As trucks are getting more and more advanced, information about their weight has become a key factor for controlling them in a more fuel efficient and safe manner. Knowing the mass of a heavy duty vehicle in real time has been a difficult challenge for the truck manufacturers. With the processing power for electronic control units in trucks steadily increasing, more advanced algorithms for calculating the mass has been developed, but at the moment there still is a wish for better performance. Since there is a lack of good information regarding the external forces acting on the vehicle, forces that depends on the slope of the road, foundation of the road and the wind, the methods have to be able to disregard these. Such an approach, based on an indirect least square solution, has been evaluated in this thesis. The results have been promising and based on these a recommendation about further evaluation has been made.

Please read the abstract (Summary) in the section, 00front of this document / Dissertation (M Eng (Electronic Engineering))--University of Pretoria, 2007. / Electrical, Electronic and Computer Engineering / MEng / unrestricted

A flight data acquisition system was developed to aid unmanned vehicle designers in verifying the vehicle's design performance. The system is reconfigurable and allows the designer to choose the correct combination of complexity, risk, and cost for a given flight test. The designer can also reconfigure the system to meet packaging and integration requirements. System functionality, repeatbility, and accuracy was validated by collecting data during multiple flights of a radio-controlled aircraft. Future work includes sensor fusion, thrust prediction methods, stability and control derivative estimation, and growing Cal Poly's small-scale component aerodynamic database.

No description available.

This thesis investigates autonomous landing of a micro air vehicle (MAV) on a nonstationary ground platform. Unmanned aerial vehicles (UAVs) and micro air vehicles (MAVs) are becoming every day more ubiquitous. Nonetheless, many applications still require specialized human pilots or supervisors. Current research is focusing on augmenting the scope of tasks that these vehicles are able to accomplish autonomously. Precise autonomous landing on moving platforms is essential for self-deployment and recovery of MAVs, but it remains a challenging task for both autonomous and piloted vehicles. Model Predictive Control (MPC) is a widely used and effective scheme to control constrained systems. One of its variants, output-feedback tube-based MPC, ensures robust stability for systems with bounded disturbances under system state reconstruction. This thesis proposes a MAV control strategy based on this variant of MPC to perform rapid and precise autonomous landing on moving targets whose nominal (uncommitted) trajectory and velocity are slowly varying. The proposed approach is demonstrated on an experimental setup.

In science, technology, and engineering, creating models of the environment to predict future events has always been a key component. The models could be everything from how the friction of a tire depends on the wheels slip  to how a pathogen is spread throughout society.  As more data becomes available, the use of data-driven black-box models becomes more attractive. In many areas they have shown promising results, but for them to be used widespread in safety-critical applications such as autonomous driving some notion of uncertainty in the prediction is required. An example of such a black-box model is neural networks (NNs). This thesis aims to increase the usefulness of NNs by presenting an method where uncertainty in the prediction is obtained by linearization of the model. In system identification and sensor fusion, under the condition that the model structure is identifiable, this is a commonly used approach to get uncertainty in the prediction from a nonlinear model. If the model structure is not identifiable, such as for NNs, the ambiguities that cause this have to be taken care of in order to make the approach applicable. This is handled in the first part of the thesis where NNs are analyzed from a system identification perspective, and sources of uncertainty are discussed. Another problem with data-driven black-box models is that it is difficult to know how flexible the model needs to be in order to correctly model the true system. One solution to this problem is to use a model that is more flexible than necessary to make sure that the model is flexible enough. But how would that extra flexibility affect the uncertainty in the prediction? This is handled in the later part of the thesis where it is shown that the uncertainty in the prediction is bounded from below by the uncertainty in the prediction of the model with lowest flexibility required for representing true system accurately.  In the literature, many other approaches to handle the uncertainty in predictions by NNs have been suggested, of which some are summarized in this work. Furthermore, a simulation and an experimental studies inspired by autonomous driving are conducted. In the simulation study, different sources of uncertainty are investigated, as well as how large the uncertainty in the predictions by NNs are in areas without training data. In the experimental study, the uncertainty in predictions done by different models are investigated. The results show that, compared to existing methods, the linearization method produces similar results for the uncertainty in predictions by NNs. An introduction video is available at https://youtu.be/O4ZcUTGXFN0 / Inom forskning och utveckling har det har alltid varit centralt att skapa modeller av verkligheten. Dessa modeller har bland annat använts till att förutspå framtida händelser eller för att styra ett system till att bete sig som man önskar. Modellerna kan beskriva allt från hur friktionen hos ett bildäck påverkas av hur mycket hjulen glider till hur ett virus kan sprida sig i ett samhälle. I takt med att mer och mer data blir tillgänglig ökar potentialen för datadrivna black-box modeller. Dessa modeller är universella approximationer vilka ska kunna representera vilken godtycklig funktion som helst. Användningen av dessa modeller har haft stor framgång inom många områden men för att verkligen kunna etablera sig inom säkerhetskritiska områden såsom självkörande farkoster behövs en förståelse för osäkerhet i prediktionen från modellen. Neuronnät är ett exempel på en sådan black-box modell. I denna avhandling kommer olika sätt att tillförskaffa sig kunskap om osäkerhet i prediktionen av neuronnät undersökas. En metod som bygger på linjärisering av modellen för att tillförskaffa sig osäkerhet i prediktionen av neuronnätet kommer att presenteras. Denna metod är välbeprövad inom systemidentifiering och sensorfusion under antagandet att modellen är identifierbar. För modeller såsom neuronnät, vilka inte är identifierbara behövs det att det tas hänsyn till tvetydigheterna i modellen. En annan utmaning med datadrivna black-box modeller, är att veta om den valda modellmängden är tillräckligt generell för att kunna modellera det sanna systemet. En lösning på detta problem är att använda modeller som har mer flexibilitet än vad som behövs, det vill säga en överparameteriserad modell.  Men hur påverkas osäkerheten i prediktionen av detta? Detta är något som undersöks i denna avhandling, vilken visar att osäkerheten i den överparameteriserad modellen kommer att vara begränsad underifrån av modellen med minst flexibilitet som ändå är tillräckligt generell för att modellera det sanna systemet. Som avslutning kommer dessa resultat att demonstreras i både en simuleringsstudie och en experimentstudie inspirerad av självkörande farkoster. Fokuset i simuleringsstudien är hur osäkerheten hos modellen är i områden med och utan tillgång till träningsdata medan experimentstudien fokuserar på jämförelsen mellan osäkerheten i olika typer av modeller.Resultaten från dessa studier visar att metoden som bygger på linjärisering ger liknande resultat för skattningen av osäkerheten i prediktionen av neuronnät, jämfört med existerande metoder. / iQdeep

