Systems Design Engineering

Permanent URI for this collectionhttps://uwspace.uwaterloo.ca/handle/10012/9914

This is the collection for the University of Waterloo's Department of Systems Design Engineering.

Research outputs are organized by type (eg. Master Thesis, Article, Conference Paper).

Waterloo faculty, students, and staff can contact us or visit the UWSpace guide to learn more about depositing their research.

Browse

Now showing 1 - 20 of 754

2D Material Based PTE Detectors with Room Temperature Operations
(University of Waterloo, 2023-12-21) Xie, Zhemiao
Real-time, room-temperature operation and self-powered photothermoelectric (PTE) detection emeries are advanced and versatile solutions for various applications. These detectors offer the advantage of not requiring external power sources, making them portable and suitable for remote or low-power environments. Additionally, their ability to operate at room temperature eliminates the need for costly and complex cooling systems, making them more accessible and cost-effective for various industries and research fields. However, issues of massive fabrication, complicated manipulations, long-term stability, and flexibility are concerned with engaging new exploration on PTE detectors with low-dimensional materials. Two-dimensional (2D) materials are emerging as leading ones due to their broadband detection from Terahertz (THz) to ultraviolet (UV), electrical conductivity with a small band gap, and strong polymer affinity for thermoelectrical conversion. This thesis aims at using 2D nanomaterials of graphene and molybdenum carbide (Mo₂C) MXene for exploring new PTE detectors, guiding 2D materials methodology, leading the investigation of polymer composites, and providing insights into various industrial, imaging, and health monitor applications. This thesis introduces three types of room operation and self-powered PTE architectures with 2D nanomaterials. First, we developed a new doped polyaniline (PANI) as the composite material with a few layers of sheets of graphene. Semi-transparent, broadband infrared (IR) detection and robust flexibility features are presented. Second, a vertical graphene/polyethylenimine (PEI) composite multi-element H-shaped detector with the spray-coating method is presented. High response time, detectivity, and a broadly responsive range are achieved with PEI concentration adjustment, lowering the thermal conductivity and enhancing compacity, focusing the realistic situations with low incident power. Finally, we propose a low-noise PTE device that operates at room temperature by Mo₂C and Poly(3,4-ethylene-dioxythiophene)-poly(styrenesulfonate) (PEDOT: PSS) nanomaterials with a flexible PET substrate. Superior energy conversion and long-term stability of the material and sustainability from surroundings are achieved through material optimizations. Based on such PTE detectors, two promising systems—the motion tracking system and the non-destructive testing (NDT) imaging system—demonstrate the time-tracking of human radiation and high-resolution imaging applications.
A 3D ellipsoidal volumetric foot–ground contact model for forward dynamics
(Springer, 2018-04-01) Brown, Peter; McPhee, John
Foot–ground contact models are an important part of forward dynamic biomechanic models, particularly those used to model gait, and have many challenges associated with them. Contact models can dramatically increase the complexity of the multibody system equations, especially if the contact surface is relatively large or conforming. Since foot–ground contact has a large potential contact area, creating a computationally efficient model is challenging. This is particularly problematic in predictive simulations, which may determine optimal performance by running a model simulation thousands of times. An ideal contact model must find a balance between accuracy for large, conforming surfaces, and computational efficiency.Volumetric contact modelling is explored as a computationally efficient model for foot–ground contact. Previous foot models have used volumetric contact before, but were limited to 2D motion and approximated the surfaces as spheres or 2D shapes. The model presented here improves on current work by using ellipsoid contact geometry and considering 3D motion and geometry. A gait experiment was used to parametrise and validate the model. The model ran over 100 times faster than real-time (in an inverse simulation at 128 fps) and matched experimental normal force and centre of pressure location with less than 7% root-mean-square error.In most gait studies, only the net reaction forces, centre of pressure, and body motions are recorded and used to identify parameters. In this study, contact pressure was also recorded and used as a part of the identification, which was found to increase parameter optimisation time from 10 to 164 s (due to the additional time needed to calculate the pressure distribution) but helped the results converge to a more realistic model. The model matched experimental pressures with 33–45% root-mean-square error, though some of this was due to measurement errors.The same parametrisation was done with friction included in the foot model. It was determined that the velocity-based friction model that was used was inappropriate for use in an inverse-dynamics simulation. Attempting to optimise the model to match experimental friction resulted in a poor match to the experimental friction forces, inaccurate values for the coefficient of friction, and a poorer match to the experimental normal force.
3D MEMS Microassembly
(University of Waterloo, 2008-09-03T17:52:49Z) Do, Chau
Due to the potential uses and advantages of 3D microelectromechanical systems (MEMS), research has been ongoing to advance the field. The intention of my reasearch is to explore different gripper designs and their interaction with corresponding components to establish a 3D microassembly system. In order to meet these goals, two grippers were designed using different mechanisms for grasping. At the same time, corresponding parts capable of being constructed into a 3D microstructure were designed to interact with the grippers. The microcomponents were fabricated using PolyMUMPS, a part of the Multi-User MEMS Processes (MUMPS), and experimentation was conducted with the goal of constructing a 3D microstructure. The results were partially successful in that both grippers were able to pick up corresonponding parts and bring them out of plane in order to make them stand up. However, a final 3D microstructure was unfortunately not achieved due to time constraints. This will be left to future researchers who continue the project. On the equpiment side a microassembly system was fully integrated using cameras for vision and motors with micro-resolution for movement. A computer program was used to control each part of the system. The cameras provided feedback from various views, allowing the operator to observe what was happening to the microcomponents. The grippers were attached to one of the motors and manipulated to pick up the parts. The final overall system proved sufficient for microassembly, but had some areas that could be improved upon.
