Marine Robotics Applications

This research of the Multi-robot Systems group aims to investigate problems and applications of  Micro Aerial Vehicle (MAV) in marine environments. In order to achieve such goals, an integration of principles and theoretical background of the behaviors of a MAV and/or a swarm of MAVs with methodology/theory describing cooperative localization and cooperative perception of autonomous robots is needed. Furthermore, multi-robot system control and principles of adaptation to the unfriendly environment for MAVs leading to a flexible stand-alone system are investigated. This research will enable the applicability of MAVs in realistic marine scenarios of surveillance, reconnaissance, search and rescue, and so on. Basically, we develop techniques that enable MAVs to interact with Unmanned Surface Vehicles (USVs) and objects on water and monitor and interact with a water environment from an aerial perspective. This enables to employ of MAVs outside laboratories equipped with a precise motion and positioning system.

Precise capture of objects on the water by a group of UAVs-USVs, optimal physical interaction between UAVs and USVs, and visually detecting any object or water phenomena are among the applications researched. To enable these multi-robot applications, theoretical principles for object detection on the water with respect to its pose, mass, and shape, a precise control system robust to uncertainties, and precise coordination in a multi-robot heterogeneous group are designed based on methods of artificial intelligence and vision systems such as deep reinforcement learning, as well as model predictive controllers. This research is aimed at the study of UAV applications in cooperation with marine robots.

In addition to the development of the theory of coordinated motion of the multi-robot system members, our research also concerns the development of the sensory equipment as well as new MAV platforms needed for real-world MAV flights, and subsequent integration of the observational constraints it induces on the MAV control in use. The research conducted in this stream is closely coordinated with the research on multi-robot systems being also realized within the Multi-robot Systems group. Our state-of-the-art approaches in the research of marine robotics applications and description of developed methods can be found in papers chronologically listed below.

MAV platform:

Tarot T650 Marine

Our newest platform based upon the Tarot T650 frame. This MAV has floating landing gear installed to secure a safe landing on water. The onboard sensory equipment is optimized to make this drone ready for an autonomous flight over any marine environment such as lakes, rivers and open sea with any harsh conditions (wind, big waves, rain, etc.). It is equipped with a GPS,  1 RGB-D camera for horizontal maneuverability, 1 RGB camera for visual inspections, and adjustable on-board light. This platform can have optional UV-lights and camera for swarming and a manipulator. We use this platform for all the marine applications.
  • 650 mm frame
  • 15″ carbon propellers
  • 4x T-Motor MN3510-13 700 kV motors
  • 6S 8000 mAh LiPo battery
  • ~3.5 kg liftoff weight
  • 20 min flight time


Examples of research results

UAV Landing on a USV in harsh winds and turbulent open-waters (video featured in RA-L 2022).

 Landing an unmanned aerial vehicle (UAV) on top of an unmanned surface vehicle (USV) in harsh sea conditions is a challenging problem owing to the forces that can damage the UAV due to severe roll and/or pitch angle of the boat during touchdown. In this research, we propose a novel model predictive control (MPC) approach that enables a UAV to land autonomously on a USV in open-waters and windy conditions. The MPC employs a novel objective function and an online decomposition of the motion of the vessel, that is oscillating in waves, in order to attempt and complete the landing during near-zero tilt of the landing platform. The nonlinear prediction of the motion of the vessel is performed using visual data from an on-board camera. And thus, the system doesn’t require any communication with the USV or any control station. The proposed method was analysed in numerous robotics simulations in harsh and extreme conditions and further validated in various real-world conditions.


Our platforms in action:


Model Predictive Control-based Trajectory Generator for Agile Landing of Unmanned Aerial Vehivles on a Moving Boat

 This work focused on the design, implementation and verification of trajectory planning for the landing of an unmanned multirotor Unmanned Aerial Vehicle (UAV) on the deck of a moving boat. The work builds on an already existing system that provides accurate estimation and prediction of vessel’s states, as well as a system of precise autonomous control of an UAV. The main focus of this work lies in designing a trajectory generator that uses Unmanned Surface Vehicle (USV)’s states and creates trajectories for the unmanned helicopter. The concept of Model Predictive Control, which takes advantage of an optimisation problem solver based on first-order solution methods, is utilised in the process. This allows the use of a relatively accurate model of the UAV to describe its dynamics. Thanks to that, it is possible to include position, orientation, linear velocities and Euler rates in the planning of the trajectory. The trajectory generator is capable of creating trajectories from take-off to touch-down while all the computations are performed onboard the UAV in real-time.



Multi-vehicle Dynamic Water Surface Monitoring

 Repeated exploration of a water surface to detect objects of interest and their subsequent monitoring is important in search-and-rescue or ocean clean-up operations. Since the location of any detected object is dynamic, we propose to address the combined surface exploration and monitoring of the detected objects by modeling spatio-temporal reward states and coordinating a team of vehicles to collect the rewards. The model characterizes the dynamics of the water surface and enables the planner to predict future system states. The state reward value relevant to the particular water surface cell increases over time and is nullified by being in a sensor range of a vehicle. Thus, the proposed multi-vehicle planning approach is to minimize the collective value of the dynamic model reward states. The purpose is to address vehicles' motion constraints by using model predictive control on receding horizon, thus fully exploiting the utilized vehicles' motion capabilities. Based on the evaluation results, the approach indicates improvement in a solution to the kinematic orienteering problem and the team orienteering problem in the monitoring task compared to the existing solutions. The proposed approach has been experimentally verified, supporting its feasibility in real-world monitoring tasks.\


State Estimation of an Unmanned Surface Vehicle by a UAV

 This work dealt with design, implementation, and verification of a system for state estimation of an Unmanned Surface Vehicle (USV) known as a boat by an Unmanned Aerial Vehicle (UAV). First, linear and nonlinear mathematical models of the boat extended by wave dynamics are presented. Introduced mathematical models of the boat are used by Kalman filters that perform estimation of boat states using data from UAV onboard sensors. Kalman filters also use data received from sensors placed on the boat, which is sent to the UAV by a wireless communication link. The result of this thesis is a UAV onboard system that provides state estimation of the boat moving on a wavy water surface. The estimation system of boat states is implemented into the UAV control system. The presented estimation system was verified by conducting real-world experiments.