| 1. |
- Alhashimi, Anas, 1978-
(författare)
-
Statistical Sensor Calibration Algorithms
- 2018
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- The use of sensors is ubiquitous in our IT-based society; smartphones, consumer electronics, wearable devices, healthcare systems, industries, and autonomous cars, to name but a few, rely on quantitative measurements for their operations. Measurements require sensors, but sensor readings are corrupted not only by noise but also, in almost all cases, by deviations resulting from the fact that the characteristics of the sensors typically deviate from their ideal characteristics.This thesis presents a set of methodologies to solve the problem of calibrating sensors with statistical estimation algorithms. The methods generally start with an initial statistical sensor modeling phase in which the main objective is to propose meaningful models that are capable of simultaneously explaining recorded evidence and the physical principle for the operation of the sensor. The proposed calibration methods then typically use training datasets to find point estimates of the parameters of these models and to select their structure (particularlyin terms of the model order) using suitable criteria borrowed from the system identification literature. Subsequently, the proposed methods suggest how to process the newly arriving measurements through opportune filtering algorithms that leverage the previously learned models to improve the accuracy and/or precision of the sensor readings.This thesis thus presents a set of statistical sensor models and their corresponding model learning strategies, and it specifically discusses two cases: the first case is when we have a complete training dataset (where “complete” refers to having some ground-truth informationin the training set); the second case is where the training set should be considered incomplete (i.e., not containing information that should be considered ground truth, which implies requiring other sources of information to be used for the calibration process). In doing so, we consider a set of statistical models consisting of both the case where the variance of the measurement error is fixed (i.e., homoskedastic models) and the case where the variance changes with the measured quantity (i.e., heteroskedastic models). We further analyzethe possibility of learning the models using closed-form expressions (for example, when statistically meaningful, Maximum Likelihood (ML) and Weighted Least Squares (WLS) estimation schemes) and the possibility of using numerical techniques such as Expectation Maximization (EM) or Markov chain Monte Carlo (MCMC) methods (when closed-form solutions are not available or problematic from an implementation perspective). We finally discuss the problem formulation using classical (frequentist) and Bayesian frameworks, and we present several field examples where the proposed calibration techniques are applied on sensors typically used in robotics applications (specifically, triangulation Light Detection and Rangings (Lidars) and Time of Flight (ToF) Lidars).
|
|
| 2. |
- Fresk, Emil
(författare)
-
The Core of Aerial Robotic Workers : Generalized Modeling, Estimation, and Control
- 2018
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- In this thesis we are going to explore what the operational core, both mathematically and algorithmically, of an Aerial Robotic Worker consists of, in order to estimate its egomotion and parameters, and adaptively control the aerial vehicle. Moreover, the aim of this thesis is to be a condensed reference for the corresponding areas of aerial robotics, in order to provide a stable and complete foundation on which one can continue research on. The areas that are covered in this Thesis are: 1) the fundamental modeling of the generalized aerial vehicle, where the kinematics, sensors and motor/thrust models will be presented together with simplified models for the motor characteristics, which will form the basis for all the future derivations, 2) how to model, calibrate and compensate for the errors existing in, and induced into, cheap accelerometers and gyroscopes, as these sensors constitute the aerial platform's core sensor suite as the inertial sensor. Successful methodologies and results are presented and evaluated to show that the cost of calibration can be dramatically reduced without loss of accuracy nor mechanical complexity. 3) How to perform inertial sensor driven egomotion and parameter estimation to lay the foundation for adaptive control strategies, where specific weight will be put on the successful development of a new profound sensory system which has the possibility to replace GPS in robotics applications, while also being able to perform indoors and in GPS denied environments, and which was the core of the localization module done in the AEROWORKS project, enabling the full, high accuracy localization around tall, GPS interfering infrastructure. And finally 4) how to utilize the estimation in low and high-level adaptive controllers, where specific results on how to successfully compensate for the movement of the center of gravity, together with the reduction of thrust over time due to declining battery voltage. Moreover we will explore the use case of Aerial Robotic Workers in real life applications and we will identify and comment on potential future directions of these aerial robotic systems and the impact theses can have in both research and society.
|
|
| 3. |
- Kadhim, Ali
(författare)
-
Selection of Decentralized Control Configuration for Uncertain Systems
- 2018
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Industrial processes nowadays involve hundreds or more of variables to be maintained within predefined ranges to achieve the production demands. However, the lack of accurate models and practical tools to design controllers for such large processes motivate the engineers/practitioners to break the processes down into smaller subsystems and applying decentralized controllers.In contrast to the centralized controller, the decentralized controller is favourable in large-scale systems due to its robustness against loop failures and model uncertainties as well as being easier to tune and update. Yet, two steps are required prior to synthesizing these single-input single-output (SISO) controllers that comprise the decentralized controller. In the first step, a set of manipulated and the controlled variables need to be selected while the second step deals with pairing these variables to close the SISO control loops in a manner that limits the interaction between the loops. The latter step, called "input-output pairing", is usually performed by means of interaction measures (IM) tools using a nominal system model. Taking model uncertainties into consideration when deciding the pairing selection of the decentralized controller is necessary since adopting the pairing based on the nominal system model might be misleading and resulting in poor system performance or instability. It is therefore essential to have tools indicating the extent to which the pairing based on the nominal model persists against gain variations due to uncertainties.The work in this thesis presents a methodology that determines whether the effect of gain uncertainty would invalidate the selected pairing. This has been done following the definition of the most established IM tool used in the industry, the relative gain array (RGA), and some of its variants. Further, a procedure has been developed to automatically obtain the optimal input-output pairing by formulating the pairing rules of relative interaction array (RIA) method as an \textit{assignment problem} (AP), and thus, simplifying the pairing selection for large-scale systems. Thereafter, uncertainty bounds of the RIA elements are employed to validate the pairing selection under the effect of given variations of the system gain. Moreover, following the RIA pairing rules, a method is proposed to calculate a minimum amount of uncertainty that renders a perturbed system for which the pairing, obtained from the nominal system model, becomes invalid.In the aforementioned methodologies, a parametric system model is assumed to be known. To relax this constraint, an approach is therefore proposed and evaluated which identifies the pairing of the decentralized controller directly from the input-output data. This approach has the advantage of exempting the user from deriving a complete parametric model of the plant to decide the input-output pairing, and hence saves the efforts by finding the parameters of the most significant subsystems in a multivariable system. The frequency response of the system and its covariance, and subsequently the dynamic RGA (DRGA) and corresponding uncertainty bounds, are estimated from the input-output data by employing a nonparametric system identification approach. In short, the work presented in this thesis provides beneficial methodologies for researchers in academia as well as engineers in industry to predict the influence of the system gain uncertainty on the pairing selection of decentralized controllers.
|
|
