A5065553528- Underwater Vehicles and Communication Systems
- Underwater Acoustics Research
- Maritime Navigation and Safety
- Fault Detection and Control Systems
- Advanced Neural Network Applications
- Image Enhancement Techniques
- Adaptive Control of Nonlinear Systems
- Robotic Path Planning Algorithms
- Hydraulic and Pneumatic Systems
- Advanced Control Systems Optimization
- Flood Risk Assessment and Management
- Vehicle Dynamics and Control Systems
- Robotics and Sensor-Based Localization
Purdue University West Lafayette
2021-2024
One of the main limiting factors in deployment marine robots is issue energy sustainability. This particularly challenging for traditional propeller-driven autonomous underwater vehicles which operate using intensive thrusters. emerging technology to enable persistent performance use recharging and retasking through docking stations. paper presents an integrated navigational algorithm facilitate reliable vehicles. Specifically, dynamically re-plans Dubins paths create efficient trajectory...
Autonomous Underwater Vehicles (AUVs) are a vital element for ocean exploration in various applications; however, energy sustainability still limits long-term operations. An option to overcome this problem is using underwater docking power and data transfer. To robustly guide an AUV into station, we propose vision algorithm short-distance detection. In paper, present Convolutional Neural Network architecture accurately estimate the dock position during terminal homing stage of docking....
Abstract Station keeping is an essential maneuver for autonomous surface vehicles (ASVs), mainly when used in confined spaces, to carry out surveys that require the ASV keep its position or collaboration with other where relative has impact over mission. However, this can become challenging classic feedback controllers due need accurate model of dynamics and environmental disturbances. This work proposes a predictive controller using neural network simulation error minimization (NNSEM–MPC)...
Obstacle detection for autonomous navigation through semantic image segmentation using neural networks has grown in popularity use unmanned ground and surface vehicles because of its ability to rapidly create a highly accurate pixel-wise classification complex scenes. Due the lack available training data, are rarely applied water scenes such as rivers, creeks, canals, harbors. This work seeks address issue by making one-of-its-kind River Segmentation En-Route By USV Dataset (ROSEBUD)...
Neural network semantic image segmentation has developed into a powerful tool for autonomous navigational environmental comprehension in complex environments. While networks have seen ample applications the ground domain, implementations surface water especially fluvial (rivers and streams) deployments, lagged behind due to training data literature sparsity issues. To tackle this problem publicly available River Obstacle Segmentation En-Route By USV Dataset (ROSEBUD) was recently published....
This work presents a framework that allows Unmanned Surface Vehicles (USVs) to avoid dynamic obstacles through initial training on an Ground Vehicle (UGV) and cross-domain retraining USV. is achieved by integrating Deep Reinforcement Learning (DRL) agent generates high-level control commands leveraging neural network based model predictive controller (NN-MPC) reach target waypoints reject disturbances. A Q Network (DQN) utilized in this trained ground environment using Turtlebot robot...
Underwater docking is a staged process in which the detection of dock crucial. It allows Autonomous Vehicles (AUVs) to recharge and transfer data, enabling long-term missions; recent work shows that deep learning can be used robustly perform at expense large amount resources for deployment on embedded devices. This paper proposes method efficiently train Convolutional Neural Network (CNN) detect station using knowledge distillation under teacher-student architecture. Additionally, augment...
In this paper, we discuss the development and deployment of a robust autonomous system capable performing various tasks in maritime domain under unknown dynamic conditions. We investigate data-driven approach based on modular design for ease transfer autonomy across different surface vessel platforms. The alleviates issues related to priori identification models that may become deficient evolving behaviors or shifting, unanticipated, environmental influences. Our proposed learning-based...
Accurately following a prescribed path is critical for safe and efficient operation of autonomous systems in the field, especially marine robots that typically operate away from human support with sporadic communication abilities. Traditionally, variations lookahead control strategy have been employed unmanned systems; however, these strategies can encounter problems when faced disturbances discontinuities localization. In underwater applications, intermittent global localization updates are...
Station keeping is an essential maneuver for Autonomous Surface Vehicles (ASVs), mainly when used in confined spaces, to carry out surveys that require the ASV keep its position or collaboration with other vehicles where relative has impact over mission. However, this can become challenging classic feedback controllers due need accurate model of dynamics and environmental disturbances. This work proposes a Model Predictive Controller using Neural Network Simulation Error Minimization...
Underwater docking is critical to enable the persistent operation of Autonomous Vehicles (AUVs). For this, AUV must be capable detecting and localizing station, which complex due highly dynamic undersea environment. Image-based solutions offer a high acquisition rate versatile alternative adapt this environment; however, underwater environment presents challenges such as low visibility, turbidity, distortion. In addition field experiments validate capabilities can costly dangerous...
This paper presents experimental ground truth data validation of the ability Underwater Gliders (UGs) to maneuver in constrained environments through starting, stopping, and maintaining turning motions on demand. capability has been validated a pool custom made highly maneuverable underwater glider, ROUGHIE, using an motion capture system for pose tracking. The experiments indicate that ROUGHIE is capable robust repeatable operation complex paths due its effectively transition between stable...