The equivalent-weights particle filter in a high-dimensional system

In general, particle filters need large numbers of model runs in order to avoid filter degeneracy in high-dimensional systems. The recently proposed, fully nonlinear equivalent-weights particle filter overcomes this requirement by replacing the standard model transition density with two different proposal transition densities. The first proposal density is used to relax all particles towards the high-probability regions of state space as defined by the observations. The crucial second proposal density is then used to ensure that the majority of particles have equivalent weights at observation time. Here, the performance of the scheme in a high, 65 500 dimensional, simplified ocean model is explored. The success of the equivalent-weights particle filter in matching the true model state is shown using the mean of just 32 particles in twin experiments. It is of particular significance that this remains true even as the number and spatial variability of the observations are changed. The results from rank histograms are less easy to interpret and can be influenced considerably by the parameter values used. This article also explores the sensitivity of the performance of the scheme to the chosen parameter values and the effect of using different model error parameters in the truth compared with the ensemble model runs.

Twin experiments with the equivalent weights particle filter and HadCM3

This paper investigates the use of a particle filter for data assimilation with a full scale coupled ocean–atmosphere general circulation model. Synthetic twin experiments are performed to assess the performance of the equivalent weights filter in such a high-dimensional system. Artificial 2-dimensional sea surface temperature fields are used as observational data every day. Results are presented for different values of the free parameters in the method. Measures of the performance of the filter are root mean square errors, trajectories of individual variables in the model and rank histograms. Filter degeneracy is not observed and the performance of the filter is shown to depend on the ability to keep maximum spread in the ensemble.

A Variance-Minimizing Filter for Large-Scale Applications

A truly variance-minimizing filter is introduced and its per for mance is demonstrated with the Korteweg– DeV ries (KdV) equation and with a multilayer quasigeostrophic model of the ocean area around South Africa. It is recalled that Kalman-like filters are not variance minimizing for nonlinear model dynamics and that four - dimensional variational data assimilation (4DV AR)-like methods relying on per fect model dynamics have dif- ficulty with providing error estimates. The new method does not have these drawbacks. In fact, it combines advantages from both methods in that it does provide error estimates while automatically having balanced states after analysis, without extra computations. It is based on ensemble or Monte Carlo integrations to simulate the probability density of the model evolution. When obser vations are available, the so-called importance resampling algorithm is applied. From Bayes’ s theorem it follows that each ensemble member receives a new weight dependent on its ‘ ‘distance’ ’ t o the obser vations. Because the weights are strongly var ying, a resampling of the ensemble is necessar y. This resampling is done such that members with high weights are duplicated according to their weights, while low-weight members are largely ignored. In passing, it is noted that data assimilation is not an inverse problem by nature, although it can be for mulated that way . Also, it is shown that the posterior variance can be larger than the prior if the usual Gaussian framework is set aside. However , i n the examples presented here, the entropy of the probability densities is decreasing. The application to the ocean area around South Africa, gover ned by strongly nonlinear dynamics, shows that the method is working satisfactorily . The strong and weak points of the method are discussed and possible improvements are proposed.

Analysis Scheme in the Ensemble Kalman Filter

This paper discusses an important issue related to the implementation and interpretation of the analysis scheme in the ensemble Kalman filter . I t i s shown that the obser vations must be treated as random variables at the analysis steps. That is, one should add random perturbations with the correct statistics to the obser vations and generate an ensemble of obser vations that then is used in updating the ensemble of model states. T raditionally , this has not been done in previous applications of the ensemble Kalman filter and, as will be shown, this has resulted in an updated ensemble with a variance that is too low . This simple modification of the analysis scheme results in a completely consistent approach if the covariance of the ensemble of model states is interpreted as the prediction error covariance, and there are no further requirements on the ensemble Kalman filter method, except for the use of an ensemble of sufficient size. Thus, there is a unique correspondence between the error statistics from the ensemble Kalman filter and the standard Kalman filter approach

Implicit equal-weights particle filter

Filter degeneracy is the main obstacle for the implementation of particle filter in non-linear high-dimensional models. A new scheme, the implicit equal-weights particle filter (IEWPF), is introduced. In this scheme samples are drawn implicitly from proposal densities with a different covariance for each particle, such that all particle weights are equal by construction. We test and explore the properties of the new scheme using a 1,000-dimensional simple linear model, and the 1,000-dimensional non-linear Lorenz96 model, and compare the performance of the scheme to a Local Ensemble Kalman Filter. The experiments show that the new scheme can easily be implemented in high-dimensional systems and is never degenerate, with good convergence properties in both systems.

