Chapter Four · failure evidence
What State Estimation & Kalman Filtering got wrong, from 86 dissertations
Across these studies, Kalman filters and related state estimation algorithms frequently suffer from divergence, excessive computational complexity, and vulnerability to model misspecification. Practitioners often observe that filters break down under severe nonlinearities, unmodeled physical dynamics, and sparse or occluded measurements, leading researchers to reject them in favor of simpler observers or empirical baselines. These records come from PhD theses at 25 institutions, 2021 to 2026. Each links to its thesis. They were extracted by language models reading the full text, so treat each as a lead to read, not a verdict.
Linearization errors and severe nonlinearities cause filter divergence
First-order Taylor series expansions and sigma-point approximations break down when applied to highly nonlinear kinematics, coordinated maneuvers, and non-Newtonian dynamics. These linearization failures lead to unbounded covariance growth, tracking divergence, and severe estimation errors.
Tried and failed
constant-acceleration Extended Kalman Filter applied to mobile emitter tracking with position error. Outcome: did not converge. Reason: diverged due to large relative position errors, performing worse than standard linear Kalman filter
Data-Driven Localization and Structure Learning in Reverberant Underwater Acoustic Environments · MIT
Tried and failed
extended Kalman filter for nonlinear depth estimation applied to visual feature depth tracking. Outcome: unstable. Reason: linearization breakdown under high measurement nonlinearity caused filter divergence and unphysical negative values
Spacecraft navigation and decision making in uncertain environments · UT Austin
Tried and failed
extended Kalman filter SLAM applied to robot localization with odometry noise. Outcome: did not converge. Reason: linearization errors from odometry uncertainty cause divergence in stationary counterexample scenarios
Onboard control, tracking and navigation for autonomous systems · UT Austin
Tried and failed
conventional extended Kalman filter state estimation applied to indirect wind velocity estimation. Outcome: did not converge. Reason: linearization errors during high-dynamic coordinated turning maneuvers caused filter divergence
An Invariant Extended Kalman Filter for Indirect Wind Estimation Using a Small, Fixed-Wing Uncrewed Aerial Vehicle · Virginia Tech
Tried and failed
particle filter with extended Kalman filter state estimation applied to jump-Markov state estimation. Outcome: worse than baseline. Reason: linearization errors in the extended Kalman filter reduced estimation accuracy
Advances in the Use of Finite-Set Statistics for Multitarget Tracking · Virginia Tech
Tried and failed
linear Kalman filtering on nonlinear kinematics applied to aircraft state and wind estimation. Outcome: did not converge. Reason: Maneuvering kinematics violated nominal equilibrium linearization assumptions, causing filter divergence
Identification of Unsteady Flight Dynamic Models and Model-based Wind Estimation with Flight Test Validation · Virginia Tech
Tried and failed
conventional extended Kalman filter applied to aerodynamic wind estimation from inertial data. Outcome: did not converge. Reason: linearization errors caused filter divergence during unaccelerated straight and level flight
An Invariant Extended Kalman Filter for Indirect Wind Estimation Using a Small, Fixed-Wing Uncrewed Aerial Vehicle · Virginia Tech
Tried and failed
extended Kalman filtering with kinematic motion models applied to trajectory prediction of maneuvering targets. Outcome: worse than baseline. Reason: linear kinematic assumptions cannot accurately capture complex nonlinear maneuvering dynamics
Data-driven Target Tracking and Hybrid Path Planning Methods for Autonomous Operation of UAV · Virginia Tech
Lost to a baseline
In the Circular Restricted Three-Body Problem (CR3BP) northern halo orbit tracking test, the standard linear-update UKF diverges right before 3 TU (loses tracking of the spacecraft), losing to the higher-order quadratic and polynomial filters (QUKF, QACUKF-4, CACUKF-6).
Polynomial Kalman filter updates · Iowa State
Considered and rejected
Considered and rejected: Rejected standard iterative Taylor linearization / Extended Kalman Filtering because range and angle measurements are highly nonlinear and underdetermined, failing to maintain Gaussianity.