Optimal input design concerns the design of an input sequence to maximize the information retrieved from an experiment. The design of the input sequence is performed by optimizing a cost function related to the intended model application. Several approaches to input design have been proposed, with results mainly on linear models. Under the linear assumption of the model structure, the input design problem can be solved in the frequency domain, where the corresponding spectrum is optimized subject to power constraints. However, the optimization of the input spectrum using frequency domain techniques cannot include time-domain amplitude constraints, which could arise due to practical or safety reasons. In this thesis, a new input design method for nonlinear models is introduced. The method considers the optimization of an input sequence as a realization of the stationary Markov process with finite memory. Assuming a finite set of possible values for the input, the feasible set of stationary processes can be described using graph theory, where de Bruijn graphs can be employed to describe the process. By using de Bruijn graphs, we can express any element in the set of stationary processes as a convex combination of the measures associated with the extreme points of the set. Therefore, by a suitable choice of the cost function, the resulting optimization problem is convex even for nonlinear models. In addition, since the input is restricted to a finite set of values, the proposed input design method can naturally handle amplitude constraints. The thesis considers a theoretical discussion of the proposed input design method for identification of nonlinear output error and nonlinear state space models. In addition, this thesis includes practical applications of the method to solve problems arising in wireless communications, where an estimate of the communication channel with quantized data is required, and application oriented closed-loop experiment design, where quality constraints on the identified parameters must be satisfied when performing the identification step. / <p>QC 20141110</p>

Generic simulations of tremor propagation through the upper limb have been achieved using a previously developed postural tremor model, but this model had not yet been compared with experimental data or utilized for subject-specific studies. This work addressed these two issues, which are important for optimizing peripheral tremor suppression techniques. For tractability, we focused on a subsystem of the upper limb: the isolated wrist, including the four prime wrist muscles (extensor carpi ulnaris, flexor carpi ulnaris, extensor carpi radialis, and flexor carpi radialis) and the two degrees of freedom of the wrist (flexion-extension and radial-ulnar deviation). Muscle excitation and joint displacement signals were collected while subjects with Essential Tremor resisted gravity. System identification was implemented for three subjects who experienced significant tremor using two approaches: 1. Generic linear time-invariant (LTI) models, including autoregressive-exogenous (ARX) and state-space forms, were identified from the experimental data, and characteristics including model order and modal parameters were compared with the previously developed postural tremor model; 2. Subject-specific parameters for the previously developed postural tremor model were directly estimated from experimental data using nonlinear least-squares optimization combined with regularization. The identified LTI models fit the experimental data well, with coefficients of determination of 0.74 ± 0.18 and 0.83 ± 0.13 for ARX and state-space forms, respectively. The optimal model orders identified from the experimental data (4.8 ± 1.9 and 6.4 ± 1.9) were slightly lower than the orders of the ARX and state-space forms of the previously developed model (6 and 8). For each subject, at least one pair of identified complex poles aligned with the complex poles of the previously developed model, whereas the identified real poles were assumed to represent drift in the data rather than characteristics of the system. Subject-specific parameter estimates reduced the sum of squared-error (SSE) between the measured and predicted joint displacement signals to be between 10% and 50% of the SSE using generic literature parameters. The predicted joint displacements maintained high coherence at the tremor frequency for flexion-extension (0.90 ± 0.10), which experienced the most tremor. We successfully applied multiple system identification techniques to identify tremor propagation models using only tremorogenic muscle activity as the input. These techniques identified model order, poles, and subject-specific model parameters, and indicate that tremor propagation at the wrist is well approximated by an LTI model.

No description available.

With recent developments in power conversion technologies and market trends that are driving those technologies toward further miniaturization and greater integration, the need for verifying an empirically based modeling methodology for proprietary power converters such as Power Supply on Chip (PwrSoC) products has risen significantly. This need motivates the investigation of black-box models which require little or no knowledge of the internal workings of a system, for those areas of industry adopting PwrSoC technology as a point-of-load solution. This thesis reports a black-box modeling method tailored to accommodate but not limited to the requirements of a specific commercially available PwrSoC technology.