The thesis deals with design of novel circuit structure suitable for hardware implementations of feedforward neural networks. The structure utilizes innovative data bus structure. The main contribution of the structure is in optimization of the utilization of implemented computing units. Proposed architecture is flexible and suitable for implementations of variety of feedforward neural network structures.

For vehicles equipped with electric motors there exist a possibility to recuperate energy during deceleration. This master’s thesis presents a driver support function, a Predictive Deceleration Control (PDC), that warns the driver when to release the accelerator pedal. If the driver follows the instructions from the function the vehicle will decelerate to an appropriate speed at the upcoming road feature, such as a roundabout. The function should both improve energy consumption and enhance driver comfort. This master’s thesis focused on how such a function can be implemented and a proof of concept was designed in a Matlab/Simulink environment. Within the scope of the proof of concept the function should warn the driver to release the accelerator pedal ahead of roundabouts, intersections, speed limit signs and stop signs. With the help of map information and the vehicle most probable path, the distance to the road features could be determined and the predicted braking distance to these road features could be calculated by the function. A feedforward controller was used to control the deceleration phase and was based on a longitudinal vehicle model. The PDC was tested both in a Lynk & Co 01 and CEVT’s dynamic simulator and the results proved that the function can be implemented in for example a Lynk & Co 01 without any additional hardware needed. However, it requires software that can provide the function with map information. During the tests performed during the master’s thesis, map information was acquired with a frequency of 1 Hz, but for the function to become more robust a higher update frequency is required. / För fordon utrustade med elektriska motorer finns det möjlighet till att återvinna energi under inbromsning. Det här examensarbetet presenterar ett förarhjälpmedel som varnar förare när denna ska släppa accelerationspedalen. Om föraren följer uppmaningen kommer bilen att minska hastigheten till en lämplig ingångshastighet för kommande vägobjekt, till exempel en cirkulationsplats. Funktionen ska både förbättra energiförbrukningen och öka förarstödet. Det här examensarbetet fokuserade på hur en sådan funktion kan implementeras och ett exempel på koncept utvecklades i en Matlab/Simulink miljö. Under utvecklingen av funktionen skulle den prediktiva retardationsregle- ringen varna förare att släppa accelerationspedalen innan cirkulationsplatser, korsningar, hastighetsskyltar och stoppskyltar. Men hjälp av kartdata och fordonets mest troliga väg kunde distansen till nästa vägobjekt bestämmas och den förväntade retardationsdistansen beräknas. Dessa värden användes sedan för att bestämma när föraren ska varnas. Framkopplingsreglering användes för att reglera retardationsförloppet och baserades på en longitudinell fordonsmodell. Den utvecklade funktionen testades både i en Lynk & Co 01 och CEVT:s dynamiska simulator och resultaten visade att funktionen kan implementeras i till exempel en Lynk & Co 01 utan någon extra hårdvara. Dock kräver en implementation av funktionen mjukvara som kan bistå funktionen med kartdata och mest trolig väg för fordonet. Under de utförda testerna i bil samlades kartdata och bilens position in med en frekvens av 1 Hz, men för att funktionens tillförlitlighet ska öka krävs en högre uppdateringsfrekvens.

Physics / M. Sc. (Physics)