Considered and rejected
Considered and rejected: Rejected classical Extended Kalman Filter (EKF) and linear observers due to Jacobian linearization errors and inability to handle abrupt feed discontinuities
Advancing bioprocessing with ensemble kalman filter: from state estimation to knowledge transfer · Imperial
Considered and rejected
Considered and rejected: Rejected using extended Kalman filter (EKF) linearizations because propagating first/second-order approximations caused large errors in highly non-linear arm dynamics
Efficient and Safe Robot Planning in Human-Robot Collaboration · DSpace at SUNY Buffalo
Considered and rejected
Considered and rejected: Rejected Extended Kalman Filter (EKF) and Particle Filter in favor of Unscented Kalman Filter due to EKF instability on highly non-linear measurement equations and Particle Filter computational inefficiency.
Advances in passive acoustic detection, localization, and tracking applied to unmanned underwater vehicles · Woods Hole
Tried and failed
Gaussian companion filters for particle flow uncertainty propagation applied to highly nonlinear state estimation. Outcome: unstable. Reason: linearized or unscented covariance approximations lead to statistically inconsistent uncertainty estimates in severe nonlinearities
Adventures in Kalman filtering : exploring methods of expanding the uses of the Kalman filter · UT Austin
Considered and rejected
Considered and rejected: Rejected using extended Kalman filtering with Jacobian linearizations due to complex nonlinear aerodynamics, selecting an Unscented Kalman Filter instead.
Real Time Local Wind Inference for Robust Autonomous Navigation · Penn
Considered and rejected
Considered and rejected: Rejected the Extended Kalman Filter (EKF) in favor of the Unscented Kalman Filter (UKF) to avoid linearisation approximation errors from Taylor series expansion while keeping computational complexity low.
Identification of nonlinear and time-varying systems under dynamic and seismic excitation · IRIS - POLITO - prod
Considered and rejected
Considered and rejected: Rejected Extended Kalman Filter (EKF) in favor of Unscented Kalman Filter (UKF) because the hydraulic model's non-Newtonian frictional pressure loss terms require numerical solutions without explicit analytical forms, making linearization impractical.
Autonomous steering and event detection : modeling, estimation, and control in drilling engineering · UT Austin
Kalman filter variants are frequently outperformed by simpler baselines
Complex Kalman filtering formulations often produce larger estimation errors and lag than simple linear regressions, moving averages, or deterministic observers. Additionally, advanced nonlinear variants such as unscented Kalman filters frequently fail to improve accuracy over basic linear filters or raw sensor signals.
Lost to a baseline
Colored noise Kalman filter with augmented random bias state loses optimality guarantee, yielding higher K-L divergence metric Eopt than an ideal Kalman filter with known disturbance input
Aerospace Applications of Noise Covariance Identification with Autocovariance Least Squares · ResearchWorks
Lost to a baseline
Manually tuned Extended Kalman filters (EKF 1, EKF 2, EKF 3) were outperformed by the LTV-ALS tuned Extended Kalman filter, which achieved lower K-L divergence metric Eopt and smaller initial transient oscillations
Aerospace Applications of Noise Covariance Identification with Autocovariance Least Squares · ResearchWorks
Lost to a baseline
Kalman Filter was outperformed by the Levant Observer and Luenberger Observer in relative velocity estimation accuracy under noisy position measurements.
Guidance and Navigation Algorithms for Spacecraft Low-Thrust Proximity Operations: Formation Flight in Circular Relative Orbit · IRIS - POLITO - prod
Lost to a baseline
Misspecified linear Kalman filter yielded higher correlations but worse relative standard deviations in non-linear state space setups compared to the partial information filter.
ESSAYS ON MACRO FINANCE · Penn
Lost to a baseline
Variance-scaled Kalman filtering (KFSA/KFSB) exhibited larger parameter estimation biases and variance than variance-scaled Gaussian (GS) and negative binomial (NB) estimators under high noise (q_p, q_m = 2).
