DalSpace Institutional Repository

DalSpace is a digital service that collects, preserves, and distributes digital material produced by the Dalhousie community.

To learn about content guidelines, policies, and how to deposit, view the Help documents.
Contact us to get started submitting content to DalSpace at dalspace@dal.ca

Communities in DSpace

Select a community to browse its collections.

Now showing 1 - 10 of 19

Recent Submissions

Access status: Open Access ,
Evolving Linear Controllers from YoLo State Capture
(2026-04-23) Hu, Zhengping; Not Applicable; Master of Computer Science; Faculty of Computer Science; Not Applicable; Dr. Andrew McIntyre; Not Applicable; Dr. Yannick Marchand; Dr. Malcolm Heywood
End-to-end deep reinforcement learning (DRL) has become a prominent paradigm for visual control, with wide application in robotics and autonomous systems. However, its monolithic architecture often presents challenges regarding interpretability, computational overhead, and deployment on resource-constrained edge devices. This thesis investigates a decoupled perception-decision framework that combines real-time object detection (YOLOv11n), Heuristic State Rectification (HSR), and Genetic Algorithms (GA). Unlike standard DRL methods that rely on large convolutional networks to map raw pixels directly to actions, our methodology compresses high-dimensional visual inputs into physically semantic, low-dimensional states, empowering minimalist controllers (e.g., linear models with fewer than 50 parameters) to handle complex non-linear dynamics. To address the inherent heteroscedasticity of visual noise in camera-based perception, we introduce a Dynamic Seed Resampling mechanism. Acting as an adaptive regularization strategy, it prevents the agent from overfitting to specific environmental initializations, thereby enhancing the robust generalization of the trained agents. Extensive evaluations against a Proximal Policy Optimization (PPO) baseline across five classic control benchmarks (CartPole, Acrobot, MountainCar, Pendulum, and LunarLander) demonstrates the efficacy of our approach. While the PPO baseline exhibited sensitivity to initialization and encountered bottlenecks in sparse-reward environments, our decoupled framework achieved highly consistent success rates on discrete action tasks with significantly reduced computational cost and thermal output. Furthermore, through an analysis of spatial drift, we empirically quantify the ``Visual Noise Barrier.'' The results elucidate that while the Vision-HSR-GA framework excels in dynamic macroscopic control, achieving absolute zero-velocity precision is fundamentally bottlenecked by the Signal-to-Noise Ratio (SNR) of the frontend perception module. Ultimately, this research validates the proposed framework as a robust, interpretable, and hardware-friendly alternative for visual control tasks.
Access status: Embargo ,
Green technologies for the derivation of bioactive compounds from sea cucumber (Cucumaria frondosa) by-products and the subsequent development of gel platforms for biomedical purposes
(2026-04-23) Lin, Jianan; Not Applicable; Doctor of Philosophy; Department of Process Engineering and Applied Science; Not Applicable; Ozan Ciftci; Yes; Su-Ling Brooks; Guangling Jiao; Azadeh Kermanshahi-pour
The Atlantic sea cucumber, Cucumaria frondosa, generates substantial processing by-products that remain underutilized despite being rich in bioactive compounds and functional biopolymers. This dissertation establishes an integrated marine biorefinery framework that couples green sequential extraction with biomaterial engineering to valorize C. frondosa viscera into both high-value extracts and marine-derived delivery platforms, advancing circular marine bioprocessing for biomedical applications. Supercritical carbon dioxide (scCO2) was employed as a green solvent-based pretreatment to recover lipid fractions and enhance downstream accessibility of polar constituents. Subsequent extractions using ethanol (EtOH) and water significantly improved saponin recovery while preserving antioxidant activity. Sequential scCO2 extraction followed by a 24-hour extraction with 70% EtOH achieved saponin yields of 16.26 ± 2.47 mg OAE/g, comparable to the conventional hexane defatting-ultrasonic extraction (17.31 ± 0.60 mg OAE/g), while coupled 24-hour hot-water extraction yielded 12.99 mg OAE/g, demonstrating