3D Mesh and Pose Recovery of a Foot from Single Image
(University of Waterloo, 2022-01-18) Boismenu-Quenneville, Frédéric
The pandemic and the major shift to online shopping has highlighted the current difficulties in getting proper sizing for clothing and shoes. Being able to accurately measure shoes using readily available smartphones would help in minimizing returns and trying to get a better fit. Being able to reconstruct the 3D geometry of a foot irregardless of the foot pose using a smartphone would help for the online shoe shopping experience. Usually, systems reconstructing a 3D foot require the foot to be in a canonical pose or require multiple perspectives. There is no system to our knowledge that allows capturing the precise pose of the foot without expensive equipment. In many situations, the canonical pose or the multiple views are not feasible. Therefore, we propose a system that can infer the 3D reconstruction and the pose estimation of the foot from any pose in only one image. Our kinematic model, based on popular biomechanical models, is made of 18 rotating joints. To obtain the 3D reconstruction, we extract the silhouette of the foot and its joint landmarks from the image space. From the silhouette and the relation between each joint landmark, we can define the shape of the 3D mesh. Most 3D reconstruction algorithms work with up-convolutions which do not preserve the global information of the reconstructed object. Using a template mesh model of the foot and a spatial convolution network designed to learn from sparse data, we are able to recover the local features without losing sight of the global information. To develop the template mesh, we deformed the meshes of a dataset of 3D feet so they can be used to design a PCA model. The template mesh is the PCA model with no variance added to its components. To obtain the 3D pose, we have labelled the vertices of the template mesh according to the joints of our kinematic model. Those labels can be used to estimate the 3D pose from the 3D reconstruction by corresponding the two meshes. To be able to train the system, we needed a good dataset. Since, there was no viable one available, we decided to create our own dataset by using the previously described PCA model of the foot to generate random 3D meshes of feet. We used mesh deformation and inverse kinematics to capture the feet in different poses. Our system showed a good ability to generate detailed feet. However, we could not predict a reliable length and width for each foot since our virtual dataset does not support scaling indications of any kind, other than the ground truths. Our experiments led to an average error of 13.65 mm on the length and 5.72 mm on the width, which is too high to recommend footwear. To ameliorate the performance of our system, the 2D joints detection method could be modified to use the structure of the foot described by our kinematic foot model as a guide to detect more accurately the position of the joints. The loss functions used for 3D reconstruction should also be revisited to generate more reliable reconstructions.
3D optical metrology by digital moiré: Pixel-wise calibration refinement, grid removal, and temporal phase unwrapping
(University of Waterloo, 2017-01-24) Mohammadi, Fatemeh
Fast, accurate three dimensional (3D) optical metrology has diverse applications in object and environment modelling. Structured-lighting techniques allow non-contacting 3D surface-shape measurement by projecting patterns of light onto an object surface, capturing images of the deformed patterns, and computing the 3D surface geometry from the captured 2D images. However, motion artifacts can still be a problem with high-speed surface-motion especially with increasing demand for higher measurement resolution and accuracy. To avoid motion artifacts, fast 2D image acquisition of projected patterns is required. Fast multi-pattern projection and minimization of the number of projected patterns are two approaches for dynamic object measurement. To achieve a higher rate of switching frames, fast multi-pattern projection techniques require costly projector hardware modification or new designs of projection systems to increase the projection rate beyond the capabilities of off-the-shelf projectors. Even if these disadvantages were acceptable (higher cost, complex hardware), and even if the rate of acquisition achievable with current systems were fast enough to avoid errors, minimization of the number of captured frames required will still contribute to reduce further the effect of object motion on measurement accuracy and to enable capture of higher object dynamics. Development of an optical 3D metrology method that minimizes the number of projected patterns while maintaining accurate 3D surface-shape measurement of objects with continuous and discontinuous surface geometry has remained a challenge. Capture of a single image-frame instead of multiple frames would be advantageous for measuring moving or deforming objects. Since accurate measurement generally requires multiple phase-shifted images, imbedding multiple patterns into a single projected composite pattern is one approach to achieve accurate single-frame 3D surface-shape measurement. The main limitations of existing single-frame methods based on composite patterns are poor resolution, small range of gray-level intensity due to collection of multiple patterns in one image, and degradation of the extracted patterns because of modulation and demodulation processes on the captured composite pattern image. To benefit from the advantages of multi-pattern projection of phase-shifted fringes and single-frame techniques, without combining phase-shifted patterns into one frame, digital moiré was used. Moiré patterns are generated by projecting a grid pattern onto the object, capturing a single frame, and in a post-process, superimposing a synthetic grid of the same frequency as in the captured image. Phase-shifting is carried out as a post-process by digitally shifting the synthetic grid across the captured image. The useful moiré patterns, which contain object shape information, are contaminated with a high-frequency grid lines that must be removed. After performing grid removal, computation of a phase map, and phase-to-height mapping, 3D object shape can be computed. The advantage of digital moiré provides an opportunity to decrease the number of projected patterns. However, in previous attempts to apply digital phase-shifting moiré to perform 3D surface-shape measurement, there have been significant limitations. To address the limitation of previous system-calibration techniques based on direct measurement of optical-setup parameters, a moiré-wavelength based phase-to-height mapping system-calibration method was developed. The moiré-wavelength refinement performs pixel-wise computation of the