A Floating-point Extended Kalman Filter Implementation for Autonomous Mobile Robots

Localization and Mapping are two of the most important capabilities for autonomous mobile robots and have been receiving considerable attention from the scientific computing community over the last 10 years. One of the most efficient methods to address these problems is based on the use of the Extended Kalman Filter (EKF). The EKF simultaneously estimates a model of the environment (map) and the position of the robot based on odometric and exteroceptive sensor information. As this algorithm demands a considerable amount of computation, it is usually executed on high end PCs coupled to the robot. In this work we present an FPGA-based architecture for the EKF algorithm that is capable of processing two-dimensional maps containing up to 1.8 k features at real time (14 Hz), a three-fold improvement over a Pentium M 1.6 GHz, and a 13-fold improvement over an ARM920T 200 MHz. The proposed architecture also consumes only 1.3% of the Pentium and 12.3% of the ARM energy per feature.

Localization of a mobile robot based on odometry and natural landmarks using extended kalman filter

SANTANA, André M.; SOUZA, Anderson A. S.; BRITTO, Ricardo S.; ALSINA, Pablo J.; MEDEIROS, Adelardo A. D. Localization of a mobile robot based on odometry and natural landmarks using extended Kalman Filter. In: INTERNATIONAL CONFERENCE ON INFORMATICS IN CONTROL, AUTOMATION AND ROBOTICS, 5., 2008, Funchal, Portugal. Proceedings... Funchal, Portugal: ICINCO, 2008.

A Discussion Related to Orbit Determination Using Nonlinear Sigma Point Kalman Filter

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Unscented Kalman Filter Applied to the Spacecraft Attitude Estimation with Euler Angles

The aim of this work is to test an algorithm to estimate, in real time, the attitude of an artificial satellite using real data supplied by attitude sensors that are on board of the CBERS-2 satellite (China Brazil Earth Resources Satellite). The real-time estimator used in this work for attitude determination is the Unscented Kalman Filter. This filter is a new alternative to the extended Kalman filter usually applied to the estimation and control problems of attitude and orbit. This algorithm is capable of carrying out estimation of the states of nonlinear systems, without the necessity of linearization of the nonlinear functions present in the model. This estimation is possible due to a transformation that generates a set of vectors that, suffering a nonlinear transformation, preserves the same mean and covariance of the random variables before the transformation. The performance will be evaluated and analyzed through the comparison between the Unscented Kalman filter and the extended Kalman filter results, by using real onboard data.

Identification of Mechanical Systems through Kautz Filter

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Comparative analysis of synchronization algorithms based on PLL, RDFT and Kalman Filter

This paper discusses the main characteristics and presents a comparative analysis of three synchronization algorithms based respectively, on a Phase-Locked Loop, a Kalman Filter and a Discrete Fourier Transform. It will be described the single and three-phase models of the first two methods and the single-phase model of the third one. Details on how to modify the filtering properties or dynamic response of each algorithm will be discussed in terms of their design parameters. In order to compare the different algorithms, these parameters will be set for maximum filter capability. Then, the dynamic response, during input amplitude and frequency deviations will be observed, as well as during the initialization procedure. So, advantages and disadvantages of all considered algorithms will be discussed. ©2007 IEEE.

