Research theme
Robust Autonomy Under Uncertainty
Robotics research presence
Robust Autonomy Under Uncertainty
Research-oriented robotics, perception, estimation, and decision-making systems for safety-critical autonomous agents.
This site is organized as a coherent research program. Each repository is framed through a scientific question, current implementation status, limitations, and next experiments, with no inflated claims about publications, affiliations, or validated benchmarks.
Research ConsoleActive
Open modules
6 repositories
Research standard
Implemented, prototype, and planned work are separated
System view
A connected autonomy research program
Research focus
Robust autonomy, visual-inertial localization, uncertainty-aware perception, risk-aware navigation, and intelligent mobility safety.
Research statement
Autonomy should expose uncertainty before it fails.
My work is organized around a practical research problem: robotic systems often fail not because every module is wrong, but because perception, localization, and planning continue operating as if degraded estimates were reliable.
The portfolio therefore treats uncertainty as a system variable. VIO health, sensor reliability, semantic perception, communication quality, and path risk are connected as parts of one autonomy stack.
Scientific framing
Every module begins from a research question, not from a visual demo.
Transparent maturity
Implemented, prototype, planned, and pending results are deliberately separated.
Reproducibility first
Experiments are expected to expose assumptions, metrics, limitations, and failure cases.
Research pillars
One coherent autonomy stack.
Perception
State Estimation
Uncertainty
Risk
Planning
Navigation
Safety
Simulation
Current research modules
Projects as components of a research agenda.
Visual-Inertial Odometry · Robust Localization
SHIELD-VIO
Can a VIO pipeline become health-aware enough to anticipate and mitigate localization degradation?
Research Prototype · research module
Visual-Inertial Odometry · Uncertainty Gating
AEGIS-VIO
When should a VIO system reject or down-weight visual measurements because they are likely to reduce estimator consistency?
Planned · research module
SLAM · Sensor Fusion · Event-Based Perception
Adaptive Multi-Modal SLAM
Can SLAM systems estimate sensor reliability online and adapt fusion weights before severe localization failure?
Active · research module
Navigation · Planning · Formal Safety
DynNav
How should a robot trade off path efficiency, uncertainty exposure, and returnability in unknown environments?
Active · research module
Autonomous Driving Safety · Trajectory Prediction
SafeCrossAI
Can infrastructure-based AI predict vulnerable road-user motion and expose uncertainty for safer intersection decisions?
Active · research module
Communications · AI · Safety-Critical Autonomy
OpenUWOC-AI
How can autonomous systems reason about communication uncertainty when operating in degraded underwater channels?
Research Prototype · research module