Small rover teams sharing compact spatial information on a planetary surface

Peter Zhi Xuan Li.

Energy-efficient spatial intelligence
through representation co-design.

MIT EECS PhD ’25 / Embedded Systems Engineer, Tycho AI

Image: Rover teams sharing maps on a planetary surface

RESEARCH VISION

Algorithm, spatial representation, and specialized hardware co-designAn algorithm graph, a spatial representation, and specialized hardware are joint design choices. Many translucent Gaussian ellipsoidal contours follow the room’s walls, floor, and furniture in the middle scene. Blue bidirectional links connect the algorithm, spatial representation, and specialized hardware.AlgorithmSpatialrepresentationSpecializedhardware
Image: Co-design of algorithms, spatial representations, and specialized hardware

Representation co-design for efficient autonomy.

A robot uses spatial models to locate itself, identify obstacles, and plan motion. Representation co-design widens algorithm–hardware co-design to include spatial representations shared across tasks and robots. These representations determine the structure and volume of the data robots store, process, and exchange. Redesigning them can reduce data movement, which can dominate computing energy.

My research extends representation co-design from efficient computing within a robot to efficient communication across collaborating robots.

Explore the research program DISCOVER THE WORK↓
Research overview

LIVE MAPPING DEMONSTRATION

Real-time 3D mapping in fabricated silicon.

Inside Gleanmer

Gleanmer implements GMMap on a system-on-chip (SoC), constructing 3D occupancy maps in real time under 6 mW of core power. At comparable map accuracy, our VLSI paper reports 44× lower power than the prior state-of-the-art mapping accelerator. It also reports at least 341× lower power than NVIDIA Jetson TX2 running GMMap.

Gleanmer at the MIT Stata Center

Gleanmer converts depth images into occupied-space and free-space Gaussians in real time. The demonstrated transfer rate is interface-limited; chip core power remains below 6 mW.
Open on YouTube

SELECTED MILESTONES

Research, recognition, and teaching.

2026 / RESEARCH

Gleanmer receives the VLSI Best Demo Award.

A live demonstration of real-time Gaussian occupancy mapping in fabricated silicon, with Zih-Sing Fu and collaborators.

See the project
2026 / COMMUNITY

Selected for RSS Pioneers.

Joining the early-career research cohort at Robotics: Science and Systems.

View the 2026 cohort
2022–2025 / TEACHING

Connecting digital design concepts with practice.

Final-project tutorials and SystemVerilog recitations connect implementation steps to course concepts.

Teaching and mentoring
Peter Zhi Xuan Li

ABOUT ME

Co-design across robotics
and specialized hardware.

I am an embedded systems engineer at Tycho AI, where I lead the specialized hardware team. I received my PhD in Electrical Engineering and Computer Science from MIT in 2025, advised by Vivienne Sze and Sertac Karaman. My research centers on representation co-design: designing spatial representations, algorithms, and hardware together for energy-efficient robot autonomy. Recognition includes selection as an RSS Pioneer (2026), the VLSI Best Demo Award (2026), and graduate first place in the ACM SRC Grand Finals (2020).