Enhancing methods for modeling and estimation of complex socio-technical systems · MIT
Lost to a baseline
Kalman Filter and Moving Average smoothing produced higher TTC MAE (0.81 s and 0.86 s) compared to raw UWB measurements (0.62 s excluding 20 m).
Utilization of Wireless Sensors for Pedestrian Safety Studies · Carleton University Institutional Repository
Lost to a baseline
Kalman with GCV hyperparameter optimization performed worse than Savitzky-Golay gridsearched optima on several benchmark systems.
Open-Source Dynamical Systems Research, with a Side of (Francis) Bacon · ResearchWorks
Tried and failed
unscented Kalman filter for motion state estimation applied to pedestrian multi-object tracking. Outcome: worse than baseline. Reason: nonlinear filtering degraded tracking accuracy compared to standard linear Kalman filter models
Improved 2D Camera-Based Multi-Object Tracking for Autonomous Vehicles · Virginia Tech
Tried and failed
Unscented Kalman Filter state estimation applied to nonlinear dynamic wind estimation. Outcome: worse than baseline. Reason: Higher computational overhead and tuning complexity yielded no significant improvement over Extended Kalman Filter.
Identification of Unsteady Flight Dynamic Models and Model-based Wind Estimation with Flight Test Validation · Virginia Tech
Tried and failed
incorporating unsteady dynamic models into kalman filters applied to wind velocity estimation in flight. Outcome: worse than baseline. Reason: unsteady aerodynamic models yielded marginally higher root mean square deviation than quasi-steady models
Identification of Unsteady Flight Dynamic Models and Model-based Wind Estimation with Flight Test Validation · Virginia Tech
Lost to a baseline
Kalman filter prediction method (185 m/year average RMSE) was beaten by simple linear regression rate extrapolation (128 m/year average RMSE).
Lost to a baseline
Kalman smoothing with GCV hyperparameter optimization was outperformed by gridsearched Kalman smoothing and gridsearched Savitzky-Golay across multiple benchmark ODEs.
Open-Source Dynamical Systems Research, with a Side of (Francis) Bacon · ResearchWorks
Lost to a baseline
Under random time series errors alone, standalone IRS (0.46995 NM/HR 95% error) outperformed the integrated Kalman filter (0.876 NM/HR 95% error)
Integration of global positioning and inertial reference system data inside a flight management computer · Cranfield
Lost to a baseline
Standard FairMOT and Faster R-CNN/YOLOv8 outperformed standard Kalman filtering in tracking and speed estimation accuracy in video processing.
New modelling approaches to analyse unsafe traffic conditions in real-time · Imperial
Considered and rejected
Considered and rejected: Rejected low-pass filters and Kalman filters for prediction output post-processing in favor of EMA filter due to superior smoothing and stability
Investigation of Future Voluntary Movement Prediction for Pathological Tremor-Alleviating Exoskeletons · Virginia Tech
Lost to a baseline
robust Kalman Filter was beaten by the Certainty Equivalent Kalman Filter on benign simulations by up to 1.64x mean squared error
Statistical Learning For System Identification, Estimation, And Control · Penn
Unmodeled physical dynamics and structural misspecification corrupt state estimates
Filters fail to maintain stable tracking when nominal equations ignore key physical phenomena such as frictional slip, actuation delays, or stiffness variations. Missing terms in the system model and unmodeled disturbances lead to continuous estimation drift, lag, and filter instability.
Tried and failed
augmented Extended Kalman Filter parameter estimation applied to structurally misspecified dynamic systems. Outcome: did not converge. Reason: missing model dynamics terms cause parameter estimates to diverge or converge to incorrect values
Online Information-Aware Motion Planning with Model Improvement for Uncertain Mobile Robotics · MIT
Tried and failed
Extended Kalman filter online parameter estimation applied to car-following model parameter identification. Outcome: did not converge. Reason: Persistent excitation order was insufficient to prevent divergence and high error covariance in linear parameter estimation
Vehicle Longitudinal Control under Autonomy, Connectivity, and Mixed-flow Traffic · Georgia Tech
Tried and failed
sub-optimal discrete Kalman filter with multirate updates applied to multisensor navigation state estimation. Outcome: did not converge. Reason: Unmatched filter dynamics failed to maintain unobservable error states, corrupting bias and tilt estimates.
