No description available.

This dissertation aims to realize the goal of developing robust control solutions that can enable legged robots to navigate complex unknown environments. The idea of creating articulated-legged machines that can mimic animal locomotion has fueled the imagination of many researchers. These legged robots are designed to assist humans in their day-to-day tasks and challenging scenarios such as monitoring remote, inhospitable environments, disaster response, and other dangerous environments. Despite several decades of research, legged robots have yet to reach the dexterity or dynamic stability needed for real-world deployments. A fundamental gap exists in the understanding and development of reliable and scalable algorithms required for the real-time planning and control of legged robots. The overarching goal of this thesis is to formally develop computationally tractable, robust controllers based on nonlinear hybrid systems theory, model predictive control, and optimization for the real-time planning and control of agile locomotion in quadrupedal robots. Toward this objective, this thesis first investigates layered control architectures. In particular, we propose a two-level hierarchical control architecture in which the higher level is based on a reduced-order model predictive control (MPC), and the lower level is based on a full-order quadratic programming (QP) based virtual constraints controller. Specifically, two MPC architectures are explored: 1) An event-based MPC scheme that generates the optimal center of mass (COM) trajectories using a reduced-order linear inverted pendulum (LIP) model, and 2) A time-based MPC scheme that computes the optimal COM and ground reaction forces (GRF) using the reduced-order single rigid body (SRB) dynamics model. The optimal COM trajectories in the event-based MPC and the optimal COM trajectories, along with the ground reaction forces in the time-based MPC, are then tracked by the low-level virtual constraints controller. The event-based MPC scheme is numerically validated on the Vision 60 platform in a physics-based simulation environment. It has significantly reduced the computational burden associated with real-time planning-based MPC schemes. However, owing to the quasi-static nature of the optimal trajectories generated by the LIP model, we explored a time-based MPC scheme using Single Rigid Body Dynamics. This time-based MPC scheme is also numerically validated using the mathematical model of the A1 quadrupedal robot. Most MPC schemes use a reduced-order model to generate optimal trajectories. However, the abstraction and unmodeled dynamics in template models significantly increase the gap between reduced- and full-order models, limiting the robot's full scope and potential. In the second part of the thesis, we aim to develop a computationally tractable robust model predictive control (RMPC) scheme based on convex QPs to bridge this gap. The RMPC framework considers the single rigid body model subject to a set of unmodeled dynamics and plans for the optimal reduced-order trajectory and GRFs. The generated optimal GRFs of the high-level RMPC are then mapped to the full-order model using a low-level nonlinear controller based on virtual constraints and QP. The key innovation of the proposed RMPC framework is that it allows the integration of the hierarchical controller with Reinforcement Learning (RL) techniques to train a neural network to compute the vertices of the uncertainty set numerically. The proposed hierarchical control algorithm is validated numerically and experimentally for robust and blind locomotion of the A1 quadrupedal robot on different indoor and outdoor terrains and at different speeds. The numerical analysis of the RMPC suggests significant improvement in the performance of the rough terrain locomotion compared to the nominal MPC. In particular, the proposed RMPC algorithm outperforms the nominal MPC by over 60% during rough terrain locomotion over 550 uneven terrains. Our experimental studies also indicate a significant reduction in the gap between the reduced full-order models by comparing the desired and actual GRFs. Finally, the last part of the thesis presents a formal approach for synthesizing robust $mathcal{H}_2$- and $mathcal{H}_infty$-optimal MPCs to stabilize the periodic locomotion of legged robots. The proposed algorithm builds on the existing optimization-based control stack. We outline the set of conditions under which the closed-loop nonlinear dynamics around a periodic orbit can be transformed into a linear time-invariant (LTI) system using Floquet theory. We then outline an approach to systematically generate parameterized $mathcal{H}_2$- and $mathcal{H}_infty$- robust controllers using linear matrix inequalities (LMIs). We subsequently established a set of conditions guaranteeing the existence of such robust optimal controllers. The proposed $mathcal{H}_2$- and $mathcal{H}_infty$-optimal MPCs are extensively validated both numerically and experimentally for the robust locomotion of the A1 quadrupedal robot subject to various external disturbances and uneven terrains. Our numerical analysis suggests a significant improvement in the performance of robust locomotion compared to the nominal MPC. / Doctor of Philosophy / Legged robots have always been envisioned to work alongside humans, assisting them in mundane day-to-day tasks to challenging