moiré wavelength based on the measured height (depth). In measurement of a flat plate at different depths, the range of root-mean-square (RMS) error was reduced from 0.334 to 0.828 mm using a single global wavelength across all pixels, to 0.204 to 0.261 mm using the new pixel-wise moiré-wavelength refinement. To address the limitations of previous grid removal techniques (precise mechanical grid translation, multiple-frame capture, moiré-pattern blurring, and measurement artifacts), a new grid removal technique was developed for single-frame digital moiré using combined stationary wavelet and Fourier transforms (SWT-FFT). This approach removes high frequency grid both straight and curved lines, without moiré-pattern artifacts, blurring, and degradation, and was an improvement compared to previous techniques. To address the limitations of the high number of projected patterns and captured images of temporal phase unwrapping (TPU) in fringe projection, and the low signal-to-noise ratio of the extended phase map of TPU in digital moiré, improved methods using two-image and three-image TPU in digital phase-shifting moiré were developed. For measurement of a pair of hemispherical objects with true radii 50.80 mm by two-image TPU digital moiré, least-squares fitted spheres to the measured 3D point clouds had errors of 0.03 mm and 0.06 mm, respectively (sphere fitting standard deviations 0.15 mm and 0.14 mm), and the centre-to-centre distance measurement between hemispheres had an error of 0.19 mm. The number of captured images required by this new method is one third that for three-wavelength heterodyne temporal phase unwrapping by fringe projection techniques, which would be advantageous in measuring dynamic objects, either moving or deforming.
Acausal Modelling of Thermal Fluid Systems with a Focus on Engine Air Path Components
(University of Waterloo, 2020-09-14) Keblawi, Amer
The automotive industry is rapidly developing more advanced vehicle propulsion systems, autonomous driving, emissions reduction, and improved fuel efficiency. Optimal control theory has evolved such that a system can be controlled in real-time based on values predicted by models over a control horizon. Controller performance depends on the accuracy of the predictive model. The purpose of this work is to develop acausal high-fidelity control-oriented plant models of engine air path components that simulate faster than real-time to design and tune the required controllers. The desired features of these models are that they be physics-based and use physically-meaningful design parameters. These features are desired to make these models extensible over different generations of the engine air path components. The considered engine in this work is the engine of the Toyota Prius 2015 Plug-in hybrid. An acausal mean value thermal engine block model is developed. The engine block model includes computations done on thermal effects in the engine block, engine air intake and exhaust streams, lubricant oil, and coolant. Model parameters are identified and validated experimentally. Acausal physics-based models, including spatial variation in variables, are introduced for the engine manifold and catalytic converter. The models are based on one-dimensional partial differential equations. A novel method based on orthogonal collocation is devised to model quasi-one-dimensional compressible flows inside engine manifolds. A similar methodology is used to model the catalytic converter, transforming the system of partial differential equations into a system of ordinary differential equations in state-space form. Both high fidelity models simulate faster than real-time. The developed models are showcased by designing and tuning a low-level engine shaft speed adaptive model predictive controller than can be used to control speed, e.g. in an adaptive cruise controller or autonomous car. The controller manipulates engine throttle and air to fuel ratio to achieve the desired engine shaft speed at minimal fuel consumption. The tuned Model Predictive controller is compared to a tuned PID controller by simulating a drive cycle followed by the Toyota engine.
Accessible Integration of Physiological Adaptation in Human-Robot Interaction
(University of Waterloo, 2021-09-21) Kothig, Austin
Technological advancements in creating and commercializing novel unobtrusive wearable physiological sensors have generated new opportunities to develop adaptive human-robot interaction (HRI). Detecting complex human states such as engagement and stress when interacting with social agents could bring numerous advantages to creating meaningful interactive experiences. Bodily signals have classically been used for post-interaction analysis in HRI. Despite this, real-time measurements of autonomic responses have been used in other research domains to develop physiologically adaptive systems with great success; increasing user-experience, task performance, and reducing cognitive workload. This thesis presents the HRI Physio Lib, a conceptual framework, and open-source software library to facilitate the development of physiologically adaptive HRI scenarios. Both the framework and architecture of the library are described in-depth, along with descriptions of additional software tools that were developed to make the inclusion of physiological signals easier for robotics frameworks. The framework is structured around four main components for designing physiologically adaptive experimental scenarios: signal acquisition, processing and analysis; social robot and communication; and scenario and adaptation. Open-source software tools have been developed to assist in the individual creation of each described component. To showcase our framework and test the software library, we developed, as a proof-of-concept, a simple scenario revolving around a physiologically aware exercise coach, that modulates the speed and intensity of the activity to promote an effective cardiorespiratory exercise. We employed the socially assistive QT robot for our exercise scenario, as it provides a comprehensive ROS interface, making prototyping of behavioral responses fast and simple. Our exercise routine was designed following guidelines by the American College of Sports Medicine. We describe our physiologically adaptive algorithm and propose an alternative second one with stochastic elements. Finally, a discussion about other HRI domains where the addition of a physiologically adaptive mechanism could result in novel advances in interaction quality is provided as future extensions for this work. From the literature, we identified improving engagement, providing deeper social connections, health care scenarios, and also applications for self-driving vehicles as promising avenues for future research where a physiologically adaptive social robot could improve user experience.