Integration of global positioning and inertial reference system data inside a flight management computer · Cranfield
Tried and failed
extended Kalman filter wheel odometry fusion applied to mobile robot track navigation. Outcome: unstable. Reason: unmodeled variable frictional slip between drive wheels and contact surfaces degraded state estimation
A novel railway maintenance robot for inspection and repair · Cranfield
Considered and rejected
Considered and rejected: Rejected the standard uncorrected Extended Kalman Predictor (EKP) for position states because ignorable coordinates with non-zero equilibrium velocity (e.g., X, Y, Z, psi) accumulate an uncompensated drift/bias shift over the delay horizon.
Time Delay Mitigation in Aerial Telerobotic Operations Using Predictors and Predictive Displays · Virginia Tech
Considered and rejected
Considered and rejected: Rejected full state-space / Kalman filter models due to severe parameter proliferation, specification errors compounding across horizons, and computational complexity.
Using Mixed Frequency Data to Forecast Recessions and GDP · ResearchWorks
Tried and failed
zero-order hold discretization in unscented Kalman filter applied to dynamic parameter estimation. Outcome: worse than baseline. Reason: insufficient approximation of continuous-time dynamics leading to significant parameter estimation errors compared to first-order hold
Finite element model updating of exponential non-viscous damping systems · Georgia Tech
Tried and failed
Kalman state estimation with underestimated stiffness parameter applied to compliant parallel robot state estimation. Outcome: unstable. Reason: Underestimating the stiffness matrix in the filter introduced lag and oscillation, destabilizing closed-loop whole-body control
Design and Control of a Structurally Elastic Humanoid Robot · Virginia Tech
Tried and failed
Kalman filtering without sensor delay compensation applied to fast dynamic robot state estimation. Outcome: unstable. Reason: ignoring uncorrected sensor latencies caused large error accumulation during rapid transient phases
Estimation and Planning for Dynamic Robot Behaviors · Harvard
Tried and failed
adaptive extended Kalman filter under dynamics mismatch applied to multi-agent state estimation during propagation. Outcome: did not generalise. Reason: adaptive covariance estimation failed to compensate for severe structural model mismatch, increasing cumulative uncertainty
Enhancing Teamwork in Multi-Robot Systems: Embodied Intelligence via Model- and Data-Driven Approaches · Georgia Tech
Tried and failed
full-state extended Kalman filter on discretized PDE applied to convection-diffusion bioreactor state estimation. Outcome: unstable. Reason: numerical instability caused by ill-conditioned covariance matrix inversion with high-dimensional discretized states
Digital Twin Design and Autonomous Control of Bioreactor Systems for Human Immune Cell Expansion · Georgia Tech
Tried and failed
finite input covariance filter for joint estimation applied to nonlinear dynamic structural state estimation. Outcome: unstable. Reason: incorrect initial input covariance causes severe low-frequency drift and underestimates residual state displacements
Input-State Estimation of Inelastic Structural Systems: Theoretical Framework and Experimental Validation · Georgia Tech
Tried and failed
ensemble Kalman filter data assimilation applied to bioreactor kinetic state estimation. Outcome: unstable. Reason: unmodeled time delays and discrete kinetic switching caused filter instability, forcing state exclusion
Advancing bioprocessing with ensemble kalman filter: from state estimation to knowledge transfer · Imperial
Tried and failed
neural network prediction of Kalman filter gains applied to multi-sensor state estimation. Outcome: did not generalise. Reason: Model assumption dependency, noise sensitivity, and failure to adapt to unmodeled environmental and sensor anomalies.
ML-Enhanced Visual Inertial Navigation System for UAVs · Cranfield
High computational overhead prevents real-time embedded execution
Full-state covariance matrix inversions, high-dimensional Jacobian calculations, and iterative ensemble queries exceed the processing capabilities of microcontrollers and real-time platforms. Consequently, researchers frequently discard full Kalman filtering architectures in favor of lower-order models or complementary filters.