Limited knowledge of the neuromechanical response to use of an ankle foot orthosis-footwear combination (AFO-FC) has created a lack of consensus in understanding orthotic motion control as a therapeutic treatment. Lack of consensus may hinder the clinician’s ability to target the motion control needs of persons with movement impairment (e.g., peripheral nerve injury, stroke, etc.). Some evidence suggests a proportional relationship between joint motion and neuromuscular activity based on the notion that use of lower limb orthoses that constrain joint motion may invoke motor slacking and decreasing levels of muscle activity. Use of AFO-FCs likely alters the biomechanical and neuromuscular output as the central control system gradually forms new movement patterns. If there is proportional relationship between muscle activation and joint motion, then it could be examined by quantifying joint motion and subsequent neuromuscular output. Considering principles of neuromechanical adjustment, my general hypothesis examines whether orthotic control of lower limb motion alters neuromuscular output in proportion to the biomechanical output as a representation of the limb’s dynamics are updated by the neural control system. The rationale for this approach is that reference knowledge of the neuromechanical response is needed to inform clinicians about how a person responds to walking with motion controlling devices such as ankle foot orthoses combined with footwear. In the first line of research, I hypothesize that a newly developed AFO which maximizes leverage and stiffness will constrain the talocrural joint and alter joint kinematics and ground reaction force patterns. To answer the hypothesis, I sampled kinematics and kinetics of healthy subjects’ treadmill walking using an AFO-FC in a STOP condition and confirmed that the AFO substantially limited the range of talocrural plantarflexion and dorsiflexion motion to 3.7° and in a FREE condition maintained talocrural motion to 24.2° compared to 27.7° in a CONTROL (no AFO) condition. A follow up controlled static loading study sampled kinematics of matched healthy subjects limbs and cadaveric limbs in the AFO STOP and FREE conditions. Findings revealed healthy and cadaveric limbs in the AFO STOP condition substantially limited their limb segment motion similar to matched healthy subjects walking in the STOP condition and in the AFO FREE condition healthy and cadaveric limbs maintained similar limb segment motion to matched healthy subjects walking in the FREE condition. In a second line of research, I hypothesize that flexibility of a newly developed footwear system will allow normal walking kinetics due to the shape and flexibility of the footwear. To answer the hypothesis, I utilized a curved-flexible footwear system integrated with an AFO in a STOP condition and sampled kinematics and kinetics of healthy subjects during treadmill walking. Results revealed subjects elicited similar cadence, stance and swing duration and effective leg-ankle-foot roll over radius compared to walking in the curved-flexible footwear integrated with the AFO in a FREE condition and a CONTROL (no AFO) condition. To validate rollover dynamics of the curved-flexible footwear system, a follow up study of healthy subjects’ treadmill walking in newly developed flat-rigid footwear system integrated with the AFO in a STOP condition revealed interrupted leg-ankle-foot rollover compared to walking in curved-flexible footwear in STOP, FREE and CONTROL conditions. In a third line of research, I hypothesize that use of an AFO that limits talocrural motion in a STOP condition will proportionally reduce activation of Tibialis Anterior, Soleus, Medial and Lateral Gastrocnemii muscles compared to a FREE and CONTROL condition due to alterations in length dependent representation of the limb’s dynamics undergoing updates to the central control system that modify the pattern of motor output. To answer the question, the same subjects and AFO-footwear presented in the first two lines of research were used in a treadmill walking protocol in STOP, FREE, and CONTROL conditions. Findings revealed the same subjects and ipsilateral AFO-footwear system presented in Aim 1 exhibited an immediate yet moderate 30% decline in EMG activity of ipsilateral Soleus (SOL), Medial Gastrocnemius (MG) and Lateral Gastrocnemius (LG) muscles in the STOP condition compared to the CONTROL condition. The reduction in EMG activity in ipsilateral SOL, MG and LG muscles continued to gradually decline during 15 minutes of treadmill walking. On the contralateral leg, there was an immediate yet small increase of 1% to 14% in EMG activity in SOL, MG, LG muscles above baseline. After 10 minutes of walking, the EMG activity in contralateral SOL, MG and LG declined to a baseline level similar to the EMG activity in the contralateral CONTROL condition. These collective findings provide compelling evidence that the moderate 30% reduction in muscle activation exhibited by subjects as they experience substantial (85%) constraint of total talocrural motion in the AFO STOP condition is not proportionally equivalent. Further, the immediate decrease in muscle activation may be due to a reactive feedback mechanism whereas the continued decline may in part be explained by a feedforward mechanism. The clinical relevance of these findings suggests that short term use of orthotic constraint of talocrural motion in healthy subjects does not substantially reduce muscle activation. These preliminary findings could be used to inform the development of orthoses and footwear as therapeutic motion control treatments in the development of motor rehabilitation protocols.

