At first, current implementations of mixed reality are often very impressive because of the immediate nature of their holograms. Looking further into the application reveals the significant limitations of not only the lack of gestures representing abstract concepts such as pulling or sliding, but also the user's need to learn extensive grammar. Both these problems make wearable technologies almost useless as regular computers are far better and easier to use for engineering and design. In an attempt to rectify this problem, the research is focused on developing ways to represent these abstract gestures and improving the intuitiveness of these gestures to minimize learning. The first step was to break down gestures into three main parts: the positioning of certain joints in the hand, the rotation of the hand itself, and the vector of motion. Consulting a movement expert and other members of the LEMUR lab, a few motions for abstract representation were parameterized based on what felt natural. A program was developed to evaluate user gestures based on how close the input was to each of these sample motions to increase the intuitiveness of control. By implementing this program, a member of the lab not involved in the development process was able quickly to perform the abstract gestures with very little instruction. Despite this success, further research is needed to both increase the number of gestures and the quality of detection.

The design and fabrication of robotic devices often requires significant funds, time, and technical expertise. Print-and-fold origami-inspired robots aim to increase the accessibility of robot creation by introducing an inexpensive, streamlined, and simplified approach to making robots. Advantages inherent to print-and-fold robots are the existence of built-in hinge mechanisms and the ability to be stored and transported compactly. Bioinspired robots use natural solutions to solve problems or complete tasks. The goal is not to copy biology exactly, but rather to adapt biological designs to the medium of robotics. The value in using a natural solution for a design problem is that it has essentially been tested already, albeit in a different context. This research aims to adapt a natural solution specifically to the realm of print-and-fold robotic body designs. My focus was on the ability of organisms, such as the three-banded armadillo and wood louse, to roll up into a ball. The imagined use of such a mechanism would be to increase the functionality of a robot by providing not only the ability for it to crawl, but also for it to enclose itself around another object and roll. The creation process entailed designing a two-dimensional template in Silhouette Studio for the folding shell, printing, cutting, and scoring the design into a piece of cardstock, and finally folding the two-dimensional sheet of cardstock into the three-dimensional robot body. The rolling-up mechanism is actuated manually with a bit of fishing line.

Hydrogels are three-dimensional polymer networks that are crosslinked using various chemical or physical crosslinkers. Their toughness, as well as swelling without dilution in aqueous environments, make them attractive candidates for medical applications such as wound dressing, drug delivery, tissue engineering, and regeneration. Gelatin methacryloyl (also known as GelMA) is a common hydrogel that can be applied in biological and medical areas because it resembles some characteristics of the native extracellular matrix, which helps the cell to grow and spread. The methacrylate modification of gelatin allows this hydrogel to be covalently crosslinked upon exposure to ultraviolet light. However, the application of this material has been limited mainly due to its weak mechanical strength prior to crosslinking, leading to undesired flows on tissue surfaces. Here we propose to overcome the low mechanical strength of GelMA by incorporating oppositely charged block polyelectrolytes that self-assemble ionically when mixed in aqueous mediums, resulting in hybrid hydrogels with mechanical robustness. To quantify the strength of adhesion interactions between the substrate and the polymer, we built and used an ASTM standard burst pressure apparatus to measure the maximum pressure that hydrogel acquires to burst. We compared the burst pressure of GelMA as well as GelMA/polyelectrolyte systems in dry and wet environments and studied the influence of dilution on both systems.

Event-Based cameras measure the timestamps of discrete changes in pixel intensity and therefore have higher temporal resolution and lower energy use than frame-based cameras. This makes them extremely useful for tracking objects moving extremely quickly, such as aircraft. Convolutional neural networks, which are typically used for frame-based object tracking, require too much computation and are too slow for event-based vision. One solution is to use Gabor filters, which are convolutional filters that are tuned to detect edges, speed, and direction of motion. We wish to benchmark the tracking accuracy of Gabor filters and determine how much accuracy is traded off for speed in their usage. Objects in frame are identified using the density based spatial clustering with noise method. Gabor Filter information is used to predict the future position of these objects. Regions of interest are created around these predicted positions. In future frames, only regions of interest are convolved, significantly speeding up compute time and allowing for the tracking of multiple objects across time. We use the Measure of Tracking Accuracy metric, which utilizes the number of false positives, false negatives, and identity switches across all frames. Preliminary results show that using Gabor filters to identify objects yields a measure of tracking accuracy of 50.05, which is surprisingly higher than the average for conventional neural networks running on the same dataset, 44.09639. This indicates that using Gabor Filters is a valid way to track objects coming from an event-based camera datastream.

Human vision offers several cues for depth perception. Notably, motion parallax -- a phenomenon referring to the difference in horizontal displacement based on how far away an object in observance is -- offers an effective visual cue for humans to determine how distant certain objects may be. Similarly, this concept can be effectively applied to current technology to improve depth estimates from multiple images. As such, tools like YOLO's are extremely useful in locating the positions of objects in each frame so that they are clearly accessible for processing depth. Taking into account the applications of this technology, it's crucial that this form of depth perception is implemented into different types of hardware; in this case, it will be applied to a robotic car, a drone, and a laser engraver that moves in the x,y direction. In the progression of this application, Virtual Reality head motion is in the process of being synced into these different pieces of hardware to provide depth information to the user while they are in motion along an (x,y,z) as well as a pitch, yaw, and roll axis.