Adaptations of green growth and degrowth in an oil-dependent economy toward a better future
(University of Waterloo, 2023-10-11) AlAteibi, Muna
Throughout the history of Newfoundland and Labrador (NL), the province has been relying on natural resources as the main sources of economic production. Consequently, NL is prone to external shocks from demand and price fluctuations. For example, the collapse of fisheries during the 1990s and the fall in global oil prices during the 2008 financial crisis have had negative impacts on the NL socioeconomic system, increasing unemployment and out-migration rates. A lack of modeling studies in the literature related to NL natural resources dependency, unemployment, and migration is the motivation for this research. This research focuses on studying the impact of oil, as a major natural resource for NL, dependency on other industries within the economy, employment, and migration through implementing green growth and degrowth policies as an alternative to decoupling the natural resources dependency and shifting away from the region’s historical sources of economic growth. This research links econometric, input-output (IO), and agent-based modeling techniques as a novel combination of methodologies to study the impact of an oil-dependent economy using oil prices and production reduction rates (scenarios of green growth and degrowth) as exogenous variables. The data used in this empirical analysis is obtained from Statistics Canada. The results help create suggestions for policymakers to steer socio-economic policies toward developing their economy for a better future.
Adapting Evolutionary Approaches for Optimization in Dynamic Environments
(University of Waterloo, 2006) Younes, Abdunnaser
Many important applications in the real world that can be modelled as combinatorial optimization problems are actually dynamic in nature. However, research on dynamic optimization focuses on continuous optimization problems, and rarely targets combinatorial problems. Moreover, dynamic combinatorial problems, when addressed, are typically tackled within an application context.

In this thesis, dynamic combinatorial problems are addressed collectively by adopting an evolutionary based algorithmic approach. On the plus side, their ability to manipulate several solutions at a time, their robustness and their potential for adaptability make evolutionary algorithms a good choice for solving dynamic problems. However, their tendency to converge prematurely, the difficulty in fine-tuning their search and their lack of diversity in tracking optima that shift in dynamic environments are drawbacks in this regard.

Developing general methodologies to tackle these conflicting issues constitutes the main theme of this thesis. First, definitions and measures of algorithm performance are reviewed. Second, methods of benchmark generation are developed under a generalized framework. Finally, methods to improve the ability of evolutionary algorithms to efficiently track optima shifting due to environmental changes are investigated. These methods include adapting genetic parameters to population diversity and environmental changes, the use of multi-populations as an additional means to control diversity, and the incorporation of local search heuristics to fine-tune the search process efficiently.

The methodologies developed for algorithm enhancement and benchmark generation are used to build and test evolutionary models for dynamic versions of the travelling salesman problem and the flexible manufacturing system. Results of experimentation demonstrate that the methods are effective on both problems and hence have a great potential for other dynamic combinatorial problems as well.
Adaptive Affective Computing: Countering User Frustration
(University of Waterloo, 2013-03-28T13:49:45Z) Aghaei, Behzad
With the rise of mobile computing and an ever-growing variety of ubiquitous sensors, computers are becoming increasingly context-aware. A revolutionary step in this process that has seen much progress will be user-awareness: the ability of a computing device to infer its user's emotions. This research project attempts to study the effectiveness of enabling a computer to adapt its visual interface to counter user frustration. A two-group experiment was designed to engage participants in a goal-oriented task disguised as a simple usability study with a performance incentive. Five frustrating stimuli were triggered throughout a single 15-minute task in the form of complete system unresponsiveness or delay. An algorithm was implemented to attempt to detect sudden rises in user arousal measured via a skin conductance sensor. Following a successful detection, or otherwise a maximum of a 10-second delay, the application resumed responsiveness. In the control condition, participants were exposed to a “please wait” pop-up near the end of the delay whereas those in the adaption condition were exposed to an additional visual transition to a user interface with calming colours and larger touch targets. This proposed adaptive condition was hypothesized to reduce the recovery time associated with the frustration response. The experiment was successfully able to induce frustration (via measurable skin conductance responses) in the majority of trials. The mean recovery half-time of participants in the first trial adaptive condition was significantly longer than that of the control. This was attributed to a possibility of a large chromatic difference between the adaptive and control colour schemes, habituation and prediction, causal association of adaptation to the frustrating stimulus, as well as insufficient subtlety in the transition and visual look of the adaptive interface. The study produced findings and guidelines that will be crucial in the future design of adaptive affective user interfaces.
