No description available.

No description available.

Dynamic pricing is a difficult problem for machine learning. The environment is noisy, dynamic and has a measurable cost associated with exploration that necessitates that learning be done in short-time horizons. These short-time horizons force the learning algorithms to make pricing decisions based on scarce data. In this work, various machine learning algorithms are compared in the context of dynamic pricing. These algorithms include the Kalman filter, artificial neural networks, particle swarm optimization and genetic algorithms. The majority of these algorithms have been modified to handle the pricing problem. The results show that these adaptations allow the learning algorithms to handle the noisy dynamic conditions and to learn quickly.

Aircraft in the US are equipped with Emergency Locator Transmitters (ELTs). In emergency situations these beacons are activated, providing a radio signal that can be used to locate the aircraft. Recent developments in UAV technologies have enabled mini-UAVs (5-foot wingspan) to possess a high level of autonomy. Due to the small size of these aircraft they are human-packable and can be easily transported and deployed in the field. Using a custom-built Radio Direction Finder, we gathered readings from a known transmitter and used them to compare various Bayesian reasoning-based filtering algorithms. Using a custom-developed simulator, we were able to test and evaluate filtering and control methods. In most non-trivial conditions we found that the Sequential Importance Resampling (SIR) Particle Filter worked best. The filtering and control algorithms presented can be extended to other problems that involve UAV control and tracking with noisy non-linear sensor behavior.

UAVs have become an essential tool in many market segments, particularly the military where critical intelligence can be gathered by them. A tailsitter aircraft is a platform whose purpose is to efficiently merge the range and endurance of fixed-wing aircraft with the VTOL capabilities of rotorcraft and is of significant value in applications where launch and recovery area is limited or the use of launch and recovery equipment is not desirable. Developing autopilot software for a tailsitter UAV is unique in that the aircraft must be autonomously controlled over a much wider range of attitudes than conventional UAVs. Assumptions made in conventional estimation and control algorithms are not valid for tailsitter aircraft because of routine operation around gimbal lock. Quaternions are generally employed to overcome the limitations Euler angles; however, adapting the attitude representation to work at a full range of attitudes is only part of the solution. Kalman filter measurement updates and control algorithms must also work at any orientation. This research presents several methods of incorporating a magnetometer measurement into an extended Kalman filter. One method combines magnetometer and accelerometer sensor data using the solution to Wahba's problem to calculate an overall attitude measurement. Other methods correct only heading error and include using two sets of Euler angles to update the estimate, using quaternions to determine heading error and Euler angles to update the estimate, and using only quaternions to update the estimate. Quaternion feedback attitude control is widely used in tailsitter aircraft. This research also shows that in spite of its effective use in spacecraft, using the attitude error calculated via quaternions to drive flight control surfaces may not be optimal for tailsitters. It is shown that during hover when heading error is present, quaternion feedback can cause undesired behavior, particularly when the heading error is large. An alternative method for calculating attitude error called resolved tilt-twist is validated, improved, and shown to perform better than quaternion feedback. Algorithms are implemented on a commercially available autopilot and validation is performed using hardware in loop simulation. A custom interface is used to receive autopilot commands and send the autopilot simulated sensor information. The final topic covered deals with the tailsitter hovering in wind. As the tailsitter hovers, wind can cause the tailsitter to turn such that the wind is perpendicular to the wings. Wind tunnel data is taken and analyzed to explain this behavior.

