Embedded Files
Scalable Uncertainty-aware Active Planning with Tightly Coupled Localization-Mapping
Scalable Uncertainty-aware Active Planning with Tightly Coupled Localization-Mapping
Proposed an active inference framework for robust, scalable, and computationally efficient uncertainty aware planning with tightly coupled localization and mapping (SLAM).
Developed an end-to-end uncertainty-aware navigation stack, encompassing perception, localization/mapping, motion planning, and decision-making.
Validated through closed-loop real-time simulations using a Clearpath Warthog UGV equipped with Ouster LiDAR and odometry in both the ROS-Unity simulation environment and a real-world setting
[Paper] : will be soon!
Active Graph-based LiDAR Inertial Odometry (LIO) SLAM
Active Graph-based LiDAR Inertial Odometry (LIO) SLAM
Developed an active graph-based LIO SLAM to guide exploration toward informative areas and reduce map drift error.
Global planner (A*) guided Model Predictive Path Integral (MPPI) control.
Validated through closed-loop real-time simulations using a Clearpath Warthog UGV equipped with Ouster LiDAR and odometry in the realistic ROS-Unity simulation environment.
[Github]
Information Geometry-aware Bayesian Filtering for Nonlinear Robot Navigation
Information Geometry-aware Bayesian Filtering for Nonlinear Robot Navigation
Proposed a nonlinear Bayesian filtering framework that integrates probabilistic estimation theory with Variational inference theory and information geometry principles, applicable to robotic systems.
Verified estimation performance in real-world experiments and identified its connection to state estimation.
[Paper]
Computationally Efficient Particle-Based MAP Trajectory Estimation
Computationally Efficient Particle-Based MAP Trajectory Estimation
Proposes a novel MAP sequence estimation method, Stein-MAP, which effectively addresses multimodality while substantially reducing computational and memory overhead.
Achieved a computational complexity of O(M^2), where M is the number of particles, compared to the O(N^2) cost of the classical MAP sequence estimation with N >> M.
[Paper]
Human-Autonomy Sensor Fusion for Target Tracking and Calibration
Human-Autonomy Sensor Fusion for Target Tracking and Calibration
Developed integrated sensor fusion of human-drawing data into autonomous target tracking systems.
Designed a probabilistic model for quantifying and performing online updates of uncertainty in human data.
Validated in extensive simulations involving multi-agent systems in a tracking scenario.
[Paper], [Poster]
Multi-agent Pedestrian Navigation System
Multi-agent Pedestrian Navigation System
Developed a Pedestrian Navigation System (PNS) based on the simultaneous integration of deterministic, probabilistic, and cooperative localization algorithms.
Demonstrated through real-world experiments in a multi-agent localization scenario.
Achieved localization accuracy within 1 meter in a GPS-denied environment during hours of operation.
[Paper]
Accelerated Trajectory Estimation using Adaptive Belief Propagation
Accelerated Trajectory Estimation using Adaptive Belief Propagation
Introduced robust and computationally efficient trajectory estimation for unknown target tracking.
Validated robustness against multimodal distributions (multiple solution candidates) and achieved computational efficiency in real-world experiments, with 500x faster computation and 10x reduced memory usage.
[IROS Presentation slides], [Paper]
Certifiable State Estimation & Sensor Fusion for Vision-based Autonomous Systems
Certifiable State Estimation & Sensor Fusion for Vision-based Autonomous Systems
Proposed a robust localization algorithm to certify machine learning-based object detection results (e.g., YOLO), leveraging principles of information geometry.
[Slides]
vision based target tracking.mp4Vision-based Target Tracking and Following on an Onboard UGV
Vision-based Target Tracking and Following on an Onboard UGV
Designed a machine learning-based vision system for target tracking and following on a UGV.
Implemented the tracking system on an onboard UGV in a hardware experiment.
Machine Learning-based Hand Gesture Classification for Robot Teleoperation
Machine Learning-based Hand Gesture Classification for Robot Teleoperation
Developed hand gesture classification using wireless sensors for robot teleoperation.
Trained hand gestures using raw sensory data with machine learning techniques, including Neural Networks and k-Nearest Neighbors.
Implemented the inference system for directional guidance of a TurtleBot.
[Report]
Convex Optimization based Nonlinear Controller Design for Flight Systems
Convex Optimization based Nonlinear Controller Design for Flight Systems
Proposed Sum Of Squares (SOS) optimization based analysis and synthesis of nonlinear flight systems.
Developed nonlinear controllers within the framework of control (Lyapunov) theory.
Demonstrated through nonlinear closed-loop 6-DOF simulations in MATLAB/Simulink.
Introduced stability analysis for linear controllers applied to nonlinear (aerodynamic) dynamics.
[Paper]: nonlinear controllers, [Paper]: stability analysis.
Autonomous Navigation Technology for Proximity Operation of Satellites
Autonomous Navigation Technology for Proximity Operation of Satellites
Developed vision-based satellite navigation systems
Introduced nonlinear filters (UKF, PF) & nonlinear controllers for vision-based satellite navigation systems.
Verified through nonlinear closed-loop simulations in a custom-developed 3D virtual space environment using MATLAB/Simulink.
[Paper]
Airdrop System using Autonomously guided Parachute
Airdrop System using Autonomously guided Parachute
Developed an autonomous parafoil system capable of delivering luggage to a designated target.
Designed guidance and navigation algorithms for a parafoil system.
Demonstrated through nonlinear closed-loop 6-DOF simulations in MATLAB/Simulink, along with hardware experiments in real-world scenarios.
Page updated
Google Sites
Report abuse