Liquid crystal elastomers (LCEs) are a class of stimuli-responsive materials capable of reversible mechanical actuation and thus hold promise for soft fluidic control devices. This study focuses on the development of photothermally responsive LCE fibers and their applications in soft pumps, regulating valves, and adaptive channels. Particular attention is given to the effects of photothermal dye type and concentration on the actuation performance of LCEs, with the aim of establishing a foundation for their use in soft robotics. LCE films were prepared through ink formulation, mixing, and UV curing. Their heating behavior and response efficiency were evaluated by thermal imaging and controlled light-irradiation experiments. On this basis, dye compositions with superior performance were identified, and LCE fibers were subsequently fabricated by extrusion. Resulting fibers were characterized by mechanical testing to quantify actuation stress and strain, response time, and cyclic stability. Optimized LCE fibers were integrated into prototypes to achieve light-controlled regulation of fluid direction and flow rate. The results demonstrate that both the type and concentration of photothermal dyes significantly influence the photothermal conversion efficiency and actuation behavior of LCEs. This study clarifies key factors governing the material design, fabrication, and fluidic device integration of LCE fibers, and provides experimental and theoretical support for advancing LCE-based soft fluidic devices and soft robotic systems.

Catechol-based bioadhesives, inspired by the versatile wet adhesion of marine organisms, have emerged as a promising class of materials capable of adhering strongly in aqueous and biological environments. The catechol functional group, found in mussel foot proteins (DOPA), enables both interfacial bonding (hydrogen bonding and metal coordination) and cohesive crosslinking (oxidation to quinones). However, achieving optimal adhesion under biological conditions, containing salts, proteins, and fluctuating redox and pH conditions, remains an ongoing challenge in biomaterial design. My project focuses on triblock polycatechol (tbPC) systems, synthetic polymer analogs to the mussel-inspired proteins that combine tunable block-copolymer structures with catechol chemistry. By systematically varying block lengths and catechol oxidation levels, I aim to understand how tbPC adhesives perform across biologically relevant surfaces and in fluids that mimic physiological and marine environments. These results will hopefully guide the design of biocompatible, stimuli-responsive sealants capable of functioning on wet tissue or underwater surfaces.

Thermal management in advanced electronics is critical as device dimensions shrink and power densities increase. This research project investigates the thermal transport properties of novel electronic materials, specifically three advanced polymer candidates (4D, 5D, and 6D), through experimental characterization and critical analysis. Thermal diffusivity measurements were conducted using a Laser Flash Analyzer (LFA 467 HyperFlash) on single and double-layer samples. Highly Oriented Pyrolytic Graphite (HOPG), Molybdenum, and Silicon wafers were utilized as reference standards to validate instrument accuracy across a range of temperatures from 20∘C to 127∘C. Experimental results for the 4-Polymer-D (4D) thin film yielded a thermal diffusivity of 0.124 mm2/s, while comparison measurements of SiO2 on Si resulted in 0.011 mm2/s. However, measurements for 5D and 6D polymers resulted in unphysical diffusivity values (>200 mm2/s), suggesting the need for improved sample preparation or modified calculation models for extremely thin organic layers on high-diffusivity substrates. These findings contribute to the fundamental understanding of phonon-mediated heat transport in organic electronics and provide a baseline for future material optimization. Next steps involve refining the measurement parameters for ultra-thin films to resolve the transient responses of 5D and 6D materials.

Scaling dexterous robot learning is constrained by the difficulty of collecting high-quality demonstrations across diverse operators. Existing wearable interfaces often trade comfort and cross-user adaptability for kinematic fidelity, while embodiment mismatch between demonstration and deployment requires visual post-processing before policy training. We present DexEXO, a wearability-first hand exoskeleton that aligns visual appearance, contact geometry, and kinematics at the hardware level. DexEXO features a pose-tolerant thumb mechanism and a slider-based finger interface analytically modeled to support hand lengths from 140 mm to 217 mm, reducing operator-specific fitting and enabling scalable cross-operator data collection. A passive hand visually matches the deployed robot, allowing direct policy training from raw wrist-mounted RGB observations. User studies demonstrate improved comfort and usability compared to prior wearable systems. Using visually aligned observations alone, we train diffusion policies that achieve competitive performance while substantially simplifying the end-to-end pipeline. These results show that prioritizing wearability and hardware-level embodiment alignment reduces both human and algorithmic bottlenecks without sacrificing task performance.

Robotic manipulation in greenhouse environments presents unique challenges owing to unstructured obstacles, diverse target appearances, and significant visual variability across spatial and temporal conditions. While imitation learning has shown promising results in robotic manipulation, existing approaches often assume relatively simple environments and lack mechanisms for obstacle-aware motion generation and failure recovery, limiting their robustness in agricultural settings. In this work, we propose GROVE, a decomposed diffusion-policy framework for greenhouse robot manipulation. Our approach separates the manipulation process into two stages: an alignment stage and an interactive stage. The alignment stage uses a conditional diffusion policy to generate end-effector trajectories toward a target-centered bottleneck pose while considering surrounding obstacles. Once the robot reaches this pose, the interactive stage executes the task-specific manipulation using demonstrations. To improve robustness in cluttered environments, we introduce a feedback-driven recovery mechanism that detects failures and adjusts the bottleneck pose to reattempt the task without additional demonstrations.