In this study, we investigated autonomous vessel obstacle avoidance using advanced techniques within the Guidance, Navigation, and Control (GNC) framework. We propose a Mixed Integer Linear Programming (MILP) based Guidance system for robust path planning avoiding static and dynamic obstacles. For Navigation, we suggest a multi-modal neural network for perception, demonstrating the identification of obstacle type, position, and orientation using imaging sensors. Additionally, the paper compares an error-based PID control strategy and a Model Predictive Control (MPC) scheme as well. This evaluation aids in better evaluating their performance and determining their applicability within the GNC scheme. We detail the implementation of these systems, present simulation results, and offer a performance evaluation using an experimental dataset. Our findings, analysed through qualitative discussion and quantitative performance indicators, contribute to advancements in autonomous navigation and the control strategies to achieve it.

@en

A novel follow-the-leader approach for azimuth-driven vessels is devised and experimentally tested in a model-scale outdoor scenario. The vessels are equipped with global navigation satellite and inertial navigation systems. A line-of-sight algorithm ensures the yaw-check ability of the follower vessel along the leader's path, while a speed-regulation allows to track its velocity. Track generation, guidance, navigation, and control modules are designed and assembled to be executed on-board in real time. The results of an outdoor experimental campaign are illustrated to show the effectiveness of the proposed approach.

@en

The framework of an autonomous vessel is typically composed of three distinct and independent blocks known as the Guidance, Navigation and Control (GNC) system. This paper presents a combination of advanced complementary techniques in the different GNC subsystems to improve upon the current common practices/state of the art in obstacle avoidance. The novel Guidance system is based on Mixed Integer Linear Programming (MILP). This optimisation technique allows quick, robust path planning with the possibility for a variety of constraints. The feasibility of this method will be investigated with the goal of providing optimal path planning in the presence of static and dynamic obstacles during autonomous sailing operations. Within the Navigation System, a multi-modal neural network architecture is proposed for the perception branch to provide high-level situational awareness for collision avoidance purposes. The computer vision approach allows for the vessel type, position and orientation all to be extracted for encounters with both dynamic and static obstacles using only imaging sensors. Two Control methods are studied in the paper, an error-based PID control strategy as well as an MPC control scheme. These two techniques will be compared to evaluate the performance and reviewing the suitability for use within the specific GNC scheme and the generic application environment. This paper details the specific implementation of each system within the overall framework, presents simulation results of the path planner and control strategy, with a performance evaluation of the navigation system using an experimental dataset. The results obtained are analysed through qualitative discussion as well as quantitative performance indicators and key conclusions are consequently drawn.

@en

Sustainability goals have reignited interest in more fuel efficient modalities for urban freight transportation. The autonomous inland vessel has the potential to enable a modality shift to the waterway, however economically-conscious solutions to the technological challenges are lacking. This paper explores the potential role of stereovision technology as a primary sensor set within the application of autonomous inland vessels. The design of a multi-camera navigation system for mid-range perception and localisation tasks is presented with a view to providing sufficient data for collision avoidance tasks. The performance of these sensors in the application is evaluated with experimental testing using small-scale vessels in an indoor tank, demonstrating their capabilities and limitations for consideration in ongoing research.@en

The development of fast and accurate intelligent vessel control systems is a necessary milestone on the path toward operating autonomous marine vehicles effectively in harsh environments and complex mission settings. One of the main problems of existing control systems is the disparity between the forecasted behaviour and how the vessel actually responds to its environment. This disparity can be partly attributed to the dependency on physics-based methods to model the response of the vessel and the fact that accurate high-fidelity physical models are too computationally expensive to be utilized in real time. One promising solution to this problem is to integrate the dynamic environmental conditions such as sea states, winds, and currents to model the response of the vessel. However, this may not be feasible with the existing physics-based controller strategies due to the high computational requirements. Instead, we propose using Artificial Intelligence (AI) based methods, which leverage Data Mining and Machine Learning, to enable fast and accurate short-term motions forecasting for autonomous marine vehicles. The AI-based approach is extremely time-aware in the forecasting phase since it does not rely on solving the physics behind the phenomenon but rather learns a phenomenon from historical examples, linking the vessel's motions to a holistic view of its real-time environment.To test our hypothesis, we will develop state-of-the-art AI-based models for the short-term motions forecasting of the roll and trim of a twin-engine commercial vessel using real-world operational data and leverage statistical methods to validate our results.@en