Prior to the wide deployment of robotic systems, they must be able to navigate autonomously. These systems cannot rely on good weather or daytime navigation and they must also be able to navigate in unknown environments. All of this must take place without human interaction. A majority of modern autonomous systems rely on GPS for position estimation. While GPS solutions are readily available, GPS is often lost and may even be jammed. To this end, a significant amount of research has focused on GPS-denied navigation. Many GPS-denied solutions rely on known environmental features for navigation. Others use vision sensors, which often perform poorly at high altitudes and are limited in poor weather. In contrast, radar systems accurately measure range at high and low altitudes. Additionally, these systems remain unaffected by inclimate weather. This dissertation develops the use of radar odometry for GPS-denied navigation. Using the range progression of unknown environmental features, the aircraft's motion is estimated. Results are presented for both simulated and real radar data. In Chapter 2 a greedy radar odometry algorithm is presented. It uses the Hough transform to identify the range progression of ground point-scatterers. A global nearest neighbor approach is implemented to perform data association. Assuming a piece-wise constant heading assumption, as the aircraft passes pairs of scatterers, the location of the scatterers are triangulated, and the motion of the aircraft is estimated. Real flight data is used to validate the approach. Simulated flight data explores the robustness of the approach when the heading assumption is violated. Chapter 3 explores a more robust radar odometry technique, where the relatively constant heading assumption is removed. This chapter uses the recursive-random sample consensus (R-RANSAC) Algorithm to identify, associate, and track the point scatterers. Using the measured ranges to the tracked scatterers, an extended Kalman filter (EKF) iteratively estimates the aircraft's position in addition to the relative locations of each reflector. Real flight data is used to validate the accuracy of this approach. Chapter 4 performs observability analysis of a range-only sensor. An observable, radar odometry approach is proposed. It improves the previous approaches by adding a more robust R-RANSAC above ground level (AGL) tracking algorithm to further improve the navigational accuracy. Real flight results are presented, comparing this approach to the techniques presented in previous chapters.

To operate, autonomous systems of necessity employ a variety of sensors to perceive their environment. Many small unmanned aerial vehicles (UAV) are unable to carry redundant sensors due to size, weight, and power (SWaP) constraints. Faults in these sensors can cause undesired behavior, including system instability. Thus, detection of faults in these non-redundant sensors is of paramount importance.The problem of detecting sensor faults in non-redundant sensors on board autonomous aircraft is non-trivial. Factors that make development of a solution difficult include both an inability to perfectly characterize systems and sensors as well as the SWaP constraints inherent with small UAV. An additional challenge is the ability of a fault-detection method to strike a balance between false-alarm rate and detection rate.This thesis explores two model-based methods of fault-detection for non-redundant sensors, a Kalman filter based method and a particle filter based method. The Kalman filter based method employs tests of mean and covariance on the normalized innovation sequence to detect faults, while the particle filter based method uses a function of the average particle weights.The Kalman filter based approach was implemented in real time on board an autonomous rotorcraft using an extended Kalman Filter (EKF). Faults tested included varied levels of bias, drift, and increased noise. Metrics included false-alarm rate, detection rate, and delay to detection. The particle filter based approach was implemented on a simulated system. This was then compared with an implementation of the EKF based approach for the same system. The same fault types and metrics were also used for these tests.The EKF based method of fault-detection performed well onboard the autonomous rotorcraft and should be generalizable to other systems for which an EKF or Kalman filter can be implemented. The theory indicates that the particle filter based algorithm should have performed better, though the simulations showed poor detection characteristics in comparison to the Kalman filter based method. Future work should be performed to explore improvements to the particle filter based method.