Efficient Kernel Computation for Volterra Filter Structure Evaluation

Despite their generality, conventional Volterra filters are inadequate for some applications, due to the huge number of parameters that may be needed for accurate modelling. When a state-space model of the target system is known, this can be assessed by computing its kernels, which also provides valuable information for choosing an adequate alternate Volterra filter structure, if necessary, and is useful for validating parameter estimation procedures. In this letter, we derive expressions for the kernels by using the Carleman bilinearization method, for which an efficient algorithm is given. Simulation results are presented, which confirm the usefulness of the proposed approach.

Synthesis Methodology Applied to a Tunable Patch Filter With Independent Frequency and Bandwidth Control

A new methodology for the synthesis of tunable patch filters is presented. The methodology helps the designer to perform a theoretical analysis of the filter through a coupling matrix that includes the effect of the tuning elements used to tune the filter. This general methodology accounts for any tuning parameter desired and was applied to the design of a tunable dual-mode patch filter with independent control of center frequency and bandwidth (BW). The bandpass filter uses a single triangular resonator with two etched slots that split the fundamental degenerate modes and form the filter passband. Varactor diodes assembled across the slots are used to vary the frequency of each degenerate fundamental mode independently, which is feasible due to the nature of the coupling scheme of the filter. The varactor diode model used in simulations, their assembling, the dc bias configuration, and measured results are presented. The theory results are compared to the simulations and to measurements showing a very good agreement and validating the proposed methodology. The fabricated filter presents an elliptic response with 20% of center frequency tuning range around 3.2 GHz and a fractional BW variation from 4% to 12% with low insertion loss and high power handling with a 1-dB compression point higher than +14.5 dB.

Modeling of diesel particulate filter filtration and regeneration for transient driving schedules

Due to their high thermal efficiency, diesel engines have excellent fuel economy and have been widely used as a power source for many vehicles. Diesel engines emit less greenhouse gases (carbon dioxide) compared with gasoline engines. However, diesel engines emit large amounts of particulate matter (PM) which can imperil human health. The best way to reduce the particulate matter is by using the Diesel Particulate Filter (DPF) system which consists of a wall-flow monolith which can trap particulates, and the DPF can be periodically regenerated to remove the collected particulates. The estimation of the PM mass accumulated in the DPF and total pressure drop across the filter are very important in order to determine when to carry out the active regeneration for the DPF. In this project, by developing a filtration model and a pressure drop model, we can estimate the PM mass and the total pressure drop, then, these two models can be linked with a regeneration model which has been developed previously to predict when to regenerate the filter. There results of this project were: 1 Reproduce a filtration model and simulate the processes of filtration. By studying the deep bed filtration and cake filtration, stages and quantity of mass accumulated in the DPF can be estimated. It was found that the filtration efficiency increases faster during the deep-bed filtration than that during the cake filtration. A “unit collector” theory was used in our filtration model which can explain the mechanism of the filtration very well. 2 Perform a parametric study on the pressure drop model for changes in engine exhaust flow rate, deposit layer thickness, and inlet temperature. It was found that there are five primary variables impacting the pressure drop in the DPF which are temperature gradient along the channel, deposit layer thickness, deposit layer permeability, wall thickness, and wall permeability. 3 Link the filtration model and the pressure drop model with the regeneration model to determine the time to carry out the regeneration of the DPF. It was found that the regeneration should be initiated when the cake layer is at a certain thickness, since a cake layer with either too big or too small an amount of particulates will need more thermal energy to reach a higher regeneration efficiency. 4 Formulate diesel particulate trap regeneration strategies for real world driving conditions to find out the best desirable conditions for DPF regeneration. It was found that the regeneration should be initiated when the vehicle’s speed is high and during which there should not be any stops from the vehicle. Moreover, the regeneration duration is about 120 seconds and the inlet temperature for the regeneration is 710K.

