Xinyang Liu

• Negotiated cross-team requirements, spec-ed, designed, and generated drawings for various IP5X sheet metal electronic enclosures for mobile robot, eliminating over-heating and cable management issues of current iteration. • Supported, debugged, and maintained mechanical and electrical hardware for tests and demos at various deployment sites.

Mechanical and procedure design for drone-based magnetic field survey of terrestrial buried crater. Advisors: Prof. Anton Ermakov and Prof. Mac Schwager

• Built strong intuition behind application of solid mechanics in structural analysis through both FEA and energy-driven methods, failure theory, structural stability, and buckling behavior of material. • Lead weekly discussions, office hours, and online Q&A forum for a class of 80 grad students; design, write, and grade weekly structure analysis assignments and class exams; manage course logistics; substitute professor for lectures on occasion.

• Deployed codebase on delivery vehicle, design/planned/carried out real-world tests, and led Monte Carlo simulation campaign to generate performance metrics for cross-functional teams under different mission parameters. • Overcame cross-platform differences, engineered a general-purpose model analysis package tailored to in-house simulator for frequency-domain controller design, allowing 10+ man-hours saved per controller iteration. • Formulated PMBLDC motor model, carried out tests and sys-ID’ed using data replays, enhanced sim accuracy by 40%.

• Produced two publications with first co-authorship and second authorship, submitted to IEEE Transactions on Robotics. • Developed first-of-its-kind aerial robot transportation and manipulation simulator, capable of simulating any number of aerial robots carrying cable-suspended or rigidly-connected cargo with full system hybrid dynamics. • Architected simulator in an object-orientated way, enable easy configuration for SITL/HITL simulation; restructured codebase to run in a multi-thread fashion under ROS, allowing scalability and faster run time. • Formulated and developed highly scalable and distributed 6D force estimator and admittance controller, achieving force-sensor-free physical human-robot interaction; eliminated safety concerns through online formation optimization. • Deployed codebase onto in-house drone platforms, iterated system design through rigorous real-world testing in MOCAP-equipped indoor flying space, confirmed and improved system stability, robustness, and estimation accuracy.

Hold weekly office hours, provide timely online Q&A, design hands-on coding tutorials for students, and collaborate with other Course Assistants on grading homework/labs/exams. Course website: https://www.chrismusco.com/machinelearning2022/

Tutor Physics and Upper-level Undergraduate Mechanical Engineering courses, including Heat Transfer, Mechanics of Material, Measurement Systems, Automatic Control, Thermodynamics, FEA, Dynamics, and Statics.

• Integrated Computer Vision based mobile projective augmented reality with Unitree A1 via ROS, enabling the autonomous quadrupedal mobile projector to display concealed construction features with position-accurate features annotations. • Localized projector by fusing lidar and IMU data with UKF, increased real-time localization accuracy by 6%-13%.

• Modeled exhaust system heat transfer scenarios with empirical testing and FEA verification; proposed cost-effective cooling solutions maximizing heat transfer on a watt-per-dollar basis, contributing annual saving of $3k per exhaust section. • Streamlined production by automating exhaust duct diverter via pneumatic actuation; produced and tested different designs for various use cases via SOLIDWORKS; handed over to production; saving 300+ secs per cycle with improved safety. • Reduced production time cost by enabling remote high-temp furnaces monitoring with self-designed CV pipeline. • Optimized high-temperature furnace oxygen lance mechanical design and assembly layout utilizing large SOLIDWORKS assembly; halved restocking lead time and reduced part and labor cost by 20% without performance degradation.

• Worked as full-time SAT teaching assistant for a class of 4, responsible for answering students’ questions during self-studying period • Communicated with parents and teachers and led discussions during mock-examination commentary lessons.