Considered and rejected
Considered and rejected: Rejected full-Jacobian all-pixel Extended Kalman Filter (EKF) covariance matrix tracking due to intractable O((2*N_pix)^3) matrix inversion complexity, restricting analysis to single-pixel EKF
Embedded Computing for Wavefront Control on Future Space Telescopes · MIT
Considered and rejected
Considered and rejected: Rejected running BPTT through the Kalman filter in a recurrent feedback hybrid architecture due to excessive compute and optimization instabilities.
Verfahren zur Zustandsschätzung im LKW-Trailer · Leibniz Universität Hannover Repository
Tried and failed
particle filter state estimation applied to battery state of charge estimation. Outcome: too slow. Reason: Substantially higher computational cost without noticeable accuracy improvement over extended Kalman filtering
Advanced state of charge estimation for lithium-sulfur batteries. · Cranfield
Tried and failed
Extended Kalman filter with linearized observation model applied to sensor state estimation on microcontrollers. Outcome: too slow. Reason: linearizing piecewise nonlinear observation equations exceeded the microcontroller's computational capacity
Modular Robots Morphology Transformation And Task Execution · Penn
Considered and rejected
Considered and rejected: Rejected nonlinear filtering (unscented Kalman filter) and Bayesian parameter estimation due to prohibitive computational expense on big data.
Considered and rejected
Considered and rejected: Rejected Kalman filter for combining accelerometer and gyroscope data due to its high computational processing cost, choosing complementary filter instead
Implementación de un sistema no invasivo para la identificación del nivel de atención en personas · Repositorio Institucional BUAP
Considered and rejected
Considered and rejected: Rejected using a single full 3-D Kalman filter with 6x6 matrices due to higher computing complexity and calculations, choosing two 2-D KFs with 4x4 matrices instead
SELECTIVE BEAMSTEERING AND 3-D TRACKING PHASED ARRAY RADAR WITH REDUCED DIMENSION KALMAN FILTERING · Calhoun
Considered and rejected
Considered and rejected: Rejected dedicated ship motion estimators (e.g. Unscented Kalman Filters, Prony analysis) because they require substantial initialization time and add computational complexity.
Robust Control for a Quadrotor Unmanned Aerial Vehicle in Complex Environments · Texas Tech
Considered and rejected
Considered and rejected: Rejected particle filter (PF) state estimation in favor of iSAM2 nonlinear least-squares due to excessive computational cost
Adaptive AUV-assisted Diver Navigation for Loosely-Coupled Teaming in Undersea Operations · MIT
Considered and rejected
Considered and rejected: Rejected a full extended Kalman filter (AHRS) running real-time on the microcontroller due to high computational demand.
Sensor-based electronic monitoring of feeding and drinking activity of nursery pigs in swine farms · Iowa State
Considered and rejected
Considered and rejected: Rejected Extended Kalman Filter (EKF) in favor of standard Kalman Filter (KF) within IMM due to EKF's computational complexity and runtime for non-real-time offline data processing.
Statistical Modeling of Air Traffic: Development of Methods and Application through a Canadian Case Study · Carleton University Institutional Repository
Considered and rejected
Considered and rejected: Decided against using the Unscented Kalman Filter in Chapter 5 in favor of the Extended Kalman Filter to reduce computation time during large simulation runs
Mission-driven Sensor Network Design for Space Domain Awareness · Virginia Tech
Considered and rejected
Considered and rejected: Rejected Unscented Kalman Filter (UKF) and Ensemble Kalman Filter (EnKF) in favor of EKF because they require multiple or ensemble model queries per update rather than a single differentiable forward pass.
An Approach for Rapid, Uncertainty-aware Damage Diagnosis of Rotating Machinery · Georgia Tech
Covariance underestimation and ensemble collapse destabilize data assimilation
Ensemble Kalman filters experience rapid loss of ensemble spread and artificial variance reduction in high-dimensional and chaotic systems. Without proper spatial localization or correction routines, spurious correlations accumulate and drive the filter toward numerical divergence.
