Reinforcement Learning and Particle Swarm Optimization Supporting Real-Time Rescue Assignments for Multiple Autonomous Underwater Vehicles

Wu, Jiehong; Song, Chengxin; Ma, Jian; Wu, Jinsong; Han, Guangjie

doi:10.1109/tits.2021.3062500

Cited by 60 publications

(23 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Application Scenario Method Limitations -Open Issues [Zeng et al, 2023] Maritime Transportation Ensemble Learning Decentralized operation and scalability [Wu et al, 2022a] Maritime Transportation Data Analysis and Risk Modelling Low automation level for collision avoidance [Hesselbarth et al, 2020] Maritime Transportation Kinematic Technique using PNT Data Low level of data analysis and ML integration [Han and Yang, 2021] Security and Privacy Part-FL Additional sets of data Security and Privacy CNN and Multi-Layer Perception Realistic channel modelling [Ma et al, 2022] Security and Privacy Credibility Mechanisms Extension to more complex orientations [Qin et al, 2022] Security and Privacy Blockchain Technology Extension to more complex orientations [Zhang et al, 2022b] Fault Detection Adaptive FL Convergence improvement [Shuangzhong and Fault Detection Interpretable FL Network in a multi-level federated center [Zhao et al, 2022] Underwater Applications Interpretable FL Limited number of nodes Underwater Applications Pac-Man inspired CPP for AUVs Low ML integration and connectivity [Wu et al, 2022b] Underwater Applications RL & Particle Swarm Optimization Low decentralization [Lin et al, 2022] Underwater Applications Multi-agent RL Low decentralization (MLP) is used, where the CNN part was responsible for the feature extraction of the data and MLP performed the attack classification. Moreover, the authors also deal with the straggler problem that might appear in FL scenarios, and refered to the inability of the involved devices to timely upload their data.…”

Section: Workmentioning

confidence: 99%

“…Simulations indicate low complexity in coverage path generation and obstacle avoidance over conventional techniques. Regarding reinforcement learning (RL) integration in underwater applications, the papers in [Wu et al, 2022b], [Lin et al, 2022] proposed relevant techniques for scenarios where AUV swarms perform rescue operations and underwater diffusion pollution tracking. These solutions provide improved convergence and performance, in terms of delay and errors, surpassing other conventional optimization and ML approaches.…”

Section: Underwater Applicationsmentioning

confidence: 99%

See 1 more Smart Citation

Federated Learning for Maritime Environments:Use Cases, Experimental Results, and Open Issues

Giannopoulos¹,

Gkonis²,

Bithas³

et al. 2023

Preprint

View full text Add to dashboard Cite

<p>Maritime transportation is vital for economic growth, since it is responsible for the vast majority of global trade. However, optimizing maritime transportation, focusing on certain performance metrics may lead to non-convex problems due to the large number and heterogeneity of network nodes and vessels. Furthermore, the harsh propagation environment, and the long propagation distances might be prohibitive for the implementation of conventional optimization. Machine learning (ML) represents a viable way towards complexity minimization but still, it might not be feasible to fully exploit its potential, since error-free feedback channels are usually assumed while the overall centralized processing delay from numerous distributed sources might render real-time deployment infeasible, due to stringent latency requirements. Meanwhile, security and privacy concerns constitute key driving factors for decentralized ML solutions, since data locality is vital to protect sensitive information. Taking into consideration all the above, this paper discusses feasibility issues, regarding the deployment of federated learning (FL) solutions in maritime environments, via the presentation and analysis of various use cases. Moreover, experimental results using datasets from an enterprise specializing in the maritime industry are provided, showing the superiority of FL over traditional ML approaches, in terms of accuracy and complexity. Finally, open issues that must be addressed to pave the way for the wide adoption of FL in maritime applications are discussed.</p>

show abstract

Section: Workmentioning

confidence: 99%

Section: Underwater Applicationsmentioning

confidence: 99%

Federated Learning for Maritime Environments:Use Cases, Experimental Results, and Open Issues

Giannopoulos¹,

Gkonis²,

Bithas³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…In addition, UGVs can navigate through debris, confined spaces, and rough terrain, facilitating victim detection and information gathering. Unmanned aerial vehicles (UAVs) or drones have emerged as valuable tools for search and rescue operations [3]. Their ability to cover large areas and access hard-toreach locations quickly makes them particularly useful for locating missing persons.…”

Section: Introductionmentioning

confidence: 99%

A Novel Feature Learning-Based Bio-Inspired Neural Network for Real-Time Collision-Free Rescue of Multirobot Systems

Li,

Yang

2024

IEEE Trans. Ind. Electron.

