Research
and Publications.

Problem Statement

Locomotion on deformable terrain, where robot–terrain motion coupling occurs, remains a gap in existing literature. Discrete probe-then-decide traversability strategies and deployment of legged robots in obstacle-dense environments like substations and construction sites, raise a further question: Can a robot successfully exploit a yielding obstacle field?

Methodology

At small-to-moderate deformation, yielding terrain behaves approximately like a spring. This motivates a spring loaded obstacle model. We designed and prototyped a novel locomotion test bed with spring-loaded obstacles mounted on a custom sliding mechanism to enable repeatable trials. Trials over variable obstacle spacings and stiffnesses with a caterpillar-inspired robot yielded observed trends in obstacle stiffness/spacing and traversal robustness.

Institution / Lab

RoboLand Lab, USC
Dr. Feifei Qian

Skills Practiced

Presented At

2026 SoCal REU Poster Symposium ↗

Materials

roboland poster (PDF) ↗

Problem Statement

Project 1: Communicating realtime, hyperlocal air quality is difficult with legacy sensor middleware and fragmented data pipelines. To inform the local community of air quality information distilled from an array of research-grade sensors, the Lab for Investigations of Local Air Quality (LILAQ) needs reliable, real-time meteorological data collection infrastructure.
Project 2: Additionally, indoor CO₂ levels impact occupant focus, alertness, and long-term health. However, popular commercial CO₂ indicators can cost upwards of $200.

Methodology

Project 1: Wrote BASH and IGOR scripts for Raspberry Pi and Windows devices to stream data to Firebase from PurpleAir and meteorological sensors on legacy middleware.
Project 2: Prototyped a CO₂ indicator with I2C interface to SCD41 sensor. Designed sensor enclosure in SolidWorks using edit-in-context assembly workflow.

Institution

Lab for Investigations of Local Air Quality, Harvey Mudd
Dr. Julie Medero and Dr. Lelia Hawkins

Skills Practiced

Problem Statement

Standard motion planners are designed for smooth continuous systems and fail on hybrid dynamical systems — those with both continuous flow and discrete jumps (e.g. a quadcopter bouncing off walls, a ball bouncing on a floor). No tools existed in the Open Motion Planning Library (OMPL) for this class of systems.

Methodology

Implemented cHyRRT and cHySST — two novel motion planners based on RRT and SST respectively — extended to handle hybrid dynamical systems with resets. Validated on collision-resilient quadcopter, bouncing ball, and pinball machine models. Results published and presented at IEEE CASE 2025. Both planners were accepted into OMPL.

Institution / Lab

Hybrid Systems Lab, UC Santa Cruz
Dr. Ricardo Sanfelice

Skills Practiced

Presented At

IEEE CASE 2025

Materials

OMPL planners source code (GitHub) ↗