Adaptive Fringe Pattern Projection Techniques for Imgae Saturation Avoidance in 3D Surface Measurement
(University of Waterloo, 2010-10-29T21:20:35Z) Waddington, Christopher
Fringe-pattern projection (FPP) techniques are commonly used for surface-shape measurement in a wide range of applications including object and scene modeling, part inspection, and reverse engineering. Periodic intensity fringe patterns with a specific amplitude are projected by the projector onto an object and a camera captures images of the fringe patterns, which appear distorted by the object surface from the perspective of the camera. The images are then used to compute the height or depth of the object at each pixel. One of the problems with FPP is that camera sensor saturation may occur if there is a large change in ambient lighting or a large range in surface reflectivity when measuring object surfaces. Camera sensor saturation occurs when the reflected intensity exceeds the maximum quantization level of the camera. A low SNR occurs when there is a low intensity modulation of the fringe pattern compared to the amount of noise in the image. Camera sensor saturation and low SNR can result in significant measurement error. Careful selection of the camera aperture or exposure time can reduce the error due to camera sensor saturation or low SNR. However, this is difficult to perform automatically, which may be necessary when measuring objects in uncontrolled environments where the lighting may change and objects have different surface reflectivity. This research presents three methods to avoid camera sensor saturation when measuring surfaces subject to changes in ambient lighting and objects with a large range in reflectivity. All these methods use the same novel approach of lowering the maximum input gray level (MIGL) to the projector for saturation avoidance. This approach avoids saturation by lowering the reflected intensity so that formerly saturated intensities can be captured by the camera. The first method of saturation avoidance seeks a trade-off between robustness to intensity saturation and low SNR. Measurements of a flat white plate at different MIGL resulted in a trade-off MIGL that yielded the highest accuracy for a single adjustment of MIGL that is uniform within and across the projected images. The second method used several sets of images, taken at constant steps of MIGL, and combined the images pixel-by-pixel into a single set of composite images, by selecting the highest unsaturated intensities at each pixel. White plate measurements using this method had comparable accuracy to the first method but required more images to form the composite image. Measurement of a checkerboard showed a higher accuracy than the first method since the second method maintains a higher SNR when the object has a large range of reflectivity. The last method also used composite images where the step size was determined dynamically, based on the estimated percentage of pixels that would become unsaturated at the next step. In measurements of a flat white plate and a checkerboard the dynamic step size was found to add flexibility to the measurement system compared to the constant steps using the second method. Using dynamic steps, the measurement system was able to measure objects with either a low or high range of reflectivity with high accuracy and without manually adjusting the step size. This permits fully automated measurement of unknown objects with variable reflectivity in unstructured environments with changing lighting conditions. The methods can be used for measurement in uncontrolled environments, for specular surfaces, and those with a large range of reflectivity or luminance. This would allow a wider range of measurement applications using FPP techniques.
Adaptive Fringe Projection and Error-Compensated Calibration for Compact 3D Shape-Measurement Systems
(University of Waterloo, 2014-09-25) Li, Dong
Measurement of the three-dimensional (3-D) shape of an object is needed for both industrial and consumer applications. In industrial applications, compact measurement systems are needed to accomplish certain tasks such as measuring an interior surface in a confined space. In consumer applications, compact measurement systems are also needed for common consumers to conveniently get access to 3D data for a wide range of everyday uses. Fringe-projection techniques have been increasingly used for 3D shape measurement due to the advantage of dense full-field measurement. For a camera-projector measurement system, system geometry (the relative camera-projector position and angle) determine the system compactness. Analysis of the relation of system geometry to measurement accuracy is challenging owing to the effect of the various factors that vary with system geometry on measurement accuracy. It is thus necessary to experimentally determine how measurement accuracy varies with system geometry, in order to determine the most compact design that satisfies a desired measurement accuracy. This has been achieved in a compactness study, in which the measurement accuracy is evaluated at different relative camera-projector positions and angles. Measurement results in the compactness study have shown that there is a tradeoff in loss of accuracy for increased compactness and loss of compactness for increased accuracy. The smallest camera-projector angle (for an industrial system) or the smallest physical distance between the camera and projector (for a consumer system) that satisfies the desired accuracy would provide the most compact design. Several new methods including 1) an improved heterodyne phase-unwrapping method, 2) an adaptive fringe-pattern projection (AFPP) method for surfaces of high variation in reflectivity and illumination, and 3) a pixel-wise adaptive fringe-pattern projection (PWAFPP) method for such surfaces, have been developed in this research to improve measurement accuracy, thus contributing to enable a more compact system design to achieve a desired measurement accuracy. First, the new improved heterodyne phase-unwrapping method detects and compensates for the spike-like errors in absolute phase maps. The method has demonstrated improved projector calibration accuracy from 18.2 to 0.2 pixels, thus ensuring usable camera-projector stereovision system calibration for 3D measurement. Second, the new AFPP method adapts the projector maximum input gray levels (MIGLs) to local surface reflectivity using only two prior fringe-pattern projection and image-capture rounds. The method demonstrated greatly improved 3D measurement accuracy by avoiding image saturation in highly-reflective surface regions while maintaining high intensity modulation of captured fringe patterns across the entire surface with large range in reflectivity. Third, the new PWAFPP method projects a MIGL adapted to the surface reflectivity and illuminance for each pixel. The method has demonstrated a 34% root-mean-square (RMS) error reduction in 3D measurement for pixels that remained saturated after applying the AFPP method. The new method can thus be used to measure surfaces with more complex variation in surface reflectivity.