View full text Add to dashboard Cite

Natural disasters and urban accidents drive the demand for rescue robots to provide safer, faster, and more efficient rescue trajectories. In this paper, a feature learning-based bio-inspired neural network (FLBBINN) is proposed to quickly generate a heuristic rescue path in complex and dynamic environments, as traditional approaches usually cannot provide a satisfactory solution to real-time responses to sudden environmental changes. The neurodynamic model is incorporated into the feature learning method that can use environmental information to improve path planning strategies. Task assignment and collision-free rescue trajectory are generated through robot poses and the dynamic landscape of neural activity. A dual-channel scale filter, a neural activity channel, and a secondary distance fusion are employed to extract and filter feature neurons. After completion of the feature learning process, a neurodynamics-based feature matrix is established to quickly generate the new heuristic rescue paths with parameter-driven topological adaptability. The proposed FLBBINN aims to reduce the computational complexity of the neural network-based approach and enable the feature learning method to achieve real-time responses to environmental changes. Several simulations and experiments have been conducted to evaluate the performance of the proposed FLBBINN. The results show that the proposed FLBBINN would significantly improve the speed, efficiency, and optimality for rescue operations.

show abstract

“…The neural network is demanding on the choice of data inputs while the UV cannot provide when encountering fault cases, which shares the same concern with the greedy algorithm, as the greedy algorithm needs to decompose the data for processing [28][29][30]. Hence the swarm intelligence algorithm for optimization stands out to be a preferable method to tackle the FTC application of UVs due to its flexibility of data inputs and fast convergence speed [31][32][33]. Zhu's group has applied Particle Swarm Optimization (PSO) based FTC on the unmanned underwater vehicle, though satisfactory torque outputs are achieved, the traditional PSO method shows poor real-time feedback, which does not conform to the online requirement of UV FTCs [34,35].…”

Section: Introductionmentioning

confidence: 99%

A GOA-Based Fault-Tolerant Trajectory Tracking Control for an Underwater Vehicle of Multi-Thruster System without Actuator Saturation

Zhu,

Wang,

Zhang

et al. 2023

Preprint

View full text Add to dashboard Cite

This paper proposes an intelligent fault-tolerant control (FTC) strategy to tackle the trajectory tracking problem of an underwater vehicle (UV) under thruster damage (power loss) cases and meanwhile resolve the actuator saturation brought by the vehicle's physical constraints. In the proposed control strategy, the trajectory tracking component is formed by a refined backstepping algorithm that controls the velocity variation and a sliding mode control deducts the torque/force outputs; the fault-tolerant component is established based on a Grasshopper Optimization Algorithm (GOA), which provides fast convergence speed as well as satisfactory accuracy of deducting optimized reallocation of the thruster forces to compensate for the power loss in different fault cases. Simulations with or without environmental perturbations under different fault cases and comparisons to other traditional FTCs are presented, thus verifying the effectiveness and robustness of the proposed GOAbased fault-tolerant trajectory tracking design.Note to Practitioners: This paper is motivated by the actuator saturation problem that exists in the trajectory tracking of an underwater vehicle (UV) when encountering power loss of the thruster system. The fault-tolerance trajectory tracking performance is affected by physical constraints of the vehicle when using the traditional methods as they may deduct excessive kinematic/dynamic requirements during the control process, thus inducing the deviation of the tracking trajectory. Therefore, the refined backstepping as well as the grasshopper optimization (GOA) are combined to eliminate the excess, where the refined backstepping is used to alleviate the speed jumps (kinematic outputs) and the GOA is to control the propulsion forces (dynamic outputs) when facing thruster fault cases. This innovates the industrial practitioners that the control design of the vehicle can be improved to avoid the tracking deviation brought by unsatisfied driving commands under fault cases through embedding optimization algorithms. Moreover, for the specific type of UV studied in this paper used for dam detection, simulations regarding practical dam detection such as the 3D polygonal line trajectory tracking and the frequently occurring UV single-fault cases are chosen, which can serve as references for practitioners working in the related field. In the future, underwater experiments of the UV will be investigated, with more effects of the practical environment involved.

show abstract

Reinforcement Learning and Particle Swarm Optimization Supporting Real-Time Rescue Assignments for Multiple Autonomous Underwater Vehicles

Cited by 60 publications

References 26 publications

Federated Learning for Maritime Environments:Use Cases, Experimental Results, and Open Issues

Federated Learning for Maritime Environments:Use Cases, Experimental Results, and Open Issues

A Novel Feature Learning-Based Bio-Inspired Neural Network for Real-Time Collision-Free Rescue of Multirobot Systems

A GOA-Based Fault-Tolerant Trajectory Tracking Control for an Underwater Vehicle of Multi-Thruster System without Actuator Saturation

Contact Info

Product

Resources

About