Tried and failed
ensemble Kalman filter with high update frequency applied to hydrodynamic state estimation. Outcome: unstable. Reason: diminishing ensemble spread caused filter divergence, leading to ignored observations
Tried and failed
multilevel ensemble Kalman filtering applied to quasigeostrophic fluid dynamics data assimilation. Outcome: did not converge. Reason: lack of forecast corrections, mean corrections, and spatial covariance localization caused filter divergence
Combining Data-driven and Theory-guided Models in Ensemble Data Assimilation · Virginia Tech
Considered and rejected
Considered and rejected: Rejected K-means hard clustering in PFGMM for light-curve tracking due to filter divergence from neglecting component covariances.
Sequential Monte Carlo filtering with Gaussian mixture models for highly nonlinear systems · UT Austin
Considered and rejected
Considered and rejected: Rejected comparing raw linear trends of the Paleoclimate Data Assimilation (PDA) ensemble mean against instrumental data because offline ensemble Kalman filtering artificially attenuates posterior variance relative to targets.
Decadal to centennial-scale climate interactions across the Indo-Pacific region · Woods Hole
Considered and rejected
Considered and rejected: Rejected evaluating linear warming trends using the Paleoclimate Data Assimilation (PDA) ensemble mean due to Kalman filter variance loss and time-dependent proxy network dropouts.
Decadal to centennial-scale climate interactions across the Indo-Pacific region · MIT
Tried and failed
ensemble Kalman filter applied to Lagrangian data assimilation in chaotic flows. Outcome: unstable. Reason: assimilation intervals exceed the Lagrangian autocorrelation timescale causing filter divergence from chaotic hyperbolic stretching
Tried and failed
estimating state error covariance directly from ensemble mean applied to ensemble Kalman filter epidemic forecasting. Outcome: unstable. Reason: led to numerical errors and ensemble collapse during forecasting
Validating Forecasting Strategies of Simple Epidemic Models on the 2015-2016 Zika Epidemic · Virginia Tech
Tried and failed
ensemble Kalman filter without covariance localization applied to large-scale high-dimensional data assimilation. Outcome: unstable. Reason: spurious correlations caused posterior variance collapse in high dimensions
Ensemble Kalman filtering and conditional normalizing flows for seismic monitoring via data assimilation · Georgia Tech
Tried and failed
decentralised Kalman filter without joint covariance tracking applied to cooperative multi-agent relative state estimation. Outcome: unstable. Reason: inter-agent errors become correlated after mutual ranging, causing filter divergence during rank reversals
The application of relative navigation to civil air traffic management · Cranfield
Tried and failed
ensemble Kalman filter on non-Gaussian spatial fields applied to subsurface permeability parameter estimation. Outcome: worse than baseline. Reason: violation of Gaussian assumptions smeared sharp facies boundaries and distorted production forecasts
Rule-based and machine learning hybrid reservoir modeling for improved forecasting · UT Austin
Tried and failed
ensemble Kalman filter with static process noise applied to state estimation of zero-concentration components. Reason: constant noise covariance induced artificial fluctuations when true physical state was strictly zero
Advancing bioprocessing with ensemble kalman filter: from state estimation to knowledge transfer · Imperial
Considered and rejected
Considered and rejected: Rejected recursive data assimilation / iterative Ensemble Kalman Filtering in EnsCGP to avoid over-conditioning, bias accumulation, and ensemble collapse.
Physics-based and data-driven inversion of magnetotelluric data for subsurface imaging · Woods Hole
Sparse observations and unobservability cause unbounded error growth
Long measurement propagation intervals, high bearing noise, and sensor occlusions prevent filters from maintaining accurate state corrections. In situations with unobservable states or excluded sensory modalities, estimation errors accumulate without bound and cause filter divergence.