scenarios such as monitoring remote locations, planetary exploration, and supporting relief programs in disaster situations. Furthermore, research into legged locomotion can aid in designing and developing powered prosthetic limbs and exoskeletons. With these advantages in mind, several researchers have created sophisticated-legged robots and even more complicated algorithms to control them. Despite this, a significant gap exists between the agility, mobility, and dynamic stability shown by the existing legged robots and their biological counterparts. To work alongside humans, legged robots have to interact with complex environments and deal with uncertainties in the form of unplanned contacts and unknown terrains. Developing robust control solutions to accommodate disturbances explicitly marks the first step towards safe and reliable real-world deployment of legged robots. Toward this objective, this thesis aims to establish a formal foundation to develop computationally tractable robust controllers for the real-time planning and control of legged robots. Initial investigations in this thesis report on the use of layered control architectures, specifically event-based and time-based Model Predictive Control(MPC) schemes. These layered control architectures consist of an MPC scheme built around a reduced-order model at the high level and a virtual constraints-based nonlinear controller at the low level. Using these layered control architectures, this thesis proposed two robust control solutions to improve the rough terrain locomotion of legged robots. The first proposed robust control solution aims to mitigate one of the issues of layered control architecture. In particular, layered control architectures rely on a reduced order model at the high level to remain computationally tractable. However, the approximation of fullorder models with reduced-order models limits the full scope and potential of the robot. The proposed algorithm aims to bridge the gap between reduced- and full-order models with the integration of model-free Reinforcement Learning (RL) techniques. The second algorithm proposes a formal approach to generate robust optimal control solutions that can explicitly accommodate the disturbances and stabilize periodic legged locomotion. Under some mild conditions, the MPC control solution is analyzed, and an auxiliary feedback control solution that can handle disturbances explicitly is proposed. The thesis also theoretically establishes the sufficient conditions for the existence of such controllers. Both the proposed control solutions are extensively validated using numerical simulations and experiments using an A1 quadrupedal robot as a representative example.

<p>In this thesis, we consider a coherent MIMO system, emphasizing on the simplicity of implementation at both the code generator and the receiver. Specifically, we consider the transmission of a space-time block code (STBC) that is a linear combination of coding matrices weighted by the information symbols through a receiver-correlated flat-fading channel and received by a linear ZF detector. Our target is the design of a code which, while maintaining full data-transmission rate, minimizes the asymptotic average (over all the random channel coefficients) bit error probability of an ZF detector. To this end, we first ensure that the full data rate of symbols is maintained, and then, based on the BER for 4-QAM signals, we derive the conditions for optimal codes and establish a code structure that minimizes the asymptotic average bit error probability. We also prove that the diversity gain of our M × N MIMO system is N − M + 1. The resulting optimum code structure requires the individual coding matrices to be mutually orthogonal when vectorized and is related to covariance matrix of correlated channel. The first optimum structural characteristics of the coding matrices is described as trace-orthogonal. A new approach to express expected value of random correlated channel has been proposed as well. From simulation results we can see that advantage of optimum code over uncoded system is more apparent as channel correlation is higher.</p> / Master of Applied Science (MASc)

<p>We study the server allocation problem for tandem queues in the absence of intermediate buffer space. Servers are assumed to be homogeneous and non-collaborative. We provide policies to maximize the throughput. We break down our work into four stages. First, we assume that all servers are dedicated. We propose an allocation algorithm that assigns servers to stations based on the mean service times and the current number of servers assigned to each station. The algorithm is proposed for stations with exponentially distributed service times, but where the service rate at each station may be different. We further study the impact on the proposed allocation method of including service time distributions with different coecients of variation. Second, we consider tandem queues with both dedicated and flexible servers. We examine policies to dynamically assign flexible servers. When there is one flexible server and two stations each with a dedicated server, we completely characterize the optimal policy. We use the insights gained from applying the Policy Iteration algorithm on systems with three, four, and five stations to devise heuristics for systems of arbitrary size. Third, we study cases where flexibility is constrained