Floating structures have raised interest in the recent years for different applications, from living and farming at sea to renewable energy production. To support the logistics on the floating structures, floating cranes are necessary and their designs are constantly improved. Increasing developments in the automation industry paved the way for automated crane operations. In this work, motion control of a smart crane is presented with particular attention to the performance under wave motion. In this research, a scaled down, two-dimensional mathematical model of a gantry crane is derived using Lagrangian mechanics and DC motors dynamics. This results in a nonlinear system that is capable of simultaneous traversing and hoisting a container. The system is simulated in MATLAB Simulink environment and a proportional-derivative control and a state feedback control are designed and implemented. Their robustness is explored by modelling sensor behavior, external disturbances and floating platform dynamics. Both control strategies were able to keep stability in a disturbed system. During simulation, the sway angles never exceed 10°. Smaller oscillations occurred using the state feedback control. Therefore, it creates a smoother response compared to the proportional derivative control, which ultimately translates to increased safety, turnover rate and durability of the crane.@en

Formation control of autonomous surface vessels (ASVs) has been studied extensively over the last few years since it offers promising advantages. In this paper, two control methods for distributed leader-follower formation control are proposed: A Nonlinear Model Predictive Control (MPC) method and an MPC method using Feedback Linearization. One agent per vessel performs planning and control. The agents exchange information on their current and predicted positions. The two proposed methods are compared with each other and also with a conventional Proportional-Integral (PI) control method. The performance of the proposed strategies is evaluated through simulations and field experiments using small scale vessels. The simulation and field experiment results show that the proposed MPC-based approaches outperform the conventional PI control method.@en

The future autonomous ships will be operating in an environment where different autonomous and non-autonomous vessels with different characteristics exist. These vessels are owned by different parties and each uses its owned unique approaches for guidance and navigation. The Collaborative Autonomous Shipping Experiment (CASE) aims at emulating such an environment and also stimulating the move of automatic ship control algorithms towards practice by bringing together different institutes researching on autonomous vessels under an umbrella to experiment with collective sailing in inland waterways. In this paper, the experiments of CASE 2020 are explained, the characteristics of different participating vessels are discussed and some of the control and perception algorithms that are planned to be used at CASE 2020 are presented. CASE 2020 will be held in parallel to iSCSS 2020 at Delft University of Technology, the Netherlands.@en

This paper presents an effective autonomous follow-the-leader strategy for Azimuthal Stern Drive vessels. The control logic has been investigated from a theoretical point of view. A line-of-sight algorithm is exploited to ensure yaw-check ability, while a speed-check feature is implemented to track the velocity of the target along the path. For this purpose, a linearised manoeuvrability model for azimuthal drive surface vessels is presented. A model-based control synthesis is proposed to ensure the stability of the closed-loop system and robust PID controllers are designed by using Linear Matrix Inequalities technique. The control strategy has been successively validated in two steps, initially by using simulation techniques, and then experimentally using an outdoor scenario with model scale tugs. The path planning, navigation, guidance and control modules are studied, detailed, and digitally implemented on-board of the model scale tugs. The models are supplied with GNSS+INS navigation system. Low-level management and control of Azimuthals angles and shaft revolutions is implemented on-board. High-level decen-tralised path planning, guidance, and control sequence evaluation are dealt with at a remote ground station. In particular, the presented follow-the-leader strategy meets the most generic needs of platooning convoys, and, in the specific instance, of Escort convoy tugs. The operative profile of the latter concerns long-lasting and routine chases with the continuous demand of tuning heading and speed to track the target vessels, until the rare occurrence of an emergency event. In a realistic scenario, the proposed control system would be beneficial for the tug master’s lucidity and alertness, while reducing avoidable risks. At the end of the paper, the results of the experimental campaign are shown to demonstrate the effectiveness of the proposed control logic.@en

