Browsing by Subject "Kalman filtering"

Now showing 1 - 6 of 6

Advanced navigation algorithms for precision landing
(2007-12) Zanetti, Renato, 1978-; Bishop, Robert H., 1957-
A detailed analysis of autonomous navigation algorithms to achieve autonomous precision landing is presented. The problem of integrated attitude determination and inertial navigation is solved. The theoretical results are applied and tested in three different applications. Optimality conditions for constrained quaternion estimation using the Kalman filter are derived. It is common in spacecraft applications to separate the attitude determination from the inertial navigation system. While this approach has worked in the past, it inevitably degrades the navigation performance when the correlations between the two systems are not correctly accounted for. It is shown how to optimally include an attitude determination algorithm into the Kalman filter. When the conditions to achieve optimality are not met, it is shown how to achieve sub-optimality by properly accounting for the correlation. The traditional approach to inertial navigation is to employ the inertial measurement unit (IMU) outputs to propagate the estimated states forward in time, rather then use them to update the state. A detailed covariance analysis of deadreckoning Mars entry navigation is performed. The contribution of various sources of IMU errors are explicitly accounted for and the filter performance is validated through Monte Carlo analysis. The drawback of dead-reckoning is that this approach prevents the inertial measurements from reducing the uncertainty of the estimated states. While this shortcoming can be compensated by the availability of other measurements, it becomes crucial when the IMU is the only sensor to provide measurements. Such a situation arises, for example, during Mars atmospheric entry. In the second application of this work, IMU measurements from a NASA mission are processed in an extended Kalman filter, and the results are compared to dead-reckoning. It is shown that is possible to reduce the uncertainty of the inertial states by filtering the IMU. The final application is lunar descent to landing navigation. In this example the IMU is filtered and the algorithms to include an attitude estimate into the Kalman filter are tested. The design performance is confirmed by Monte Carlo analysis.
Development of multisensor fusion techniques with gating networks applied to reentry vehicles
(2003) Dubois-Matra, Olivier; Bishop, Robert H., 1957-
The problem of model inaccuracy for Extended Kalman Filters (EKF) is addressed in the case of vehicle atmospheric entry tracking and navigation with a filter bank architecture, also called mixture-of-experts, regulated by gating network, which is then tested in two different applications. First, a wind-frame based flight model is developed, which allows for maneuvers, and inclusion of atmospheric and gravity models. This level of complexity allows in theory for better estimation accuracy when used in an EKF, but the filter performance is in part dependent on the accuracy of the vehicle and environment models. The problem is how to deal with imperfect models. The approach treated here, which as already been applied in other domains, is to create a population of filters, each representing a particular modeling of the vehicle and/or environment. The discriminating device between the expert filters is a gating network, which is a simplified single-layer neural network learning in real-time with the help of the statistical information from the filters. The gating network is used to compute a weighted sum of the state estimate from each filter, which is therefore an optimal estimate. The gating network can also be used as an hypothesis tester, which is the case in the first example. The system was applied to the tracking and identification at high altitude of reentering spiraling objects accompanied by decoys. The object is being tracked at high altitude by three ground radars providing a variety of measurements which are treated in parallel by two filters, one being an expert tuned for the real target and the other tuned for the decoy. Experiments show that the regulated bank can rapidly correctly identify the object as being the real target. The second application is precision Mars entry navigation, where the on-board navigation system of a maneuvering Mars lander used a bank of expert EKF, each processing inertial acceleration as measurement, and each designed around a specific realization of the imperfectly known atmospheric density profile. The objective here is less to identify the best performing model than optimizing the overall state estimate by combining the estimate from every filter. The system also periodically restarts the filters with the current optimal estimate so as to keep all the filters competitive during all of the descent. The result is that this mixture-of-experts does not perform better than a dead-reckoning scheme unless one of the density model happens to be relatively close from the real density profile, but that it is more robust than dead-reckoning to loss of data, and can readily adapt additional sources of measurements.
An ensemble Kalman filter module for automatic history matching
(2007-12) Liang, Baosheng, 1979-; Sepehrnoori, Kamy, 1951-
The data assimilation process of adjusting variables in a reservoir simulation model to honor observations of field data is known as history matching and has been extensively studied for few decades. However, limited success has been achieved due to the high complexity of the problem and the large computational effort required by the practical applications. An automatic history matching module based on the ensemble Kalman filter is developed and validated in this dissertation. The ensemble Kalman filter has three steps: initial sampling, forecasting through a reservoir simulator, and assimilation. The initial random sampling is improved by the singular value decomposition, which properly selects the ensemble members with less dependence. In this way, the same level of accuracy is achieved through a smaller ensemble size. Four different schemes for the assimilation step are investigated and direct inverse and square root approaches are recommended. A modified ensemble Kalman filter algorithm, which addresses the preference to the ensemble members through a nonequally weighting factor, is proposed. This weighted ensemble Kalman filter generates better production matches and recovery forecasting than those from the conventional ensemble Kalman filter. The proposed method also has faster convergence at the early time period of history matching. Another variant, the singular evolutive interpolated Kalman filter, is also applied. The resampling step in this method appears to improve the filter stability and help the filter to deliver rapid convergence both in model and data domains. This method and the ensemble Kalman filter are effective for history matching and forecasting uncertainty quantification. The independence of the ensemble members during the forecasting step allows the benefit of high-performance computing for the ensemble Kalman filter implementation during automatic history matching. Two-level computation is adopted; distributing ensemble members simultaneously while simulating each member in a parallel style. Such computation yields a significant speedup. The developed module is integrated with reservoir simulators UTCHEM, GEM and ECLIPSE, and has been implemented in the framework Integrated Reservoir Simulation Platform (IRSP). The successful applications to two and three-dimensional cases using blackoil and compositional reservoir cases demonstrate the efficiency of the developed automatic history matching module.