Addressing Data Scarcity in Domain Generalization for Computer Vision Applications in Image Classification
(University of Waterloo, 2024-08-30) Kaai, Kimathi
Domain generalization (DG) for image classification is a crucial task in machine learning that focuses on transferring domain-invariant knowledge from multiple source domains to an unseen target domain. Traditional DG methods assume that classes of interest are present across multiple domains (domain-shared), which helps mitigate spurious correlations between domain and class. However, in real-world scenarios, data scarcity often leads to classes being present in only a single domain (domain-linked), resulting in poor generalization performance. This thesis introduces the domain-linked DG task and proposes a novel methodology to address this challenge. This thesis proposes FOND, a "Fairness-inspired cONtrastive learning objective for Domain-linked domain generalization," which leverages domain-shared classes to learn domain-invariant representations for domain-linked classes. FOND is designed to enhance generalization by minimizing the impact of task-irrelevant domain-specific features. The theoretical analysis in this thesis extends existing domain adaptation error bounds to the domain-linked DG task, providing insights into the factors that influence generalization performance. Key theoretical findings include the understanding that domain-shared classes typically have more samples and learn domain-invariant features more effectively than domain-linked classes. This analysis informs the design of FOND, ensuring that it addresses the unique challenges of domain-linked DG. Furthermore, experiments are performed across multiple datasets and experimental settings to evaluate the effectiveness of various current methodologies. The proposed method achieves state-of-the-art performance in domain-linked DG tasks, with minimal trade-offs in the performance of domain-shared classes. Experimental results highlight the impact of shared-class settings, total class size, and inter-domain variations on the generalizability of domain-linked classes. Visualizations of learned representations further illustrate the robustness of FOND in capturing domain-invariant features. In summary, this thesis advocates future DG research for domain-linked classes by (1) theoretically and experimentally analyzing the factors impacting domain-linked class representation learning, (2) demonstrating the ineffectiveness of current state-of-the-art DG approaches, and (3) proposing an algorithm to learn generalizable representations for domain-linked classes by transferring useful representations from domain-shared ones.
Advancing the Extraction of Mechanical Properties from Biaxial Data
(University of Waterloo, 2023-08-24) Azzi, Carl
Mechanical characterization is vital to understand soft tissue behaviour in health and pathology. In aortic aneurysms, for instance, it is used to develop techniques for rupture risk assessment. In skin, for instance, it is used to assess the effects of freezing and anatomic location, the information useful for donor tissue banks. Planar biaxial testing is one of the most common tools for the mechanical characterization of soft tissues. In this experiment, loadings such as displacements or forces are applied to the edges of the square specimens yielding deformations at the center of the specimens that are measured digitally. Biaxial testing is a great tool as it captures anisotropy and nonlinearity of soft tissues’ behaviour and is capable of exploring a wide range of deformation states as it can apply different combinations of loadings. However, because the deformations at the center of the specimens are not controlled, no two mechanical tests are equivalent, complicating the consistency in data extraction and comparison. In this study, we propose a new approach for biaxial data analysis. First, a surface is fitted to the biaxial data. Second, the mechanical response is interpolated at the true equi-biaxial stretch deformation state (the state at which the deformations at the center of the sample are equal). Third, the effective mechanical properties such as high/low elastic moduli, and transition stretches/stress are extracted from the interpolated response. Other studies, in contrast, extract properties at the equi-biaxial displacement deformation state (the state at which equal strains are applied at the sample edges), which is due to the anisotropy of soft tissues and experimental setup, varies from specimen to specimen. We argue that our proposed approach of data extraction is more robust from the mechanical point of view. To demonstrate that our proposed approach can result in drastically different data sets, we apply it to previously tested aortic tissues from human donors and pigs. We compare effective mechanical properties extracted from the interpolated equi-biaxial stretch deformation state and conventionally used equi-biaxial displacement deformation state. Statistical analysis shows a significant difference between two groups of mechanical measures whether the measures are compared individually within each group or the general comparison of groups is conducted. Particularly, measures related to the transition zones (stress/stretch) linked to collagen fibres’ engagement behaviour were affected the most. Overall, the results indicate that the way the data is extracted can impact the outcome of biaxial studies. This further highlights the advantage of using the proposed approach of biaxial data extraction at equivalent deformation states versus the conventional approach. The proposed approach of the data extraction was also applied to human skin samples that came from the same donor’s back. Prior to biaxial testing, the samples were frozen/stored using three different freezing protocols (wet freezing in Phosphate Buffered Salin alone and with the cryoprotectant Glycerol as well as dry freezing using Liquid Nitrogen). Then, after testing, the effective mechanical properties were extracted and the effects of freezing and the anatomic locations were evaluated. We found very little quantitive evidence that the freezing storage approach mattered, although some qualitative observations were made to highlight the distinct behaviour of the samples frozen using Liquid Nitrogen. In the case of heterogeneity analysis, samples closer to the spine were different from samples further away from the spine, with the transition zone properties affected the most, especially for the samples subjected to Liquid Nitrogen freezing protocol. Future studies should assess each of the effect of heterogeneity and the effects of freezing separately, however, the overall approach of data extraction seems promising for intra-patient analysis.