The current and future trends of modern wireless communication systems place heavy demands on fast data transmissions in order to satisfy end users’ requirements anytime, anywhere. Such demands are obvious in recent applications such as smart phones, long term evolution (LTE), 4 & 5 Generations (4G & 5G), and worldwide interoperability for microwave access (WiMAX) platforms, where robust coding and modulations are essential especially in streaming on-line video material, social media and gaming. This eventually resulted in extreme exhaustion imposed on the frequency spectrum as a rare natural resource due to stagnation in current spectrum management policies. Since its advent in the late 1990s, cognitive radio (CR) has been conceived as an enabling technology aiming at the efficient utilisation of frequency spectrum that can lead to potential direct spectrum access (DSA) management. This is mainly attributed to its internal capabilities inherited from the concept of software defined radio (SDR) to sniff its surroundings, learn and adapt its operational parameters accordingly. CR systems (CRs) may commonly comprise one or all of the following core engines that characterise their architectures; namely, adaptive coded modulation (ACM), automatic modulation classification (AMC) and spectrum sensing (SS). Motivated by the above challenges, this programme of research is primarily aimed at the design and development of new paradigms to help improve the adaptability of CRs and thereby achieve the desirable signal processing tasks at the physical layer of the above core engines. Approximate modelling of Rayleigh and finite state Markov channels (FSMC) with a new concept borrowed from econometric studies have been approached. Then insightful channel estimation by using Kalman filter (KF) augmented with interacting multiple model (IMM) has been examined for the purpose of robust adaptability, which is applied for the first time in wireless communication systems. Such new IMM-KF combination has been facilitated in the feedback channel between wireless transmitter and receiver to adjust the transmitted power, by using a water-filling (WF) technique, and constellation pattern and rate in the ACM algorithm. The AMC has also benefited from such IMM-KF integration to boost the performance against conventional parametric estimation methods such as maximum likelihood estimate (MLE) for channel interrogation and the estimated parameters of both inserted into the ML classification algorithm. Expectation-maximisation (EM) has been applied to examine unknown transmitted modulation sequences and channel parameters in tandem. Finally, the non-parametric multitaper method (MTM) has been thoroughly examined for spectrum estimation (SE) and SS, by relying on Neyman-Pearson (NP) detection principle for hypothesis test, to allow licensed primary users (PUs) to coexist with opportunistic unlicensed secondary users (SUs) in the same frequency bands of interest without harmful effects. The performance of the above newly suggested paradigms have been simulated and assessed under various transmission settings and revealed substantial improvements.

[pt] Neste trabalho aplica-se o método de inovações ao problema de estimação de um processo Gauss-Markov provindo de um sistema multivariável descrito por uma equação de estado. Após a dedução das fórmulas gerais de estimação em termos do processo de inovações obtém-se os algoritmos recursivos do filtro de Kalman-Bucy para o caso não linear contínuo, bem como, para o caso linear continuo e discreto. A seguir, faz-se a representação do processo como saída de um sistema causal e causalmente reversível excitado por um ruído branco, chamada representação por inovações (RI). Os parâmetros deste sistema são determinados a partir da covariância do processo. Finalmente, é desenvolvido um algoritmo para a determinação de uma RI de um processo estacionário provindo de um sistema desconhecido, invariante no tempo. / [en] In this work the innovations method is applied to the estimation problem of a Gauss-Markov process, output of a multivariable system described by a state equation. After obtaining general estimation formulas in terms of the innovations process, the Kalman-Bucy filter recursive algorithms are derived for the nonlinear continuous case as well as for the linear discrete and continuous case. Next, it is given a representation of the process as an output of a causal and causally reversible system to a white noise, known as the innovation representation. The parameters of such a system are determined from the process covariance. Finally, an algorithm is built to obtain an IR of a stationary process coming from an unknown time-invariant system.

This thesis contains the development of a deformation monitoring software based on undifferenced GPS observations. Software like this can be used in alarm systems placed in areas where the earth is unstable. Systems like this can be used in areas where people are in risk of getting hurt, like in earthquake zones or in land slide areas, but they can also be useful when monitoring the movements in buildings, bridges and other artefacts. The main hypotheses that are tested are whether it is possible to detect deformations with undifferenced observations and if it is possible to reach the same accuracy in this mode as when working in a traditional mode where the observations are differenced. The development of a deformation monitoring software based on undifferenced GPS observations is presented. A complete mathematical model is given as well as implementation details. The software is developed in Matlab together with a GPS observation simulator. The simulator is mainly used for debugging purposes. The developed software is tested with both simulated and real observations. Results from tests with simulated observations show that it is possible to detect deformations in the order of a few millimetres with the software. Calculations with real observations give the same results. Further, the result from calculations in static mode indicates that the commercial software and the undifferenced software diverge a few millimetres, which probably depends on different implementations of the tropospheric corrections. In kinematic mode the standard deviation is about 1 millimetre larger in the undifferenced mode than in the double differenced mode. An initial test with different observation weighting procedures indicates that there is a lot of potential to improve the result by applying correct weights to the observations. This is one of the aims in the future work within this project. This thesis are sponsored by the Swedish Research Council for Enviroment, Agricultural Sciences and Spatial Planning, FORMAS within the framework “Monitoring of construction and detection of movements by GPS ref no. 2002-1257" / QC 20101108