Liquid crystal elastomers (LCEs) are soft stimuli-responsive materials that contract along the orientation of mesogen, called director, upon heating due to a transition from the nematic to isotropic phase. Direct Ink Writing 3D printing allows fabrication of LCE structures with variable spatial orientation and order parameter due to the shear forces acting during the extrusion process. An understanding of complex LCE orientations is yet to be achieved to get extreme shape morphing, snapping and locomotion. We designed structures with complex print patterns by developing custom G-Codes. We prepare a standard LCE ink by mixing the mesogen, RM-82, cross-linker, n-Butylamine, and a photoinitiator, HHMP, and oligomerizing the mixture at high temperatures. To reduce the temperature-dependent viscosity of the ink during printing, we fabricate an in-house syringe heater that uniformly heats the ink. UV curing was performed during and after printing to fix the director orientation. We first printed several unidirectional rectangular specimens to characterize the effect of nozzle velocity and syringe pressure on the actuation stretch. Higher printing velocities and pressures result in greater shear forces, which produce higher actuation shrinkage. We also found that the shrinkage increases with the actuation temperature. More complex shapes such as a disk with spiral print pattern were printed, which actuates from a planar disc to a 3D cone, due to circumferential shrinkage and radial expansion. We also computationally simulated bilayer structures that can snap to a new configuration and instantly release energy in this process.

Low Density Parity Check (LDPC) codes are linear block codes with high throughput and error-correction capabilities, making them relevant for transmission of information over constrained or noisy transmission channels in communications. Iterative message passing algorithms are used to decode LDPC codes passed between variable nodes corresponding to the received channel bits and check nodes. While Sum-Product Algorithm (SPA) achieves high decoding performance at the expense of high computational complexity, reduced complexity decoding algorithms such as Min-Sum Algorithm (MSA) meanwhile can suffer decoding performance degradation as a tradeoff for simplified computation of outgoing check node messages. Through an implementation of a modified MSA with correction term, also known as min* algorithm, in a LDPC decoding on both software and hardware, we seek to demonstrate and verify improved decoding performance over MSA. The min* LDPC decoder is implemented in a MATLAB testing script via a forward-backward algorithm -based message passing operation, demonstrating clear improved performance over standard min-sum decoding in both bit-error-rate (BER) tests for parameters of 100 trials, 50 maximum iterations and frame-error rate (FER) tests measures across set parameters of 20 error frames calculation threshold, 20 maximum iterations. BER and FER curves generated from these tests verify the postulation of improved decoding performance in min*- based LDPC decoding over min-sum LDPC decoding. Future work is expected to continue on the development of a hardware-based min* LDPC decoder that is being written in SystemVerilog to run on a ZCU106 FPGA.

Brain-computer interfaces (BCI) enable humans to control technology, such as prostheses, through imagined movement. The highest performing BCIs require brain implants that read signals from individual neurons, a surgery that can be risky and unaffordable. Electroencephalography (EEG) is a non-invasive alternative used to detect neural activity from one’s scalp with a wearable cap that contains 64 electrodes. While more accessible, this technology is underutilized due to the challenge of shifting cap positions and external hair artifacts that add impedance to the EEG signals. To develop a robust decoder that interprets the variant signals and translates them into intended directions of motion, we repeatedly trained an EEGNet convolutional neural network across multiple sessions of closed loop BCI control. We created a computer game where the user moves the cursor left, right, up, and down by imagining their left hand, right hand, tongue, and feet moving, respectively. After testing a variety of methods, we found success using target acquisition, in which the user uses all four directions to move the cursor to the goal. Rotation decoding maps the EEG inputs to their intended directions using the cursor’s relative position to the target. With this technique, our subject achieved a high validation accuracy of 85% and consistently controls movement on a 2-D plane. After validating this among a variety of subjects, we wish to apply this signal processing to 3-D systems, which will enable control over prostheses in a manner as natural as moving one’s own body.

5G technology has been enabled by recent advances in coding theory, such as polar codes. However, many applications remain out of reach. Without sufficiently low error-rate and low latency, a class of applications, dubbed “mission-critical” applications due to their strict error and latency requirements, remain out of reach of current wireless communication technology. However, much progress has been made to close this gap. One such area, list-decoding, is a subject of research at Professor Wesel’s Communications Systems Laboratory. It has already been shown that both high-rate and low-rate zero-terminated and tail-biting convolutional codes (ZTCCs and TBCCs) with cyclic- redundancy-check (CRC)-aided list decoding techniques closely approach the random-coding union (RCU) bound for short blocklengths. However, current program implementations have limited our ability to design higher-rank CRCs in these papers. In our research, we use software engineering techniques to improve the performance of the current CRC search algorithm and mitigate an important memory-bandwidth bottleneck. We are then able to use these performance improvements to design higher-rank CRCs.

In the 1900s, photographer Eadweard Muybridge rigged 12 consecutive cameras with a tripwire to produce sequential pictures depicting a horse's motion. These images were monumental in proving that horses amid their gallop are momentarily airborne, but more importantly, they developed the concept known as time-resolved imaging. Today's frontier in sequential motion photography resides on fundamental questions about quantum dynamics utilizing X-ray free electron lasers (XFELs) to understand interactions between light and matter on the femtosecond scale–a quadrillionth of a second. This is game-changing for our basic understanding of nature’s smallest, fastest, and most elusive constituents that play a fundamental role in chemistry, biology, and quantum physics. High-quality electron beams are at the heart of these unique sources, which currently rely on decades-old technologies, thus hampering their advancement. Our project explores emerging theories in quantum electrodynamics combined with nonlinear optical techniques (e.g. four-wave mixing) to enhance the quality of electron beams and XFELs dramatically. Through vigorous calculations and computer simulations, we first prove physical theories and follow by building instrumentation informed by this theory. This sparked the design of our hollow-core fiber system, where high-energy light can be tailored to travel according to a wider range of parameters while maintaining its integrity, differing from traditional optical approaches. By completely redesigning the mechanisms of our electron source, we reach the potential for higher peak energies in our beamlines than ever before, creating a generation of unprecedented XFEL technology.