such that flexible servers can only service two adjacent stations. We provide optimal policies for tandem queues with three and four stations and compare them with optimal policies for corresponding non-constrained cases. Fourth, we consider two parallel tandem queues with both dedicated servers and vertical flexible servers - servers that can move between corresponding stations of the two tandem queues. The workload allocations are the same for each line and each vertical flexible server moves only between two corresponding stations. We examine policies for dynamic allocation of these vertical flexible servers. When each tandem queue has two stations, each station possesses a dedicated server, and a vertical flexible server exists for each pair of stations, we specify the optimal policy. For cases with more than two stations, heuristic assignments are proposed. We also analyze the throughput improvement gained from adding flexible servers within a tandem queue or between two parallel tandem queues. Numerical results are used to verify the heuristics provided in each of the stages.</p> / Doctor of Philosophy (PhD)

The objectives of this project encompass the design and development of a drug delivery system to continuously deliver therapeutic agents from the stomach to the proximal region of the intestine. The delivery system designed would have sufficient gastric residence time together with near zero-order release kinetics. The physicochemical properties pertaining to the formulation development of the model drug (alfuzosin HCl) were evaluated. Excipients were selected based on the studies of their physicochemical properties and compatibility with the active ingredient. Gastro-retentive dosage forms have been the topic of interest in recent years as a practical approach in drug deliveries to the upper GI tract or for release prolongation and absorption. These dosage forms are particularly suitable for drugs that have local effects on the gastric mucosa in the stomach. Other candidates include drugs that are likely to be absorbed in the upper small intestine, or drugs that are unstable in basic environment of distal intestine and colon or those with low solubility at elevated pH conditions (i.e. weak bases). To develop a gastro-retentive delivery system the following steps were taken. First, to investigate the possible incompatibility issues between the model drug and excipients to be used for the delivery system. Stability and physicochemical properties of the active agent and its mixture with excipients were studied using analytical techniques such as Raman spectroscopy and Differential scanning calorimetry (DSC). No incompatibility issues were detected. Second, Kollidon SR as a relatively new release-rate controlling polymer was incorporated in the final formulation. For solid dosage form the ability of the final powder mix to flow well during manufacturing and the intrinsic characteristics that make it compressible are critical. The in-depth compaction study of Kollidon SR was assessed with the help of a compaction simulator. The flowability, swelling and erosion behavior together with release-rate retarding properties of Kollidon SR were also assessed. The final oral delivery system was based on Kollidon SR and Polyethylene Oxide (PEO) 303 as a monolithic matrix system. The noneffervescent monolithic matrix was made by direct compression. In vitro evaluation of the designed system released the active content in a near zero manner. The dosage form was bouyant in pH 2.0 acidic buffer with no floatation lag time which minimizes the possibility of early gastric emptying. / Pharmaceutics

Data is a critical component in a supervised machine learning system. Many successful applications of learning systems on various tasks are based on a large amount of labeled data. For example, deep convolutional neural networks have surpassed human performance on ImageNet classification, which consists of millions of labeled images. However, one challenge in conventional supervised learning systems is their generalization ability. Once a model is trained on a specific dataset, it can only perform the task on those \emph{seen} classes and cannot be used for novel \emph{unseen} classes. In order to make the model work on new classes, one has to collect and label new data and then re-train the model. However, collecting data and labeling them is labor-intensive and costly, in some cases, it is even impossible. Also, there is an enormous amount of different tasks in the real world. It is not applicable to create a dataset for each of them. These problems raise the need for Transfer Learning, which is aimed at using data from the \emph{source} domain to improve the performance of a model on the \emph{target} domain, and these two domains have different data or different tasks. One specific case of transfer learning is Zero-Shot Learning. It deals with the situation where \emph{source} domain and \emph{target} domain have the same data distribution but do not have the same set of classes. For example, a model is given animal images of `cat' and `dog' for training and will be tested on classifying 'tiger' and 'wolf' images, which it has never seen. Different from conventional supervised learning, Zero-Shot Learning does not require training data in the \emph{target} domain to perform classification. This property gives ZSL the potential to be broadly applied in various