Adversarial Machine Learning and Defenses for Automated and Connected Vehicles
(University of Waterloo, 2024-04-18) Zhang, Dayu
This thesis delves into the realm of adversarial machine learning within the context of Connected and Automated Vehicles (CAVs), presenting a comprehensive study on the vulnerabilities and defense mechanisms against adversarial attacks in two critical areas: object detection and decision-making systems. The research firstly introduces a novel adversarial patch generation technique targeting the YOLOv5 object detection algorithm. It presents a comprehensive study in the different transformations and parameters that change the effectiveness of the patch. The patch is then implemented within the CARLA simulation environment to assess robustness under varied real-world conditions, such as changing weather and lighting. With all the transformation applied during generation, the patch is able to reduce the confidence of YOLO5 detecting the stop sign by 70% comparing to the original stop sign if the lighting condition is good. However if the lighting condition is sub-optimal, for example, during a raining weather, the patch only reduce the confidence by 38% due to the patch being harder to be detected. Overall, the optimized patch still shows a greater effect on detection evasion compares to a random noise patch on any environment conditions. Overall, this part of the research showcase a novel way of generating adversarial patches and a new approach of testing the patches in a open-source simulator, CARLA, for better autonomous vehicle testing against adversarial attacks in the future. Simultaneously, this thesis investigates the susceptibility of Deep Reinforcement Learning (DRL) algorithms, in particular, Deep Q-Network (DQN) and Deep Deterministic Policy Gradient (DDPG) algorithms, to black-box adversarial attacks executed through zeroth-order optimization methods like ZO-SignSGD in a lane-changing scenario. The research first train the policies with finely turned hyper-parameters in the lane-changing environment and achieving a high performance. With a good policy as a base, the black-box attack successfully fooled both algorithms by optimally changing the state value to force the policy going straight while maintaining a small perturbation size compare to the original. While under attack, both DQN and DDPG are unable to perform, achieving an average of reward 108 and 45 comparing to their original performance of 310 and 232 respectively. A preliminary study on the effect of adversarial defense is also performed, which shows resistance against the attack and achieving slightly increase in average reward. This part of research uncovers significant vulnerabilities, demonstrating substantial performance degradation in DRL when used in the decision making of an autonomous vehicle. At last, the study underscores the importance of enhancing the security and resilience of machine learning algorithms embedded in CAV systems. Through a dual-focus on offensive and defensive strategies, including the exploration of adversarial training, this work contributes to the foundational understanding of adversarial threats in autonomous driving and advocates for the integration of robust defense mechanisms to ensure the safety and reliability of future autonomous transportation systems.
Affine and Regional Dynamic Time Warping
(University of Waterloo, 2014-11-11) Chen, Tsu-Wei Webber
Time series are a ubiquitous form of data prevalent in everyday life, and their analysis has gathered immense interest in many domains. Pointwise matches between two time series are of great importance in time series analysis, and dynamic time warping (DTW) has been widely known to provide reasonable matches. There are situations where time series alignment should be invariant to scaling and offset in amplitude or certain regions of a time series should be strongly reflected in the pointwise matches. Two different variants of DTW, affine DTW (ADTW) and regional DTW (RDTW), are proposed to handle scaling and offset in amplitude and regional emphasis respectively. Furthermore, ADTW and RDTW can be combined in two different ways to generate alignments that incorporate advantages from both methods. In global-affine regional DTW (GARDTW), the affine model is applied globally to the entire time series with regional emphasis, whereas in local-affine regional DTW (LARDTW), the affine model is applied locally to each region which are then emphasized. Alignments produced by the proposed methods are evaluated on simulated datasets and their associated difference measures are tested on real datasets. The proposed methods are found to significantly outperform DTW when an evaluated dataset meets the models or preferences of the proposed methods.
Age Differences in the Situation Awareness and Takeover Performance in a Semi-Autonomous Vehicle Simulator
(University of Waterloo, 2022-04-26) Murzello, Yovela
Research on young and elderly drivers indicates a high crash risk amongst these drivers in comparison to other age groups of drivers. Young drivers have a greater propensity to adopt a risky driving style and behaviors associated with poor road safety. On the other hand, age-related declines can negatively impact the performance of older drivers on the road leading to crashes and risky maneuvers. Thus, autonomous vehicles have been suggested to improve the road safety and mobility of younger and older drivers. However, the difficulty of manually taking over control from semi-autonomous vehicles might vary in different driving conditions, particularly in those that are more challenging. Hence, the present study aims to examine the effect of road geometry and scenario, by investigating young, middle-aged and older drivers' situation awareness (SA) and takeover performance when driving a semi-autonomous vehicle simulator on a straight versus a curved road on a highway and an urban non-highway road when engaged in a secondary distracting task. Due to the impact of COVID-19, data from only the young (n=24) and middle-aged (n=24) adults were collected and analyzed. Participants drove a Level 3 semi-autonomous simulator vehicle and performed a secondary non-driving related task in the distracted conditions. The results indicated that the participants had significantly longer hazard perception times on the curved roads and autopilot drives, but there was no significant effect of driver age and road type. Their Situation Awareness Global Assessment Technique (SAGAT) scores were higher in the highway scenarios, on the straight roads, and in the manual drive compared to the autopilot with distraction drive. Young drivers were also found to have significantly higher SAGAT scores than middle-aged drivers. While there was a significant interaction effect between road type and road geometry on takeover time, there was no significant main effect of road geometry, drive type and driver’s age. For the takeover quality metrics, road geometry and drive type had an effect on takeover performance. The resulting acceleration was higher for the straight road and in the autopilot drives, and the lane deviation was higher on the curved road and autopilot only drive compared to the autopilot with distraction drive. There was no significant main effect of road type and driver’s age on resulting acceleration and lane deviation. Overall, while there were age differences in some aspects of SA, young and middle-aged drivers did not differ in their takeover performance. The participants' SA was impacted by the road type and geometry and their takeover quality varied according to the road geometry and drive type. The outcomes of this research will aid vehicle manufacturing companies that are developing Level 3 semi-autonomous vehicles with appropriately designing the lead time of the takeover request to meet the driving style and abilities of younger and middle-aged drivers. This will also help to improve road safety by reducing the crash rate of younger drivers.