The Heating, Ventilation and Air Conditioning (HVAC) systems are nowadays in almost every new building, develop or improve better control strategies for them is very common, looking to have more energy efficiency and require less input parameters from the user. In this project, new control strategies based in previous theory models has been used with a new approach in order to find a good solution for irregular occupied spaces. In this new approach a feed-forward filter with a fixed preheating time, using an algorithm based on an identified model, calculates how much degrees the temperature room can be decreased and regulate the power of the radiators to do it.The results of this project displays that the chosen model have to be changed but the idea is interesting, because the simulations of the reference building give, with a preheating timeof 2 hours, around 3ºC of temperature reduction during 18 days and savings of 33% of the heat energy needed for the whole month.Considering that buildings and the residential sector currently account for 40 percent of Sweden's energy consumption and around 25 percent of other countries like USA or Spain, and that irregular spaces are more or less a 10% of the governmental, institutional, academic or public buildings, the potential savings are not negligible. The evaluation of this control strategy with its mathematical model as well as its resultsduring the month of January and the behavior of the system along the year have been made with the help of IDA program for simulation of the reference building and its energy system.

Laser-initiierte Freie Elektronen Laser (FEL) und Energy Recovery Linearbeschleuniger (ERL) erfordern höchste Stabilität der beschleunigten Elektronenpakete. Die zeitliche Streuung der Elektronenpakete und die mittlere Energieabweichung in den Undulatoren sollten in der Grössenordnung von Femtosekunden bzw. im Promille Bereich liegen. Das erfordert eine Regelung der Hochfrequenz (HF) Beschleunigungsfelder in den supraleitenden Hohlraumresonatoren bis zu 0.02° Phasen- und 1e-4 Amplitudengenauigkeit. Die TESLA Resonatoren des 2.3 GeV Linearbeschleunigers des geplanten BESSY-FEL sollen im Dauerstrichbetrieb bei geringer Strahllast betrieben werden. Die HF Resonanzbreite ist folglich sehr schmalbandig und liegt im Bereich von 10 Hertz. Um die erreichbare Feldstabilität zu erfassen, wurden die Resonatoren einem umfangreichen Messprogramm in der HoBiCaT Testanlage unterzogen. Eine Charakterisierung der vollständigen Resonatoreinheit hinsichtlich der mechanischen Verstimmung durch Mikrophonie, statische -und dynamische Lorentzkraftverstimmung, ihrer mechanischen Eigenschaften und HF-System Rauschen erbrachte wichtige Daten, um die zu erwartende Feld -und somit Strahlstabilität im Linac zu simulieren. Die gemessene Mikrophonie betrug 1-5 Hz rms, ist somit eine dominante Fehlergröße und wirkt sich limitierend auf die Strahlstabilität im Linac aus. Um sie zu minimieren, wurden aktive Dämpfungsmethoden entwickelt. Dazu wurden unterschiedliche mechanische Abstimmungssysteme mit integrierten Piezoelementen getestet. Ein adaptiver, vorauskompensierender Regelungsalgorithmus wurde entwickelt, welcher die gemessene Transferfunktion der Abstimmvorrichtung beinhaltet. Damit wurde eine Kompensierung der Mikrophonie um einem Faktor von zwei bis sieben erreicht. Die Einbeziehung dieser Regelung in die Linacsimulationen zeigte, dass diese einen wichtigen Beitrag zur Erreichbarkeit der benötigten Strahlstabilität für zukünftige FELs und ERLs darstellt. / In seeded High-Gain-Harmonic-Generation free electron lasers or energy recovery linear accelerators the requirements for the bunch-to-bunch timing and energy jitter of the beam are in the femtosecond and per mill regime. This implies the ability to control the cavity radio-frequency (RF) field to an accuracy of 0.02° in phase and up to 1e-4 in amplitude. For the planned BESSY-FEL it is envisaged to operate 144 superconducting 1.3 GHz cavities of the 2.3 GeV driver linac in continuous wave mode and at a low beam current. The cavity resonance comprises a very narrow bandwidth of the order of tens of Hertz. Such cavities have been characterized under accelerator like conditions in the HoBiCaT test facility. It was possible to measure the error sources affecting the field stability in continuous wave (CW) operation. Microphonics, the main error source for a mechanical detuning of the cavities, lead to an average fluctuation of the cavity resonance of 1-5 Hz rms. Furthermore, the static and dynamic Lorentz force detuning and the helium pressure dependance of the cavity resonance have been measured. Single cavity RF control and linac bunch-to-bunch longitudinal phase space modeling containing the measured properties showed, that it is advisable to find means to minimize the microphonics detuning by mechanical tuning. Thus, several fast tuning systems have been tested for CW operation. These tuners consist of a motor driven lever for slow and coarse tuning and a piezo that is integrated into the tuner support for fast and fine tuning. Regarding the analysis of the detuning spectrum an adaptive feedforward method based on the least-mean-square filter algorithm has been developed for fast cavity tuning. A detuning compensation between a factor of two and up to a factor of seven has been achieved. Modeling the complete system including the fast tuning scheme, showed that the requirements of the BESSY-FEL are attainable.