applications where a system is expected to tackle unexpected situations. In this dissertation, we develop algorithms that can help a model effectively transfer visual and semantic knowledge learned from \emph{source} task to \emph{target} task. More specifically, first we develop a model that learns a uniform visual representation of semantic attributes, which help alleviate the domain shift problem in Zero-Shot Learning. Second, we develop an ensemble network architecture with a progressive training scheme, which transfers \emph{source} domain knowledge to the \emph{target} domain in an end-to-end manner. Lastly, we move a step beyond ZSL and explore Label-less Classification, which transfers knowledge from pre-trained object detectors into scene classification tasks. Our label-less classification takes advantage of word embeddings trained from unorganized online text, thus eliminating the need for expert-defined semantic attributes for each class. Through comprehensive experiments, we show that the proposed methods can effectively transfer visual and semantic knowledge between tasks, and achieve state-of-the-art performances on standard datasets. / Computer and Information Science

Diabetes is a leading cause of death and disability in the United States. Diabetes requires a lifetime medical treatment. Some diabetes drugs could be taken orally, while others require daily injection or inhalation to maximize bioavailability and minimize toxicity. Parenteral delivery is a group of delivery routes which bypass human gastrointestinal track. Among all the parenteral methods, we chose subcutaneous implant based on its fast act and high patient compliance. When using subcutaneous implant, drug release needs to be strictly controlled. There are three major groups of controlled release methods. Solvent controlled system is already used as osmotic implant. Matrix controlled system is used in Zoladex® implant to treat cancer. Membrane controlled systems is widely used in coating tablets, but not that popular as an implant. Based on the research reported by previous scientists, we decided to build a hybrid system using both matrix and membrane control to delivery human insulin and other small molecule drugs. Subcutaneous environment is different from human GI track. It has less tolerance for external materials so many polymers cannot be used. From the FDA safe excipient database, we selected albumin as our primary polymer and gelatin as secondary choice. In our preliminary insulin diffusion study, we successfully found that insulin mixed with albumin provided a slower diffusion rate compared with control. In addition, we added zinc chloride, a metal salt that can precipitate albumin. The insulin diffusion rate is further reduced. The preliminary study proved that matrix control using albumin is definitely feasible and we might add zinc chloride as another factor. In order to fabricate an implant with appropriate size, we use lyophilisation technology to produce uniformly mixed matrix. Apart from albumin and human insulin, we added sucrose as protectant and plasticizer. The fine powder after freeze-dry was pressed as a form of tablet. The tablets were sealed in Falcon® cell culture insert. Cell culture insert provide a cylinder shape and 0.3 cm2 surface area for drug release. Insulin release study provided a zero order kinetics from prototypes with zinc chloride or 0.4 micron pore size membrane. Caffeine was used as a model drug to investigate the releasing mechanism. Three pore size membranes (0.4, 3 and 8 micron) were tested with same formulation. While 0.4 micron prototypes provided the slowest release, 3 micron ones surprisingly released caffeine faster than 8 micron implants. We calculated the porosity with pore size and concluded that the percentage of open area on a membrane is the key point to control caffeine release. 0.4 micron membranes were used for future research. We increased the percentage of albumin in our excipient, and achieved a slower caffeine release. However, the zero order release could only last for 3 days. After we replaced sucrose with gelatin, a 5 day zero order release of caffeine was achieved. With all the results, we proposed our “Three Phase” drug release mechanism controlled by both membrane and matrix. Seven other small molecule drugs were tested using our prototype. Cloudy suspension was observed with slightly soluble drugs. We updated our “Three Phase” drug release mechanism with the influence of drug solubility. Data shows that releasing rate with same formulation and membrane follows the solubility in pH 7.4. This result proves that our prototype might be used for different drugs based on their solubility. Finally, with all the information of our prototype, we decided to build a “smart insulin implant” with dose adjustment. We proposed an electrical controlled implant with different porosity membranes. Solenoid was used as the mechanical arm to control membrane porosity. 3-D printing technology was used to produce the first real prototype of our implant. Finally, insulin implant with clinically effective insulin release rate was achieved. / Pharmaceutical Sciences

