Current surface vehicle tracking systems often rely on GPS, prior knowledge of the target, or external communication links—making them vulnerable in denied or adversarial environments. This research investigates the development of a real-time perception and decision-making architecture for a follower USV, enabling it to autonomously track and follow an uncooperative vessel using only onboard passive sensing. The system combines LiDAR point cloud processing with machine learning–based vision to detect and track targets without relying on GPS, comms, or centralized control. Field experiments validate the system’s ability to perform in real-world maritime scenarios.
Integrate LiDAR and onboard camera systems for passive sensing
Apply K-Means clustering to LiDAR point clouds to identify the lead vessel
Deploy a custom-trained YOLOv9 object detector to validate visual targets
Fuse detections with real-time Kalman filtering to estimate target motion
Implement the architecture on a WAM-V USV using ROS 2 for full onboard autonomy
Validate performance through field trials simulating adversarial conditions
Technical information can be found in the Testing section.
Note: This research is currently under development
Current multi-USV navigation strategies often rely on centralized planning, inter-vessel communication, or make assumptions about the behavior of nearby vessels—limiting their reliability in dynamic or communication-limited environments. This research explores a decentralized navigation system that enables multiple USVs to autonomously follow the International Regulations for Preventing Collisions at Sea (COLREGs). The approach integrates Artificial Potential Fields (APF) with COLREGs-compliant behavior rules, allowing each USV to make independent, onboard decisions based on real-time sensor data without prior coordination or connectivity. The system is validated through simulated encounter scenarios using the Virtual RobotX (VRX) environment.
Implement decentralized path planning based on Artificial Potential Fields
Integrate COLREGs rules for head-on, overtaking, and crossing situations
Distinguish between static objects and vessels using LiDAR and camera fusion
Develop onboard decision-making logic for each USV without centralized control
Simulate multi-USV interaction scenarios in the VRX Gazebo simulator
Evaluate behavior through various encounter types involving shared navigation goals
Technical information can be found in the Testing section.
Note: This research is currently under development
Current AUV navigation methods suffer from cumulative errors, sometimes require additional infrastructure, and are often affected by environmental conditions. This thesis investigates the creation of high-resolution geomagnetic survey maps and evaluates the feasibility of integrating geomagnetic navigation with existing AUV systems. The goal is to use these geomagnetic maps to enhance the performance and reliability of AUV navigation in challenging underwater environments.
Objectives:
Integrate Geomagnetic Measurement onboard an AUV
Conduct Geomagnetic Surveys
Develop maps of the magnetic field
Analyze and define the specific requirements associated with conducting geomagnetic surveys
Develop algorithms to support a navigation system for AUVs by integrating geomagnetic data from the AUV’s magnetometer and pre-existing geomagnetic field models
Technical information can be found in the Publications section.
This research focuses on the modeling, implementation, and field testing of a power takeoff (PTO) system equipped with a ball-type continuously variable transmission (B-CVT) for a small marine hydrokinetic (MHK) turbine. The turbine, a partially submerged multi-blade undershot waterwheel (USWW), is deployed from a floating unmanned autonomous mobile catamaran platform. A validated numerical torque model for the MHK turbine has been developed, and a speed controller has been implemented and tested in the field to optimize power generation based on blade configurations and submergence levels.
This project aims to harness marine current energy efficiently, with potential applications such as recharging aerial drones (UAVs) and supporting sustainable power solutions for unmanned systems. Bench and field testing have been conducted to assess the turbine's power conversion capabilities, and detailed results on turbine performance and controls architecture are provided.
Technical information can be found in the Publications section.
This research project focuses on the development of a Proportional-Integral-Derivative (PID) auto-tuning controller for Unmanned Surface Vehicles (USVs), specifically the WAM-V 16, integrating Artificial Potential Field (APF) methods for enhanced path planning. While traditional PID controllers are widely used for low-level control due to their simplicity and effectiveness, they often require manual tuning to adapt to changes in the vehicle's dynamics and environmental conditions. This project aims to address this limitation by implementing a PID auto-tuning mechanism that utilizes AI algorithms to automatically adjust the controller parameters in real-time, thereby ensuring optimal performance under varying operational conditions.
Additionally, the research incorporates APF techniques to facilitate efficient path planning, allowing the WAM-V 16 to navigate complex environments while avoiding obstacles. The APF approach creates a virtual field that guides the vehicle towards its target while repelling it from undesired obstacles. By combining PID auto-tuning with APF for path planning, the project seeks to enhance the navigational capabilities of USVs, making them more robust and adaptive to real-world challenges. The study is structured into several sections, including an overview of existing research, the proposed methodology, and a detailed timeline for completion.
Note: This research is currently under development, papers on low-level control for WAMV-16 can be found at Publications section.
Collaborative Unmanned Robots for Autonomy
The FAU SCUBA (Scalling Collaborative Unmanned Robots for Autonomoy) Lab focuses on advancing autonomy in unmanned vehicles, particularly Unmanned Aerial Vehicles (UAVs) and Unmanned Surface Vehicles (USVs). The primary goal is to improve the cooperative behaviors of these vehicles through innovative algorithms and methodologies.
Autonomy Scaling: The lab aims to develop scalable autonomy solutions that enable UAVs and USVs to operate collaboratively in complex environments.
Sensor Integration: Research emphasizes the integration of various sensors, enhancing situational awareness and decision-making capabilities of unmanned systems.
Multi-Vehicle Coordination: The lab explores techniques for coordinating multiple UAVs and USVs, optimizing their collaborative tasks such as exploration, mapping, and data collection.
Artificial Intelligence: Utilizing AI to enhance decision-making processes in UAVs and USVs, allowing for improved adaptability and responsiveness in dynamic environments.
Machine Learning: Implementing machine learning techniques to develop predictive models for environmental interactions, enabling unmanned systems to learn from past experiences and improve their operational efficiency.
Path Planning: Research includes the development of advanced path planning algorithms, focusing on efficient and safe navigation in collaborative settings.
The research conducted at SCUBA Lab has practical applications in marine exploration, disaster response, and environmental conservation, aiming to contribute significantly to the field of autonomous robotics.
For more information, visit: https://mukhe027.github.io/scubalab/