SHIELD-VIO
Active research repository connected to robust perception, uncertainty-aware estimation, or safe decision making.
Research vision
Reliable autonomy requires perception systems that know when they are uncertain.
My research direction studies how robots can detect degraded sensing, represent uncertainty, adapt estimation and fusion, and make safer navigation decisions before failures become safety-critical.
Current projects
A connected research program.
Adaptive Multi-Modal SLAM
Active research repository connected to robust perception, uncertainty-aware estimation, or safe decision making.
DynNav
Active research repository connected to robust perception, uncertainty-aware estimation, or safe decision making.
SafeCrossAI
Active research repository connected to robust perception, uncertainty-aware estimation, or safe decision making.
Themes
Research areas.
SLAM
Robust mapping and localization under perceptual degradation and incomplete sensing.
Visual-Inertial Odometry
Health-aware VIO with degradation monitoring, diagnosis, and recovery policies.
Sensor Fusion
Adaptive weighting of visual, inertial, event-based, and semantic signals under uncertainty.
Autonomous Driving
Prediction and safety reasoning for vulnerable road users at intelligent intersections.
Risk-Aware Robotics
Planning methods that consider collision risk, returnability, and uncertainty propagation.
UAV Autonomy
Future extension of robust perception and VIO methods to aerial platforms.
Computer Vision
Perception modules that expose reliability, calibration, and failure modes.
AI for Robotics
Learning systems evaluated by decision utility, robustness, and reproducibility.
Underwater Communications
Future Work: communication-constrained autonomy and sensor-network reliability.
Future directions
Next research steps.
- Benchmark degraded VIO and SLAM pipelines against reproducible baselines.
- Connect localization health estimates to planner-level risk and recovery behavior.
- Develop visualization tools for uncertainty, trajectories, semantic perception, and point clouds.
- Keep publication sections ready while marking research outputs as forthcoming until accepted or public.