This project focuses on developing a high-fidelity digital twin of the UCLA campus to advance autonomous vehicle research within the UCLA Mobility Lab. This quarter, we completed the RoadRunner map, establishing a fully validated and simulation-ready roadway network as the foundation of the digital twin. Using built-in visualization tools, manual inspection, and topology checks, we resolved geometry and connectivity issues including lane height discontinuities, misaligned intersections, inconsistent lane directions, and discontinuous segments. With the roadway network finalized and verified, our efforts have transitioned to constructing the virtual UCLA environment in Blender. We are now developing detailed 3D models of campus buildings, surrounding infrastructure, terrain features, and environmental assets to accurately replicate the physical campus layout. These assets are being optimized for scale, coordinate alignment, and compatibility to ensure seamless integration with the completed road network. The virtual world will then be imported into CARLA, where traffic signals, dynamic actors, and autonomous vehicle scenarios will be implemented. With the map complete and world-building underway, the project has entered the full-scene development phase, positioning the UCLA digital twin for advanced simulation testing and AV validation.

This abstract has been withheld from publication.

Magnetic topological insulators provide a route to control topological electronic states by altering their magnetic order. Here, we report a sharp magnetic-field-induced electronic phase-transition in the bulk magnetic topological insulator MnSb4Te7, observed in real space using low-temperature scanning tunnelling microscopy and spectroscopy, and explained by our first-principles DFT+U calculations. Our calculations identify MnSb4Te7 as an axion insulator in both antiferromagnetic and ferromagnetic configurations. At zero field, we observe a small, spatially homogeneous gap of approximately 5 meV pinned to the Fermi level, which is consistent with an antiferromagnetic axion-insulating ground state. Upon applying an out-of-plane magnetic field, the electronic spectrum remains largely unchanged until a threshold value near 6 T, where it undergoes an abrupt renormalization: the low-energy gap at the Fermi level increases discontinuously to a value of approximately 80 meV. Concomitantly, a second, large gap opens at the surface Dirac point roughly 180 meV above the Fermi level. This transition is accompanied by a qualitative reconstruction of quasiparticle-interference patterns. The magnitude and coincidence of these two field-induced gaps cannot be attributed solely to a simple Zeeman or exchange-driven splitting of Dirac states, and instead point to an electronic phase transition. Such a field-induced transition is unprecedented in an axionic material and can signal the emergence of a sub-leading order as discussed in the cont

Magnetic topological insulators provide a route to control topological electronic states by altering their magnetic order. Here, we report a sharp magnetic-field-induced electronic phase-transition in the bulk magnetic topological insulator MnSb4Te7, observed in real space using low-temperature scanning tunnelling microscopy and spectroscopy, and explained by our first-principles DFT+U calculations. Our calculations identify MnSb4Te7 as an axion insulator in both antiferromagnetic and ferromagnetic configurations. At zero field, we observe a small, spatially homogeneous gap of approximately 5 meV pinned to the Fermi level, which is consistent with an antiferromagnetic axion-insulating ground state. Upon applying an out-of-plane magnetic field, the electronic spectrum remains largely unchanged until a threshold value near 6 T, where it undergoes an abrupt renormalization: the low-energy gap at the Fermi level increases discontinuously to a value of approximately 80 meV. Concomitantly, a second, large gap opens at the surface Dirac point roughly 180 meV above the Fermi level. This transition is accompanied by a qualitative reconstruction of quasiparticle-interference patterns. The magnitude and coincidence of these two field-induced gaps cannot be attributed solely to a simple Zeeman or exchange-driven splitting of Dirac states, and instead point to an electronic phase transition. Such a field-induced transition is unprecedented in an axionic material and can signal the emergence of a sub-leading order as discussed in the cont

This project focuses on the design and development of a small-scale humanoid robot through iterative modification of an existing platform. Instead of building a new system from scratch, the team investigates how structural redesign, actuator integration, and system optimization can improve performance and enable stable locomotion. The central question explores how mechanical configuration, particularly in the lower body, torso, and arms, affects load distribution, robustness, and modularity. The methodology follows a systems engineering approach, dividing the robot into subsystems including actuators, structure, flesh, and reinforcement learning. Key efforts include redesigning the hip, ankle, and foot to increase degrees of freedom and mobility, and optimizing the torso for efficient packaging using carbon fiber or aluminum frame architectures. Existing actuators are reused and modified with custom gearing and parallel actuation. Prototyping involves CAD modeling, finite element analysis, and rapid fabrication such as 3D printing and machining. Results show successful actuator validation, improved structural designs, and early hardware prototypes, forming a foundation for future locomotion control. This project demonstrates a practical and scalable approach to humanoid robot design, advancing efficient mechanical integration and system performance.

As Moore’s Law slows due to the fundamental physical limits of 2D manufacturing, the semiconductor industry is shifting toward modular 3D stacked architectures known as "Chiplets." This transition requires new frameworks for design analysis, as standards like 3Dblox lack integrated cost analysis and simplified design rule checking for heterogeneous components. This research presents tools to bridge these gaps, specifically finalizing a conversion tool between 3Dblox and CATCH (Cost-Analysis Tool for Chiplets). We generate a library of eight base chiplets to create a suite of 20 benchmark 3Dblox designs to validate tool accuracy and performance. To ensure compatibility, these designs are validated through 3D IC standard tools such as Cadence Integrity. Specifically, we resolve alignment errors and generate LEF macro definitions to accommodate 3Dblox’s orientation and centering logic. We address problems including unique bump labeling, centering, location alignment, mirroring, and pitch alignment at chiplet interfaces. These have resulted in a working design suite, a GitHub release of the conversion scripts, and a pending release of the design suite database. Future work will include refining the conversion tool from 3Dblox to CATCH and enriching the benchmark suite with details such as per-bump net and signal names. Ultimately, this project enhances the modularity of 3D integrated circuits by providing an automated means to ensure design validity and cost-effectiveness across complex chiplet systems.