that scCO2-assisted sequential extraction provides an efficient and environmentally favourable alternative for saponin isolation. The residual biomass was further repurposed into sea cucumber polysaccharide crudes (SCPSC), which served as a marine-derived structural precursor for gel fabrication. To establish a scalable, crosslinker-free gelation platform, injectable and self-healing polyelectrolyte complex (PEC) hydrogels were developed using chitosan, marine chondroitin sulphate or algal fucoidan, hydrolyzed collagen, and cellulose nanocrystals (CNC). These physically crosslinked networks exhibited tunable swelling (~178-766%), low E-factors down to 0.98, and solid-like rheological behaviour (G' ≫ G"), with mesh sizes of ~5-10 nm and strength modulated through formulation control. Building on this framework, chitosan/SCPSC/CNC/sodium dodecyl sulfate (SDS)-based PEC porous networks were prepared via a semi-dissolution acidification sol-gel transition (SD-A-SGT) followed by poor-solvent-induced aggregation. Despite the compositional heterogeneity of the crude extracts, the resulting organogels formed robust physically crosslinked networks, as evidenced by G' ≫ G", with no crossover between 25 and 50 °, rubbery plateau moduli reaching ~30 kPa, and compressive Young’s moduli up to 18.36 kPa. Equilibrium swelling ratios ranged from ~170 to 770%, with CNC-reinforced formulations typically reaching ~290-350%, while low SDS-containing organogels showed exceptional uptake (~770%) driven by electrostatic chain expansion. Controlled solvent exchange and drying converted these organogels into aerogels and structure-preserving hexane-exchanged xerogels with tunable architectures and functions. Using curcumin as a hydrophobic model payload, the dried gels achieved encapsulation efficiencies up to 45.56 ± 5.37% and loading capacities up to 41.92 ± 3.85 mg/g, while densified conventional xerogels reached a maximum apparent loading of 45.91 ± 4.19 mg/g. Structural and thermal analyses confirmed suppression of curcumin crystallinity via nanoconfinement within the porous matrices. In vitro release studies demonstrated architecture-dependent kinetics, with aerogels reaching ~94% cumulative release within 24 h, compared with ~73% for densified xerogels, enabling tunable burst-to-sustained delivery profiles. Encapsulation also markedly improved curcumin stability against light and heat relative to free curcumin.
Access status: Open Access ,
An Urban Common On Gananoque’s Waterfront
(2026-04-15) Treanor, Roan; Not Applicable; Master of Architecture; School of Architecture; Not Applicable; NA; Not Applicable; Steve Parcell; David Correa; Cristina Verissimo
As rural waterfront economies in Ontario transition from industrial manufacturing to leisure tourism, public infrastructure is often misaligned with the needs of permanent residents. This thesis investigates how an urban common can respond to the intermittency and privatization of Ontario small-town waterfronts. It proposes that reprogramming underutilized industrial land through strategic design interventions and localized tactics can create an inclusive, year-round public realm that supports both residents and visitors. Gananoque, Ontario, serves as a testing ground, where seasonal tourism, an aging population, increasing privatization, and underused land generate social and spatial imbalance. Through a design-led methodology, this research develops a framework for integrating seasonal economies with permanent civic infrastructure. The resulting masterplan demonstrates how small towns can accommodate growth, reinforce local identity, and transform waterfronts from seasonal destinations into resilient urban common supporting diverse users throughout the year.
Access status: Open Access ,
Towards Understanding Multilevel Building Navigation for the Blind and Low Vision Individuals
(2026-04-17) Oladipupo, Ridwan; No; Master of Computer Science; Faculty of Computer Science; Received; n/a; No; Rita Orji; Yujie Tang; Rina Wehbe
