Research publications

Publications

Publications organized by the four GRAPPLE research modules.

Module 01

Environment Abstraction

Module overview ↗

CLEAR: A Semantic-Geometric Terrain Abstraction for Large-Scale Unstructured Environments

Pranay Meshram, Charuvahan Adhivarahan, Ehsan Tarkesh Esfahani, Souma Chowdhury, Chen Wang, and Karthik Dantu.

IEEE Robotics and Automation Letters (RA-L), 2026. Accepted August 2026.

CLEAR converts landcover and elevation maps into reusable, semantically aligned terrain regions connected as a graph for large-scale off-road planning. Evaluated on 9–100 km² digital terrain maps with physics-based simulation, it reduces per-query planning time by 2–19.6× relative to raw-grid A* after a one-time abstraction, with 6.7% cost overhead.

Read paper (IEEE Xplore) ↗

Module 02

Physics Informed Human Model

Module overview ↗

Publication list coming soon.

Module 03

Neuro-Symbolic Vehicle Model

Module overview ↗

Learning When to Jump for Off-road Navigation

Zhipeng Zhao, Taimeng Fu, Shaoshu Su, Qiwei Du, Ehsan Esfahani, Karthik K. Dantu, Souma Chowdhury, and Chen Wang.

Proceedings of Robotics: Science and Systems (RSS), Sydney, Australia, July 2026. DOI: 10.15607/RSS.2026.XXII.066.

This paper introduces Motion-aware Traversability (MAT), which models terrain cost as a function of vehicle velocity to decide when to cross, jump, or avoid obstacles. Perception predicts terrain-dependent Gaussian parameters once, allowing the planner to update costs efficiently as motion changes. Simulation and real-world experiments demonstrate agile off-road navigation with 75% less path detour while maintaining safety in the evaluated terrains.

Read paper (RSS proceedings) ↗

Module 04

COA Generation

Module overview ↗

Publication list coming soon.

QR code linking to https://onr-grapple.github.io/

Take the research with you

Scan to open this website on your phone.

onr-grapple.github.ioDownload QR code ↓