Buildings account for up to a third of total world greenhouse gas GHG emissions, and this pattern is expected to persist. By 2050, cities will be home to 70 % of the world's population, demanding a significant number of buildings to be constructed. Efforts to reduce these emissions in the past had varied performance. However, several examples indicate that well thought and adequately executed mix of building technology coupled with environmental policies may reduce emissions. Therefore, cities worldwide are joining the race to decarbonize their buildings to become net-zero carbon and support green economies through a diversified bundle of policies. However, designing and selecting the appropriate mix of building technology and environmental policies is challenging to generate the most outlast net-zero carbon impacts. This research aims to uncover the intersection between science and policy's role in achieving a global net-zero energy building sector. First, an urban comparative analysis for ten environment-leading cities has been made to understand the latest progress in the building sector and draw on future recommendations. The findings are thematically grouped into five themes a) Building's energy efficiency (energy demand sector). (b) Electrified renewable grids (energy supply sector). (c) Green fiscal incentives (d) Education and capacity building. (e) Governance and collaboration. Second, the University of Ottawa has been utilized as a part of the campus as a living lab initiative to examine installing photovoltaic panels over the campus buildings as part of the university expansion program to achieve net-zero operations by 2040. The following parameters have been considered to address the PV systems viability, 1) the expected electricity output. 2) the initial and operational costs. 3) the GHG reductions in operational energy. 4) the PV system embodied carbons. RETScreen Expert software has been used to perform the Life Cycle Cost Analysis (LCCA) to assess PV system output and financial viability. One Click-LCA software to carry-out Life Cycle Assessment (LCA) to assess embodied carbons. The results indicate from analyzing 31 buildings that 20% - 107% of electricity can be offset depending on each building's energy use and solar collector area. Additionally, the 31 buildings analyzed for electricity generation collectively have the potential to save around 23% of the total campus electricity consumption with a production capacity of 18 million units (kWh) annually, including 21,108 solar panels. Also, the project shows financial viability only if the PV systems are installed as part of the whole campus with a Net Present Value (NPV) of $4,985,89 and an Internal Rate of Return (IRR) of 11.4%. The analysis shows 24% and 18% maximum sensitivity to increased initial cost and decreased electricity generation/rate. Finally, the GHG estimated reductions over 25 years from generated electricity are 14,445 tCO2, and the estimated increased embodied carbons from the Life Cycle Assessment are set to be 1,023 tCO2. Additionally, drawing upon urban analysis and the case study, the research highlights the dynamic nature of the building sector emissions reduction and city initiatives. Thirdly, a detailed analysis was carried out in the System Advisor Model (SAM) software to integrate the solar system with energy storage in the Advanced Research Complex (ARC) Building at the University of Ottawa. The study assesses the system viability and helps the university to reduce its monthly electricity bill and help Ontario to maintain its grid reliability by keeping the electricity demand low at peak times. The findings show that using an integrated solar system with an energy storage system by mitigating 100%, 90%, 75%, and 50% of the building electricity demand during the Ontario gird peak could lead to a Net Present Value of $2,01, $1.70, $1.30, and $0.864 million over 25 years the lifetime of the project through the Ontario Global Adjustment Program. The study also shows that with the absence of the Ontario Global Adjustment Program as a fiscal reform tool and relying only on the time of use electricity rates, the solar panels with an energy storage system could lead to a negative Net Present Value of $-550 thousand.

This work explores the design, construction, and analysis of a novel DC-DC converter which incorporates combinations of switching capacitors and inductors to achieve an integer voltage divider function, without the need for a feedback loop controller to achieve the desired output voltage. The proposed Hybrid Voltage Divider additionally provides zero voltage switching (ZVS) at turn on transitions which yields improved overall efficiency of the converter. Besides a proof-of-concept via computer simulations, another primary goal of this thesis is to demonstrate the functionalities of the proposed Zero Voltage Switching Hybrid Voltage Divider (ZVS-HVD) through hardware prototyping. The proposed ZVS-HVD was designed and constructed to provide a 2:1 division with 24V input voltage at 120W maximum output power utilizing 500kHz switching frequency. Findings from simulations and hardware tests verify that the converter effectively provides the desired 12V output at varying loads with less than 5% voltage ripple. The efficiency of the converter reaches 95.02% at full load and peak efficiency of 96.33% at 55% load. Moreover, the converter consistently maintains the ZVS operations across all switches under varying loads. Overall, results verify the feasibility of the proposed ZVS HVD converter as an alternative solution in providing high efficiency DC voltage division without the need for complex feedback circuitry.