Experimental and modeling study of the filtration and oxidation characteristics of a diesel oxidation catalyst and a catalyzed particulate filter

A diesel oxidation catalyst (DOC) with a catalyzed diesel particulate filter (CPF) is an effective exhaust aftertreatment device that reduces particulate emissions from diesel engines, and properly designed DOC-CPF systems provide passive regeneration of the filter by the oxidation of PM via thermal and NO2/temperature-assisted means under various vehicle duty cycles. However, controlling the backpressure on engines caused by the addition of the CPF to the exhaust system requires a good understanding of the filtration and oxidation processes taking place inside the filter as the deposition and oxidation of solid particulate matter (PM) change as functions of loading time. In order to understand the solid PM loading characteristics in the CPF, an experimental and modeling study was conducted using emissions data measured from the exhaust of a John Deere 6.8 liter, turbocharged and after-cooled engine with a low-pressure loop EGR system and a DOC-CPF system (or a CCRT® - Catalyzed Continuously Regenerating Trap®, as named by Johnson Matthey) in the exhaust system. A series of experiments were conducted to evaluate the performance of the DOC-only, CPF-only and DOC-CPF configurations at two engine speeds (2200 and 1650 rpm) and various loads on the engine ranging from 5 to 100% of maximum torque at both speeds. Pressure drop across the DOC and CPF, mass deposited in the CPF at the end of loading, upstream and downstream gaseous and particulate emissions, and particle size distributions were measured at different times during the experiments to characterize the pressure drop and filtration efficiency of the DOCCPF system as functions of loading time. Pressure drop characteristics measured experimentally across the DOC-CPF system showed a distinct deep-bed filtration region characterized by a non-linear pressure drop rise, followed by a transition region, and then by a cake-filtration region with steadily increasing pressure drop with loading time at engine load cases with CPF inlet temperatures less than 325 °C. At the engine load cases with CPF inlet temperatures greater than 360 °C, the deep-bed filtration region had a steep rise in pressure drop followed by a decrease in pressure drop (due to wall PM oxidation) in the cake filtration region. Filtration efficiencies observed during PM cake filtration were greater than 90% in all engine load cases. Two computer models, i.e., the MTU 1-D DOC model and the MTU 1-D 2-layer CPF model were developed and/or improved from existing models as part of this research and calibrated using the data obtained from these experiments. The 1-D DOC model employs a three-way catalytic reaction scheme for CO, HC and NO oxidation, and is used to predict CO, HC, NO and NO2 concentrations downstream of the DOC. Calibration results from the 1-D DOC model to experimental data at 2200 and 1650 rpm are presented. The 1-D 2-layer CPF model uses a ‘2-filters in series approach’ for filtration, PM deposition and oxidation in the PM cake and substrate wall via thermal (O2) and NO2/temperature-assisted mechanisms, and production of NO2 as the exhaust gas mixture passes through the CPF catalyst washcoat. Calibration results from the 1-D 2-layer CPF model to experimental data at 2200 rpm are presented. Comparisons of filtration and oxidation behavior of the CPF at sample load-cases in both configurations are also presented. The input parameters and selected results are also compared with a similar research work with an earlier version of the CCRT®, to compare and explain differences in the fundamental behavior of the CCRT® used in these two research studies. An analysis of the results from the calibrated CPF model suggests that pressure drop across the CPF depends mainly on PM loading and oxidation in the substrate wall, and also that the substrate wall initiates PM filtration and helps in forming a PM cake layer on the wall. After formation of the PM cake layer of about 1-2 µm on the wall, the PM cake becomes the primary filter and performs 98-99% of PM filtration. In all load cases, most of PM mass deposited was in the PM cake layer, and PM oxidation in the PM cake layer accounted for 95-99% of total PM mass oxidized during loading. Overall PM oxidation efficiency of the DOC-CPF device increased with increasing CPF inlet temperatures and NO2 flow rates, and was higher in the CCRT® configuration compared to the CPF-only configuration due to higher CPF inlet NO2 concentrations. Filtration efficiencies greater than 90% were observed within 90-100 minutes of loading time (starting with a clean filter) in all load cases, due to the fact that the PM cake on the substrate wall forms a very efficient filter. A good strategy for maintaining high filtration efficiency and low pressure drop of the device while performing active regeneration would be to clean the PM cake filter partially (i.e., by retaining a cake layer of 1-2 µm thickness on the substrate wall) and to completely oxidize the PM deposited in the substrate wall. The data presented support this strategy.