Tried and failed
extended Kalman filter with increased measurement noise covariance applied to indirect wind velocity estimation. Outcome: did not converge. Reason: state estimates diverged when measurement noise covariance was increased
An Invariant Extended Kalman Filter for Indirect Wind Estimation Using a Small, Fixed-Wing Uncrewed Aerial Vehicle · Virginia Tech
Tried and failed
unscented Kalman filter applied to orbit determination with sparse measurements. Outcome: did not converge. Reason: long propagation gaps between measurement updates caused divergence under nonlinear dynamics
Sequential Monte Carlo filtering with Gaussian mixture models for highly nonlinear systems · UT Austin
Tried and failed
Extended Kalman filter simultaneous localization and mapping applied to multi-agent fused directional sensing. Outcome: did not converge. Reason: Excessive bearing measurement noise and wide sensor opening angles degrade state estimator convergence
Dynamics of Multi-Agent Systems with Bio-Inspired Active and Passive Sensing · Virginia Tech
Tried and failed
Extended Kalman filter using sparse line-of-sight ranging applied to spacecraft orbit determination. Outcome: did not converge. Reason: measurement visibility below twenty percent caused unbounded filter divergence and extreme state estimation errors
Lunar Laser Ranging for Autonomous Cislunar Spacecraft Navigation · Virginia Tech
Tried and failed
sensor exclusion during Kalman filter state estimation applied to satellite attitude and torque estimation. Outcome: unstable. Reason: loss of absolute attitude measurements causes immediate filter divergence
Modeling and Analysis of a Thermospheric Density Measurement System Based on Torque Estimation · Virginia Tech
Tried and failed
Kalman-filter-based motion tracking without appearance features applied to multi-object tracking under occlusions. Reason: Kalman filter error accumulates during occlusions, preventing re-identification upon target reappearance
Intelligent perception frameworks and algorithms for social good: food security and public safety · Iowa State
Tried and failed
retaining cross-correlation covariance terms in EKF applied to vision-based spacecraft state estimation. Outcome: worse than baseline. Reason: cross-correlation terms produced less consistent state estimates and degraded overall filter performance
Tried and failed
reducing sensor noise parameter below model accuracy threshold applied to Kalman filter state estimation. Outcome: worse than baseline. Reason: underlying reference model resolution limits dominated error when sensor noise was over-optimistically tuned
Lunar Surface Navigation Using Gravity and Star Tracker Measurements · Virginia Tech
Considered and rejected
Considered and rejected: Rejected standard Kalman filtering for post-intervention Stage II MoodZoom data due to high missingness (29%), using time-adjusted RMSSD (tRMSSD) instead
Tried and failed
extended and unscented Kalman filtering applied to nonlinear spacecraft relative navigation. Outcome: did not converge. Reason: point-wise unobservability and multimodal non-Gaussian posterior state distributions
Violations of physical bounds and manifold constraints degrade performance
Additive Kalman filter updates violate non-Euclidean manifold properties, causing quaternion norm drift and singular covariance matrices. Furthermore, unconstrained filtering variants generate infeasible states across contact boundaries and physical admissibility limits, destabilizing downstream models.
Considered and rejected
Considered and rejected: Additive Extended Kalman Filter (AEKF) was rejected because quaternion norm constraints are violated, covariance singularities occur, and error quaternions cannot form valid covariance representations.
Attitude Determination using Asynchronous MultiSensor Fusion · YorkSpace
Considered and rejected
Considered and rejected: Rejected standard additive Kalman filtering for quaternion states due to non-unit norm drift and renormalization inaccuracies, adopting MEKF error vectors instead
Modeling and Analysis of a Thermospheric Density Measurement System Based on Torque Estimation · Virginia Tech
Considered and rejected
Considered and rejected: Rejected model reduction for enforcing Kalman filter state constraints due to loss of physical state meaning and lack of general tractability across complex or time-varying constraints.
