and Edge Sparsifica- tion for Graph-Based SLAM,

    K. Eckenhoff, L. Paull, and G. Huang, “Decoupled, Consistent Node Removal and Edge Sparsifica- tion for Graph-Based SLAM,”IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Daejeon, Korea (South), Oct. 2016, 10.1109/IROS.2016.7759505

    2016

  13. [21]

    MS-SLAM: Memory Efficient Visual SLAM with Sliding Window Map Sparsification,

    X. Zhang, J. Dong, Y . Zhang, and Y .-H. Liu, “MS-SLAM: Memory Efficient Visual SLAM with Sliding Window Map Sparsification,”Journal of Field Robotics, 2024, 10.1002/rob.22431

    2024 doi

  14. [22]

    A Covariance Control Theory,

    A. F. Hotz and R. E. Skelton, “A Covariance Control Theory,”IEEE Conference on Decision and Con- trol (CDC), Fort Lauderdale, FL, Dec. 1985, 10.1109/CDC.1985.268547. 18

    1985

  15. [23]

    Optimal Covariance Steering for Discrete-Time Linear Stochastic Systems,

    F. Liu, G. Rapakoulias, and P. Tsiotras, “Optimal Covariance Steering for Discrete-Time Linear Stochastic Systems,”IEEE Transactions on Automatic Control, V ol. 70, April 2005, 10.1109/TAC.2024.3472788

    2005

  16. [24]

    Optimal Covariance Control for Discrete-Time Stochastic Linear Systems Subject to Constraints,

    E. Bakolas, “Optimal Covariance Control for Discrete-Time Stochastic Linear Systems Subject to Constraints,”IEEE Conference on Decision and Control (CDC), Las Vegas, NV , Dec. 2016, 10.1109/CDC.2016.7798422

    2016

  17. [25]

    Input Hard Constrained Optimal Covariance Steering,

    K. Okamoto and P. Tsiotras, “Input Hard Constrained Optimal Covariance Steering,”IEEE Conference on Decision and Control (CDC), Dec. 2019, 10.1109/CDC40024.2019.9029353

    2019

  18. [26]

    Optimal Covariance Control for Stochastic Sys- tems Under Chance Constraints,

    K. Okamoto, M. Goldshtein, and P. Tsiotras, “Optimal Covariance Control for Stochastic Sys- tems Under Chance Constraints,”IEEE Control Systems Letters, V ol. 2, April 2018, 10.1109/LC- SYS.2018.2826038

    2018

  19. [27]

    Covariance Steering with Optimal Risk Allocation,

    J. Pilipovsky and P. Tsiotras, “Covariance Steering with Optimal Risk Allocation,”arXiv.org, Apr. 2021, arXiv:2009.09554v2

    2021 arXiv

  20. [28]

    Stochastic Model Predictive Control for Constrained Linear Systems Using Optimal Covariance Steering,

    K. Okamoto and P. Tsiotras, “Stochastic Model Predictive Control for Constrained Linear Systems Using Optimal Covariance Steering,”arXiv.org, Nov. 2019, arXiv:1905.13296v2

    2019 arXiv

  21. [29]

    Finite-Horizon Separation-Based Covariance Control for Discrete-Time Stochastic Lin- ear Systems,

    E. Bakolas, “Finite-Horizon Separation-Based Covariance Control for Discrete-Time Stochastic Lin- ear Systems,”IEEE Conference on Decision and Control (CDC), Miami Beach, FL, Dec. 2018, 10.1109/CDC.2018.8619542

    2018

  22. [30]

    Terminal Guidance System for Satellite Rendezvous,

    W. H. Clohessy and R. S. Wiltshire, “Terminal Guidance System for Satellite Rendezvous,”Journal of the Aerospace Sciences, V ol. 27, No. 9, 1960

    1960

  23. [31]

    D. A. Vallado,Fundamentals of Astrodynamics and Applications. Microcosm Press, 2013

    2013

  24. [32]

    J. P. Hespanha,Linear Systems Theory. Princeton University Press, 2018

    2018

  25. [33]

    J. S. Bendat and A. G. Piersol,Random Data: Analysis and Measurement Procedures. John Wiley & Sons, 1971

    1971

  26. [34]

    https://www.mathworks.com/help/optim/ug/fmincon.html

    The MathWorks, Inc., Natick, Massachusetts, United States,fmincon, 2025. https://www.mathworks.com/help/optim/ug/fmincon.html

    2025

  27. [35]

    Frey,Advances in Constrained Spacecraft Relative Motion Planning

    G. Frey,Advances in Constrained Spacecraft Relative Motion Planning. PhD thesis, University of Michigan, 2018. https://deepblue.lib.umich.edu/items/a15cc144-8ce9-494d-85e8-26a878c7d753. 19

    2018

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.