Safe and Open-Space Aware Navigation
Human-aware occupancy mapping and reinforcement learning for narrow dynamic environments.
Explore Mobile Robot →
RV Lab · Konkuk University
Intelligence that moves.
We build robots that perceive uncertainty, learn from interaction, coordinate with other agents, and perform useful work in environments designed for people.
5integrated research programs
2023–26current featured work
Sim + Realevaluation environments
Physical AIshared research direction
Our mission
Close the loop between what a robot senses, understands, and does.
RV Lab develops embodied intelligence across mobile robots, legged systems, manipulators, coordinated robot teams, and humanoids. The embodiments differ, but the central question remains the same: how can a robot make reliable decisions when the physical world is uncertain, dynamic, and shared?
Our work combines multimodal perception, geometric representation, reinforcement learning, multi-agent coordination, simulation, and careful real-world validation. Research is organized as connected programs so that advances in safe navigation, human-motion understanding, or demonstration data can strengthen more than one robot platform.
Research programs
Five embodiments, one connected agenda
Explore each program for methods, verified results, full-size paper figures, current directions, and related applications.
01 · Perception to navigation
Mobile Robot
Human-aware autonomy, semantic mapping, RGB–LiDAR fusion, and safe reinforcement learning.
02 · Predictive safety
Quadruped Robot
Future-risk prediction, stability-aware learning, and robust locomotion.
03 · Perception to contact
Manipulator
Robot learning, visual quality inspection, and surface-aware automation.
04 · Cooperative intelligence
Bimanipulator
Multi-agent control, collision-free coordination, demonstration data, and VLA.
05 · Integrated Physical AI
Humanoid Robot
Human-motion understanding, whole-body control, and language-conditioned skills.
Shared research loop
From multimodal data to physical impact
A common research architecture keeps perception, learning, control, and deployment connected.
01 · PERCEIVE
Sense the world
Fuse vision, geometry, motion, contact, language, and robot state.
02 · REPRESENT
Structure uncertainty
Build maps, safety fields, compact geometry, and embodied state.
03 · LEARN & CONTROL
Select useful action
Optimize policies that balance task progress, coordination, and safety.
04 · VALIDATE
Return to reality
Compare baselines, probe failure modes, and close the sim-to-real loop.
Recent highlights
Representative results across the lab
These highlights are drawn from verified journal and conference publications, with program pages providing fuller context.
88–94%SOAR-RL navigation success
97.06%production inspection accuracy
83–95%multi-arm task success
Multi-Agent Collision-Free Posture Control
Compact link geometry and cooperative learning for shared multi-arm workspaces.
Explore Bimanipulator →Emerging Embodied AI Systems
Current work spans stability-aware locomotion, multimodal demonstrations, VLA, and learned picking.
Browse latest publications →Research in context
Built for environments where robots must matter
Projects connect fundamental methods with mobility, manufacturing, human support, and digital-twin applications.
Autonomous systems
Mobility & logistics
Shared-space navigation, semantic maps, localization, and mobile autonomy.
Advanced manufacturing
Inspection & manipulation
Machine vision, robot learning, and adaptive surface processes.
Human-centered AI
Motion & interaction
Human behavior understanding and embodied systems designed around people.
Research, people, and results
Build the next embodied system with us.
We welcome students, academic collaborators, and industry partners working at the intersection of robotics, learning, simulation, and real-world deployment.