Floating structures have raised interest in the recent years for different applications, from living and farming at sea to renewable energy production. To support the logistics on the float- ing structures, floating cranes are necessary and their designs are constantly improved. In- creasing developments in the automation industry paved the way for automated crane opera- tions. In this work, motion control of a smart crane is presented with particular attention to the performance under wave motion. In this research, a scaled down, two-dimensional math- ematical model of a gantry crane is derived using Lagrangian mechanics and DC motors dy- namics. This results in a nonlinear system that is capable of simultaneous traversing and hoist- ing a container. The system is simulated in MATLAB Simulink environment and a proportional-derivative control and a state feedback control are designed and implemented. Their robustness is explored by modelling sensor behavior, external disturbances and floating platform dynamics. Both control strategies were able to keep stability in a disturbed system. During simulation, the sway angles never exceed 10 degrees. Smaller oscillations occurred us- ing the state feedback control. Therefore, it creates a smoother response compared to the pro- portional derivative control, which ultimately translates to increased safety, turnover rate and durability of the crane.@en

The maneuvering control of autonomous vessels has been under extensive investigations by academic and industrial communities since it is one of the primary steps towards enabling unmanned shipping. In this paper, a model predictive control (MPC) approach is presented for trajectory tracking control of vessels which takes into account the thrust allocation (TA) problem in the presence of rotatable thrusters. In this approach, the TA problem is formulated over a finite horizon and solved with regard to the power consumption, changes in the angle and speed of actuators, and the operating constraints. In the proposed control approach, several linearization techniques have been employed to enable the adoption of quadratic programming approaches for solving the MPC's and TA's optimization problems. The performance of the proposed approach is evaluated through several simulation experiments using a replica vessel model.

@en

In this paper, a distributed control approach is proposed to enable fuel-efficient Vessel Train Formations (VTF) in inland waterways and port areas for addressing the efficiency and environmental issues of transport over water. For path tracking, collision avoidance, and consensus over the VTF speed a distributed Model Predictive Control (MPC) algorithm is adopted which uses the Alternating Direction Method of Multipliers (ADMM) to guarantee path following and consensus between vessels. The all-electric Direct Current (DC) configuration is considered for the Power and Propulsion Systems (PPS) of the autonomous vessels under study. Considering their PPS specification, the vessels negotiate with each other to agree on the most efficient speed for all the vessels in the VTF. Simulation results suggest that a significant amount of fuel saving can be obtained by using the proposed approach.

@en

Real-time collision avoidance with full consideration of ship maneuverability, collision risks and International Regulations for Preventing Collisions at Sea (COLREGs) is difficult in multi-ship encounters. To deal with this problem, a novel method is proposed based on model predictive control (MPC), an improved Q-learning beetle swarm antenna search (I-Q-BSAS) algorithm and neural networks. The main idea of this method is to use a neural network to approximate an inverse model based on decisions made with MPC for collision avoidance. Firstly, the predictive collision avoidance strategy is established following the MPC concept incorporating an I-Q-BSAS algorithm to solve the optimization problem. Meanwhile, the relative collision motion states in typical encounters are collected for training an inverse neural network model, which is used as an approximated optimal policy of MPC. Moreover, to deal with uncertain dynamics, the obtained policy is reinforced by long-term retraining based on an aggregation of on-policy and off-policy data. Ship collision avoidance in multi-ship encounters can be achieved by weighting the outputs of the neural network model with respect to different target ships. Simulation experiments under several typical and multi-ship encounters are carried out using the KVLCC2 ship model to verify the effectiveness of the proposed method.

@en