Near-field wireless power transfer (WPT) technology facilitates the energy autonomy of heterogeneous systems, significantly augmenting complementary metal-oxide-semiconductor field-effect-transistor (CMOS) technology. In low-power wearable devices, existing power conditioning integrated circuits do not maximize the power factor (PF) for rectification and power conversion efficiency (PCE) due to multiple conversion. Additionally, there is no core power management for the entire power flow. The majority of the research focuses on active rectifiers, which reduce the turn-on voltage for rectification. Certain studies target the output voltage regulation via feedback to the transmitter or direct battery charging without power maximization. Firstly, this study investigates a high-power factor WPT front-end circuit that is namely the mono-periodic switching rectifier (MPSR) and implemented in a 0.18µm 1.8V/5V CMOS process. Integrated phase synchronizers are used to align the waveshape of a wirelessly-coupled sinusoidal voltage source in a receiving coil to the corresponding conducting current. Using this approach, the PF can be increased from roughly 0.6 to unity without requiring any wireless or wired feedback to the transmitter. The proposed MPSR can also provide AC-DC rectification, and step up and down the sinusoidal voltage source's peak amplitude using a pulse-width modulator. Measured voltage conversion ratios range between 0.73X and 2X, and the PF can be boosted up to unity. Secondly, the wireless power system-on-chip (WPower-SoC) is proposed and implemented in a 0.18µm 1.8V/3.3V CMOS process. The WPower-SoC integrating power management can provide rectification, output voltage regulation, and battery charging. Additionally, the implementation of feedforward envelope detection (FED) can reduce the variation in a wireless power link and improve load transient responses. Simulated results demonstrate that 5% of the output voltage regulation is improved when an output load changes. Moreover, the FED reduces approximately 40% of the transient response time. Overshoot and undershoot voltages are decreased by 23% and 26.5%, respectively. The measured output voltage regulates at 3.42V and can supply output power up to 342mW. A temperature sensor as part of the power management core remains active when the WPT receivers enter sleep mode to prolong the battery usage time. In the final part of this study, a nano-watt high-accuracy temperature sensing core is implemented in a 0.18µm 1.8V/3.3V CMOS process that can self-compensate the temperature shift without the need for additional compensating techniques that consume extra power.

The Littoral Acoustic Demonstration Center has collected passive acoustic monitoring data in the northern Gulf of Mexico since 2001. Recordings were made in 2007 near the Deepwater Horizon oil spill that provide a baseline for an extensive study of regional marine mammal populations in response to the disaster. Animal density estimates can be derived from detections of echolocation signals in the acoustic data. Beaked whales are of particular interest as they remain one of the least understood groups of marine mammals, and relatively few abundance estimates exist. Efficient methods for classifying detected echolocation transients are essential for mining long-term passive acoustic data. In this study, three data clustering routines using k-means, self-organizing maps, and spectral clustering were tested with various features of detected echolocation transients. Several methods effectively isolated the echolocation signals of regional beaked whales at the species level. Feedforward neural network classifiers were also evaluated, and performed with high accuracy under various noise conditions. The waveform fractal dimension was tested as a feature for marine biosonar classification and improved the accuracy of the classifiers. [This research was made possible by a grant from The Gulf of Mexico Research Initiative. Data are publicly available through the Gulf of Mexico Research Initiative Information & Data Cooperative (GRIIDC) at https://data.gulfresearchinitiative.org.] [DOIs: 10.7266/N7W094CG, 10.7266/N7QF8R9K]