This dissertation is devoted to a unified and comprehensive study of zero-current-transition (ZCT) soft-switching techniques for high-power three-phase PWM inverter applications. Major efforts in this study are as follows: 1) Conception of one new ZCT scheme and one new ZCT topology; 2) Systematic comparison of a family of ZCT inverters; 3) Design, implementation and experimental evaluation of two 55-kW prototype inverters for electric vehicle (EV) motor drives that are developed based on the proposed ZCT concepts; and 4) Investigation of the ZCT concepts in megawatts high-frequency power conversions. The proposed ZCT techniques are also applicable to three-phase power-factor-correction (PFC) rectifiers. In order to minimize switching losses, this work first proposes a new control scheme for an existing three-phase ZCT inverter circuit that uses six auxiliary switches. The proposed scheme, called the six-switch ZV/ZCT, enables all main switches, diodes and auxiliary switches to be turned off under zero-current conditions, and in the meantime provides an opportunity to achieve zero-voltage turn-on for the main switches. Meanwhile, it requires no modification to normal PWM algorithms. Compared with existing ZCT schemes, the diode reverse-recovery current is reduced significantly, the switching turn-on loss is reduced by 50%, the resonant capacitor voltage stress is reduced by 30%, and the current and thermal stresses in the auxiliary switches are evenly distributed. However, a big drawback of the six-switch ZV/ZCT topology, as well as of other types of soft-switching topologies using six auxiliary switches, is the high cost and large space associated with the auxiliary switches. To overcome this drawback, this work further proposes a new three-phase ZCT inverter topology that uses only three auxiliary switches-- the three-switch ZCT. The significance of the proposed three-switch ZCT topology is that, among three-phase soft-switching inverters developed so far, this is the only one that uses fewer than six auxiliary switches and still has the following three features: 1) soft commutation for all main switches, diodes and auxiliary switches in all operation modes; 2) no modification to normal PWM algorithms; and 3) in practical implementations, no need for extra resonant current sensing, saturable cores, or snubbers to protect the auxiliary switches. The proposed six-switch ZV/ZCT and three-switch ZCT inverters, together with existing ZCT inverters, constitute a family of three-phase ZCT inverters. To explore the fundamental properties of these inverters, a systematic comparative study is conducted. A simplified equivalent circuit is developed to unify common traits of ZCT commutations. With the visual aid of state planes, the evolution of the family of ZCT inverters is examined, and their differences and connections are identified. Behaviors of individual inverters, including switching conditions, circulating energy, and device/component stresses, are compared. Based on the proposed six-switch ZV/ZCT and three-switch ZCT techniques, two 55-kW prototype inverters for EV traction motor drives have been built and tested to the full-power level with a closed-loop controlled induction motor dynamometer. The desired ZCT soft-switching features are realized together with motor drive functions. A research effort is carried out to develop a systematic and practical design methodology for the ZCT inverters, and an experimental evaluation of the ZCT techniques in the EV motor drive application is conducted. The design approach integrates system optimization with characterizations of the main IGBT device under the ZCT conditions, selection, testing and characterization of the auxiliary devices, design and selection of the resonant inductors and capacitors, inverter loss modeling and numerical analysis, system-level operation aspects, and layout and parasitic considerations. Different design aspects between these two ZCT inverters are compared and elaborated. The complexity of the 55-kW prototype implementations is compared as well. Efficiencies are measured and compared under a group of torque/speed points for typical EV drive cycles. Megawatts high-frequency power conversion is another potential application of the ZCT techniques. The integrated gate commutated thyristor (IGCT) device is tested and characterized under the proposed six-switch ZV/ZCT condition, and the test shows promising results in reducing switching losses and stresses. Improvements in the IGCT switching frequency and simplification of the cooling requirements under ZCT operations are discussed. In addition, a generalized ZCT cell concept is developed based on the proposed three-switch ZCT topology. This concept leads to the discovery of a family of simplified multilevel soft-switching inverters that reduce the number of auxiliary switches by half, and still maintain desirable features. / Ph. D.