Introducing principled approximation and online control into streaming applications
(2021-08) Pei, Yan, Ph. D.; Pingali, Keshav; Fussell, Donald; Stone, Peter; Ding, Ke
The ubiquity of streaming applications in important domains such as deep learning, computer vision/graphics, Internet of Things has opened up opportunities for the use of approximate computing to enable efficient execution of these applications on a wide range of platforms. This dissertation explores the use of ideas from machine learning and control theory to exploit approximation in streaming applications in a principled way. We first present online control techniques of introducing principled approximation into Simultaneous Localization and Mapping (SLAM) algorithms, which are used in emerging domains like robotics and autonomous driving in which autonomous agents build a map while navigating through unknown environments. Existing studies of approximation in SLAM have mostly used offline control, assuming the trajectory is known before the agent starts to move, which is impractical. The proposed methodology controls approximation in an adaptive manner without causing unacceptable quality degradation, enabling efficient deployment of SLAM on a wider range of resource-constrained platforms. We also propose Sonic, a sampling-based online controller for general constrained optimization problems in long-running streaming applications. Within each phase of a streaming application’s execution, Sonic utilizes the beginning portion to sample the knob space sequentially and aims to pick the optimal knob setting for the rest of the phase, given a user-specified constrained optimization problem. Machine learning regressors and Bayesian optimization are applied in Sonic for better sampling choices. Our experiments show that Sonic is able to find near-optimal knob settings at run time for the applications we studied. For online control, state estimation is a key component. In this dissertation, we introduce a novel derivation of Kalman filtering, a classic state estimation technique that can be used in online control to combine noisy estimates of quantity of interest. It is presented from an abstract perspective with key assumptions and concepts clarified. We then exploit these insights and propose RASR, a performance-oriented super-resolution program for rendered content that takes advantage of internal sub-pixel states of graphics hardware. RASR is a deep learning approach to fusing frames of low-fidelity into one highfidelity frame, designed to meet the increasing demand for high throughput, high quality and low latency in real-time rendering.
Navigation algorithms and observability analysis for formation flying missions
(2006) Huxel, Paul John; Bishop, Robert H., 1957-
Navigation algorithms and the corresponding observability analysis for formation flying missions are developed. The methodology of the observability analysis relates the physical geometry of the observers, as well as the spacecraft formation, to several measures of system observability. Relationships between these observability measures and the state error covariance are then derived to provided estimated bounds or forecasts for the expected navigation accuracy. These methods range from conservative time-invariant analytic bounds to more representative numerical forecasts using common dilution of precision metrics. The research also examines the robustness of the extended Kalman filter when simultaneously processing inertial and relative range measurements. It has been shown that processing relative range measurements in conjunction with inertial range measurements can directly increase the accuracy of the inertial state estimate. However, it has also been shown that when there is relatively large uncertainty in the state estimate the addition of relative measurements can cause an otherwise convergent filter to diverge. This dissertation considers several methods for preventing this divergence, as well as an in-depth examination of second-order terms to explain the basis of the problem. In particular, to illustrate their potential significance, analytical bounds are derived for the second-order terms.
Spacecraft precision entry navigation using an adaptive sigma point Kalman filter bank
(2007) Heyne, Martin Cornelius, 1973-; Bishop, Robert H., 1957-
This work documents the development of a sigma point Kalman filter for the purpose of precision spacecraft navigation during the atmospheric entry, descent and landing phase. The use of the sigma point Kalman filter is driven by the desire to avoid complex partial derivatives associated with the standard extended Kalman filter. The strategy increases the likelihood that the navigation algorithm will be compatible with the Electra. Using Mars Exploration Rover Spirit (MER-A) and the Mars Science Laboratory (MSL) data, experiments were conducted to validate the proposed navigation concept. Beginning at atmospheric entry interface, the hypersonic entry phase is considered and the navigation architecture performance is quantified. Using the sigma point Kalman filter as the main computational unit, a filter bank for environmental parameter identification is investigated. The focus of the investigation is atmospheric parameter identification. The MERA mission is used to verify the ability of the filter bank to make appropriate selections. The navigation architecture is implemented on the Electra programmable radio, a flight hardware communication node available on spacecraft build for Mars exploration. The investigations show that the sigma point Kalman filter structure is very applicable to the atmospheric entry navigation problem. When used in conjunction with the filter bank concept, the overall navigation architecture is shown to be able to improve navigation accuracy over standard dead-reckoning, while providing robustness to uncertainties in the atmosphere. The navigation algorithm is successfully hosted on the Electra programmable radio and is capable of processing actual MER inertial measurement data.