In this thesis, we report the observations of optical squeezing from second harmonic generation (SHG), optical parametric oscillation (OPO) and optical parametric amplification (OPA). Demonstrations and proposals of applications involving the squeezed light and electro-optic control loops are presented. ¶ In our SHG setup, we report the observation of 2.1 dB of intensity squeezing on the second harmonic (SH) output. Investigations into the system show that the squeezing performance of a SHG system is critically affected by the pump noise and a modular theory of noise propagation is developed to describe and quantify this effect. Our experimental data has also shown that in a low-loss SHG system, intra-cavity nondegenerate OPO modes can simultaneously occur. This competition of nonlinear processes leads to the optical clamping of the SH output power and in general can degrade the SH squeezing. We model this competition and show that it imposes a limit to the observable SH squeezing. Proposals for minimizing the effect of competition are presented. ¶ In our OPO setup, we report the observation of 7.1 dB of vacuum squeezing and more than 4 dB of intensity squeezing when the OPO is operating as a parametric amplifier. We present the design criteria and discuss the limits to the observable squeezing from the OPO.We attribute the large amount of squeezing obtained in our experiment to the high escape efficiency of the OPO. The effect of phase jitter on the squeezing of the vacuum state is modeled. ¶ The quantum noise performance of an electro-optic feedforward control loop is investigated. With classical coherent inputs, we demonstrate that vacuum fluctuations introduced at the beam splitter of the control loop can be completely cancelled by an optimum amount of positive feedforward. The cancellation of vacuum fluctuations leads to the possibility of noiseless signal amplification with the feedforward loop. Comparison shows that the feedforward amplifier is superior or at least comparable in performance with other noiseless amplification schemes. When combined with an injection-locked non-planar ring Nd:YAG laser, we demonstrate that signal and power amplifications can both be noiseless and independently variable. ¶ Using squeezed inputs to the feedforward control loop, we demonstrate that information carrying squeezed states can be made robust to large downstream transmission losses via a noiseless signal amplification. We show that the combination of a squeezed vacuum meter input and a feedforward loop is a quantum nondemolition (QND) device, with the feedforward loop providing an additional improvement on the transfer of signal. In general, the use of a squeezed vacuum meter input and an electro-optic feedforward loop can provide pre- and post- enhancements to many existing QND schemes. ¶ Finally, we proposed that the quantum teleportation of a continuous-wave optical state can be achieved using a pair of phase and amplitude electro-optic feedforward loops with two orthogonal quadrature squeezed inputs. The signal transfer and quantum correlation of the teleported optical state are analysed. We show that a two dimensional diagram, similar to the QND figures of merits, can be used to quantify the performance of a teleporter.

Les opérateurs de centres d'appels sont exposés à un bruit de fond relativement élevé, principalement composé des conversations de leurs collègues. Ce bruit de fond gêne l'intelligibilité des conversations téléphoniques et induit chez les opérateurs des comportements dommageables. Une solution pour remédier à ce problème consiste à équiper leur casque téléphonique d'un système de contrôle actif afin de réduire le bruit directement au niveau de leurs oreilles. En raison de la limitation du contrôle par feedback au contrôle des basses fréquences, nous avons envisagé l'étude d'un dispositif de contrôle par feedforward, le principe étant d'utiliser un ou plusieurs capteurs de référence détectant le bruit gênant en amont de la zone de réduction du bruit. Le contrôle par feedforward en centre d'appels peut être limité par des facteurs relevant de la physique et du traitement du signal en raison de la complexité du champ acoustique et des caractéristiques instationnaire et large-bande du bruit. Cette thèse présente l'étude acoustique, algorithmique et technologique d'un système de contrôle par feedforward sous casque, en se basant sur des simulations de contrôle réalisées à partir d'enregistrements sonores et d'une expérience de contrôle temps-réel. Nous y montrons que malgré la faisabilité acoustique et algorithmique du contrôle par feedforward dans les centres d'appels, la technologie actuelle des casques téléphoniques ne permet pas d'obtenir un contrôle satisfaisant.