Age-Related Changes in Vibro-Tactile EEG Response and Its Implications in BCI Applications: A Comparison Between Older and Younger Populations
(IEEE, 2019-04) Chen, Mei Lin; Fu, Dannie; Boger, Jennifer; Jiang, Ning
The rapid increase in the number of older adults around the world is accelerating research in applications to support age-related conditions, such as brain-computer interface (BCI) applications for post-stroke neurorehabilitation. The signal processing algorithms for electroencephalogram (EEG) and other physiological signals that are currently used in BCI have been developed on data from much younger populations. It is unclear how age-related changes may affect the EEG signal and therefore the use of BCI by older adults. This research investigated the EEG response to vibro-tactile stimulation from 11 younger (21.7 ± 2.76 years old) and 11 older (72.0 ± 8.07 years old) subjects. The results showed that: 1) the spatial patterns of cortical activation in older subjects were significantly different from those of younger subjects, with markedly reduced lateralization; 2) there is a general power reduction of the EEG measured from older subjects. The average left vs. right BCI performance accuracy of older subjects was 66.4 ± 5.70%, 15.9% lower than that of the younger subjects (82.3 ± 12.4%) and statistically significantly different (t(10) = -3.57, p = 0.005). Future research should further investigate age-differences that may exist in electrophysiology and take these into consideration when developing applications that target the older population.
Age-related effects on EEG and Brain-computer Interface Classification
(University of Waterloo, 2018-08-13) Chen, Mei Lin
There is a rapid increase in the number of older adults around the world. This directly translates to an increase in the number of people living with health complications that are more prevalent in the elderly population, such as post-stroke conditions. Current rehabilitation techniques for stroke and other disorders are limited in effectiveness, which calls for the development of new approaches such as brain-computer interface (BCI) applications in neurorehabilitation. The majority of BCI applications are based on electroencephalogram (EEG) and other physiological signals to detect user intention and provide feedback. However, many of the signal processing algorithms currently used have been developed on data from a much younger population. There is a need to investigate how age-related changes directly affect EEG signals and extend to BCI control, specifically for older adults. In this thesis research, EEG response to vibro-tactile stimulation from 11 younger adults (21.7±2.76 years old) and 11 older adults (72.0±8.07 years old) were investigated. The results showed that, firstly, the spatial pattern of the cortical activation in older subjects was significantly different from that in younger adults (older adults had a reduced lateralization in activation); and secondly, there is a general overall power reduction in the EEG from older adults compared to younger adults. This suggests that the approach for designing BCI applications for older adults must be fundamentally differently than that for younger adults. This need is further shown in the average BCI performance accuracy classifying left vs. right was 64.5±7.75% for the older adults, which was more than 20% lower and statistically different (t(20)= -4.3, p <0.001) than that in the younger subjects, which was 85.3±14.1%. Compared to current works in the field, this research is unique in its examination of age-related differences in EEG signals and is the only work we are aware of that examines the age-related differences in EEG response to vibro-tactile stimulation. This finding should be further investigated with other BCI paradigms such as motor imagery in order to confirm the impact of age on BCI control. Further, provided that this age-related difference persists across different modalities, it is then necessary to fine-tune the algorithmic approaches to fit the intended application to the target population.
Age-Sensitive Features for Detection of Muscle Fatigue using the High-Density Electromyogram
(University of Waterloo, 2022-01-18) Krishnan, Bharath
The processes behind fatigue development within the muscles have been a topic of interest for exercise scientists for decades. This is because fatigue is one of the primary reasons for a decrease in performance and increase in likelihood of injury during exercise[1]. Typically, muscle fatigue is detected through modifications of the amplitude and spectral characteristics of a surface electromyogram (sEMG), or the variability of torque signals recorded throughout a sustained contraction. However, the behaviour of these parameters with the generation of fatigue depends on a variety of factors. One major factor is age, where the age-related loss of muscle fibers, and changes in neuromuscular system impact how muscles adapt to and develop fatigue. The purpose of this study was to examine age-sensitive High Density Surface Electromyogram (HD-sEMG) features and investigate the effect of spatial filter type on intramuscular coherence analysis in fatigue detection. Fatiguing submaximal isometric contractions of the bicep brachii was performed by eight young (24.40 ± 2.42 years) and five elderly (72.90 ± 2.21 years) males, while HD-sEMG recorded signals from the biceps brachii and a dynamometer recorded torque signals. The task was performed at 20% maximal voluntary contraction (MVC). From the HD-sEMG signals, the mean intramuscular coherence was calculated in the alpha (11-15Hz), beta (16-29Hz), and gamma (30-50Hz) frequency bands each of which stems from different neurological origins. Statistical differences were only found in the alpha (p=0.0006), and beta (p=0.0207) bands between the pre-and post-fatigue conditions of the young group. Furthermore, a correlation between mean coherence and torque variability during the final 25% of the contraction before task failure revealed that both the age groups had positive correlation in the alpha band. Different correlations were found in the beta and gamma bands, with positive correlations being observed in the elderly group and negative correlations in the young group. These results suggest that age-related changes in the corticospinal pathway exist causing the elderly to be less fatigable when compared to the young population. This proposes that the introduced intramuscular coherence analysis can be used to obtain fatigue related features from HD-sEMG signals that are age-sensitive.

Browse

Browsing Systems Design Engineering by Title