Investigating state estimation methods and sensor calibration pipelines to enable robust positioning in GNSS-degraded or signal-denied environments.
Modern UAV, vehicular, and submarine systems rely heavily on precise positioning. When GPS/GNSS signals are lost or actively jammed, systems must fall back on internal sensors. My research focuses on leveraging low-cost Micro-Electromechanical Systems (MEMS) Inertial Measurement Units (IMUs) by developing advanced estimation algorithms and real-time calibration routines to minimize drift.
Specifically, I develop robust Particle Filter frameworks tailored to two distinct map-matching paradigms: one fusing vertical gravity gradients from quantum gravimeters with inertial navigation systems (INS), and another fusing magnetic field anomalies with INS data. Fusing these geophysical fields with UAV flight dynamics, vehicular motion models, and submarine depth/velocity constraints allows vehicles to bind drift and navigate without GPS.
Experience range-only landmark tracking in real time. Guide the robot manually or let it patrol autonomously to watch the particle cloud predict, weight, and resample states.
Open Live SimulatorFusing inertial navigation systems (INS) with geophysical databases. I develop Particle Filter frameworks for two distinct modes: matching vertical gravity gradients from quantum gravimeters, and matching magnetic anomalies. Fusing these fields with UAV, vehicle, and submarine dynamics constrains unbounded inertial drift during GNSS outages.
Research FocusImplementing Rao-Blackwellised Particle Filters (RBPF) to calibrate sensors online without filter divergence. This decouples low-dimensional navigation states (like position and heading) from high-dimensional sensor calibration parameters (like magnetometer biases and scale factors) estimated via per-particle Kalman filters.
Research FocusFusing deep learning surrogate models (DNNs) with Particle Filters to replace computationally expensive online numerical integrations (such as gravity gradient evaluation) with sub-millisecond forward passes, optimized in real-time C++ (Eigen) for low-power onboard processors.
Research FocusRao-Blackwellised Particle Filtering (RBPF), Deep Learning Surrogate Models, EKF/UKF, and Factor Graph Optimization.
UAV flight dynamics, non-holonomic vehicle models, submarine depth/velocity constraints, and Lie Groups (SO(3), SE(3)) for 3D rotation representation.
Real-Time C++ (Eigen), Python (PyTorch/NumPy), MATLAB/Simulink, and hardware-in-the-loop testing.
See the software packages, sensor models, and peer-reviewed papers detailing state estimation and navigation systems.