Tried and failed
unscented Kalman filter applied to systems with contact constraints. Reason: sampled sigma points fell into infeasible regions, biasing state estimates away from the contact manifold
Estimation and Planning for Dynamic Robot Behaviors · Harvard
Tried and failed
unscented quaternion state estimation filter applied to rigid body attitude tracking with bias. Outcome: did not converge. Reason: residual sensor biases and disturbance torques prevented convergence to high accuracy
Tried and failed
unconstrained extended Kalman filter with geometric error measurements applied to vehicle heading and pose estimation. Outcome: unstable. Reason: large initial heading errors caused non-unique cross-track error solutions leading to continuous heading oscillations
Autonomous Vehicle Pose Estimation in GNSS-Denied Areas Using Cross-Track Error Measurements · Virginia Tech
Tried and failed
unconstrained Kalman filtering variants applied to nonlinear physical state estimation. Outcome: unstable. Reason: estimates violated physical admissibility bounds, causing numerical instability and process model crashes
On traffic state estimation and control in the world of connected vehicles · UT Austin
Tried and failed
thresholded Kalman filter for intent disambiguation applied to human-robot physical interaction state estimation. Reason: could not distinguish passive mechanical compliance from active voluntary motion, causing false negative cooperativeness spikes
Modeling the Sit-to-Stand Transition using Koopman Lifting Linearization and Human State Estimation · MIT
Left open by the authors
Problems the authors named and did not get to.
Left open
Develop a principled residual uncertainty tracking method for low-rank Kalman filters to prevent overconfidence without heuristic covariance inflation. Blocker: Lacks a concrete mathematical formulation or specific algorithmic mechanism for tracking truncated residual uncertainty.
Probabilistic Inference for Spatiotemporal Dynamics · Publikationssystem UB Tuebingen
Left open
Extend the Maximum Correntropy Criterion Extended Kalman Filter to secure distributed vehicle-to-vehicle state estimation in multi-robot systems. Blocker: None
Correntropy: Answer to non-Gaussian noise in modern SLAM applications? · unevada
Left open
Integrate an Extended Kalman Filter with realistic navigation sensor models into the docking MPC simulation to reduce trajectory bouncing. Blocker: None
Application of model predictive control for the autonomous rendezvous and docking of small satellites · Georgia Tech
Left open
Implement bias-aware sequential Kalman filters to simultaneously estimate orbital state and systematic along-track biases from deep space TLE pseudo-observations. Blocker: None
Left open
Implement an Extended Kalman Filter using a CTRV motion model for tracklet reconnection in subviral particle tracking. Blocker: None
Motion patterns of subviral particles: Digital tracking, image data processing and analysis Bewegungsmuster subviraler Partikel: Digitales Tracking, Bilddatenverarbeitung und -anal · open_UMR Marburg DSpace 10.0
Left open
Develop multifidelity square root filters, specifically extending the linear control variates ensemble Kalman filtering framework to a multifidelity LETKF. Blocker: None
Combining Data-driven and Theory-guided Models in Ensemble Data Assimilation · Virginia Tech
Left open
Derive theoretical convergence rates for extended Kalman, unscented Kalman, and particle ODE filters. Blocker: None
Uncertainty-Aware Numerical Solutions of ODEs by Bayesian Filtering · Publikationssystem UB Tuebingen
Left open
Estimate full posterior state distribution uncertainty beyond Gaussian linear approximations for multi-modal vision-based state estimation. Blocker: No concrete approach, target benchmark, or formulation specified for non-Gaussian posterior estimation
Methods for Vision-Based State Estimation and Online Motion Model Adaptation Using Multi-Modal Measurements and Motion Constraints · Publikationssystem UB Tuebingen
Left open
Incorporate multimodal belief distributions over contact modes into contact-constrained Kalman filtering to prevent filter divergence in simulation. Blocker: None
Estimation and Planning for Dynamic Robot Behaviors · Harvard
Left open
Incorporate Kalman filtering or Bayesian tracking into the PEB-annealed 3D UAV path planner to combine noisy localization measurements over time. Blocker: None
Management and Analysis of Localization Information in Uncrewed Aerial Systems · Virginia Tech
Checking a claim in this area?
We can run the same search on any method or claim. If nothing turns up, we will say so, and that proves nothing on its own.