Over 2.2 billion blind and low vision people globally struggle to navigate multi-level buildings independently. This thesis addresses this problem through three studies. First, we tested low-cost cameras for computer vision applications. Camera C5 per- formed well (0.96 accuracy, $27.99 cost), but all cameras required proper lighting to work. Without light, detection failed completely. Second, we interviewed 20 blind and low vision users about navigating buildings. Most (85%) could not find elevators in unfamiliar buildings without help, and 100% relied on sighted guides, even though they had good navigation skills. The problem was lack of information about where elevators were located. Third, we develop SmartEye, a prototype navigation system achieving 78.96% usability and 91.7% recommendation rates. However, evaluation reveals a critical “last-meter navigation gap”: accurate elevator detection alone fails to ensure successful call button location (r = 0.092). All 12 participants identified hands-free operation as non-negotiable, with guide dog and white cane users reporting physical impossibility of operating handheld devices while maintaining mobility aids. Key findings reveal that effective multi-level navigation requires: (1) hands-free wear- able form factors compatible with existing mobility aids, (2) multimodal audio and haptic feedback, (3) fine-grained directional guidance beyond proximity detection, and (4) adequate environmental lighting. Computer vision can help blind people find elevators in unfamiliar buildings, which is the first step toward accessible multi-level navigation.
Access status: Embargo ,
INTEGRATED BLOOD IMMUNOLOGY FROM VACCINATION TO CRITICAL ILLNESS: BIOMARKERS, MACHINE LEARNING, AND TRANSCRIPTOMIC PATHWAYS
(2026-04-17) Toloue Ostadgavahi, Ali; Yes; Doctor of Philosophy; Department of Microbiology & Immunology; Received; Dr. Ignacio Rubio; Yes; Dr. Christopher Richardson; Dr. Jean Marshall; Dr. Shashi Gujar; Dr. David J. Kelvin
Infection-related critical illness remains a major cause of morbidity and mortality in intensive care units, where COVID-19, bacteremia, sepsis, and septic shock often converge on similar clinical trajectories despite distinct etiologic triggers. Peripheral blood offers a practical window into these systemic responses. However, the molecular programs that distinguish protective immunity from maladaptive inflammation, and early sensing from later tissue injury, remain incompletely resolved. This publication-format thesis investigates blood immune activity along a continuum from controlled antigen exposure in vaccination to dysregulated host response in critical illness, aiming to define shared and syndrome-specific signals that are relevant for risk stratification and treatment. A central challenge in the field is that current clinical labels and single-analyte biomarkers capture hemodynamic consequences of disease better than they capture underlying biology, leading to limited reproducibility and weak guidance on when to apply host-directed interventions. To address this, the thesis integrates complementary blood-based approaches that connect functional humoral immunity, interpretable multiplex biomarker patterns at ICU presentation, and whole-blood transcriptomic pathways measured side by side across major ICU syndromes. By organizing these data into a unified, pathway-centered, time-aware framework, the work tests which programs are conserved across pathogen classes, which are syndrome-specific, and how they evolve with increasing severity. Across the integrated studies, heterologous SARS-CoV-2 vaccination elicits strong binding and neutralizing antibody responses, providing a quantitative benchmark for later blood-signal interpretation. Interpretable machine-learning analyses of ICU biomarker panels identify a small set of stable severity drivers across clinically overlapping infections, with programmed death ligand-1 and myeloperoxidase emerging as conserved indicators linked to worsening multi-organ failure. Comparative whole-blood RNA sequencing of adults with COVID-19, bacteremia, sepsis, and septic shock reveals a shared early core of innate sensing and cytokine activity, with syndrome-specific modulation of complement–coagulation and progressive engagement of metabolic stress, proteostasis, and cytoskeletal remodelling modules as illness severity increases. Together, these findings support a phased model of critical-illness biology and provide a practical scaffold for endotype- and timing-aware biomarker development and therapeutic trials.
Access status: Open Access ,
Conceptualizing Sexual Self-Schemas: A Review of Different Approaches and Their Implications for Understanding Women’s Sexual Function
(2021) Huberman, J. S; Bouchard, K. N.; Wang, G. A.; Rosen, N. O.