Special issue on ‘Artificial Intelligence & big data in shipping’

engineering is advancing quickly nowadays and leading the transformation of the embark diligence. A huge sum of data regarding every aspect of ship could be collected that provides fat avenues and huge opportunities to improve the occupation and operation processes of shipping. meanwhile, the boom of artificial Intelligence ( AI ) makes it possible to effectively reap the benefits of boastfully data and create huge impacts on the industry. It has been widely acknowledged that AI and big data play an ever more important role in driving efficiency and coping with approaching regulations ampere well as mounting commercialize atmospheric pressure in the transportation industry. The application of AI or big data is gaining grounds in diverse segments of the transportation industry and has pushed the fat efficiency boundaries of nautical firms. Nevertheless, newfangled challenges and concerns over their wide custom have besides surfaced. It changes the ways in which ship firms make decisions and challenges their traditional management strategies and operation modes. More importantly, it demands new AI methods, big data analytics algorithm, and optimization models to facilitate the decision-making processes at each phase of the ship industry. This special issue aims to bring together holocene technical and methodological advancements and findings concerning AI and big data in the transport industry. The first base article by Munim et aluminum. conducted a bibliometric review of 279 studies related to big data and AI in the nautical industry. interestingly, their review shows that studies on adult data and AI in the nautical industry have increased sharply since 2014 which highlights the growing importance of the subject. Further, their psychoanalysis reveals four clusters of studies which include ( 1 ) digital transformation, ( 2 ) Automatic Identification System ( AIS ), ( 3 ) energy efficiency, and ( 4 ) predictive analytics. The common methods used by the studies include support vector machine, convolutional neural net, and atom teem optimization. The university and author collaboration networks are besides depicted to facilitate future collaboration within the networks. In addition, the authors drew attention to investigating the socio-cultural and commercial aspects of AI and boastfully data to balance technical advancements, which have been the stress of most existing studies. The second and third base articles are related to achieving operational efficiency through AI. The second article by Chu et aluminum. introduced a set partitioning-based model to solve a ferry network design problem with candidate serve arch using Zhuhai Islands as a shell study. Further, they developed a hybrid variable vicinity descent-based algorithm to solve and optimise the network. Using data obtained from past fares and future demand from a survey questionnaire, their numeric analysis shows that the algorithm can lead to considerable cost savings for ferry operators. This demonstrates the superiority of the algorithm over a benchmark algorithm ( i.e. tabu search ) or manual timetables. consequently, the algorithm can be used by ferry operators to design their schedule and service network.

In the third gear article, Luan et alabama. applied a multilayer perceptron artificial nervous net to predict ship energy performance. Using operational data of over 100 container ships obtained from a major korean embark company, the accuracy of the algorithm was evaluated and compared with multiple linear arrested development. Based on respective predictive performance criteria, their analysis shows that the multilayer perceptron artificial neural network outperforms multiple linear regression. The find is invariant across different ship sizes. In accession, their results show that average sailing accelerate, voyage clock, cargo weight and ship capability can serve as commodity predictors of ships ’ fuel consumption. consequently, shipping companies can adopt the algorithm to define the optimum accelerate for their ships to minimise fuel cost and improve environmental operation. The remaining articles are related to using AI to analyse and prevent ship accidents. In the fourth article, Wang and Yin applied text mine and association rule mining on 536 inland waterborne exile accident reports between 2000 and 2018 to identify and examine the risk variables. The risk variables can be classified into four categories, namely, embark factors, environment factors, homo factors and accident factors. From analysing the categories, maritime guard recommendations such as prohibiting clog, paying care to navigation visibility, standardising shipping companies management and improving sailor competence are provided. overall, the article offers new methodological contributions and insights to maritime safety management and accident prevention by employing text mining algorithm.

In the fifth article, Jiang and Lu introduced a dynamic bayesian network model to estimate dynamic emergency hazard in ocean lanes. First, a bayesian network was developed from diachronic emergency probe reports and adept interviews. thereafter, the anterior, conditional, and transition probabilities were obtained using evidence hypothesis approach, expectation maximization algorithm, and a Markov model, respectively. ultimately, the Viterbi algorithm was adopted to estimate emergency risks using actual emergencies that occurred in the indian Ocean from 2009 to 2018. The exercise contributes to a better understanding of the propagation of respective hazard factors and the identification of those factors that contribute to sea lane emergencies. Some of the key contributors to sea lane emergencies include tip, waves, visibility, and pirate attacks.

In the sixth article, Ma et aluminum. proposed a deep support Learning ( DRL ) exemplary to solve the collision-avoidance problem of unman Surface Vehicles ( USVs ) in complex scenarios. specifically, the collision-avoidance-related rules of International Regulations for Preventing Collision at Sea ( COLREGs ) were modelled and integrated into the DRL model, which ensured the render decisions could abide the basic regulations. Besides, a collision-avoidance model based on Deep Deterministic Policy Gradient ( DDPG ) was developed as the effect of DRL model, where the reward function was designed from perspectives of current collision gamble, COLREGs residency, and the correctness of collision-avoidance action. The DRL model was proved to be effective through simulation scenarios with five USVs. This employment could facilitate the adoption of autonomous ships by providing effective methods to avoid collisions in complex and dynamic environments. In the seventh article, Xiao and Ma focused on building a nautical traffic simulator that could good reflect the realistic transport behavior during collision-avoidance maneuver. The artificial impel was developed to realise autonomous decision-making of ships under different scenarios, whose parameters were learned from AIS data in both head-on and overtake situations. This method could well capture the ship behavior in restrict waters and make the nautical dealings simulator more realistic. For the practitioners of the nautical industry, this sour could be beneficial as the simulator could provide more dependable datum for port and watercourse plan, and hazard psychoanalysis and moderation. overall, the collection of articles presents the latest research advancements in AI and boastfully data in transport. The findings and recommendations shall guide future academic research and assist the maritime industry or regulators in their decisions to improve commercial enterprise outcomes or externalities .

reservoir : https://mindovermetal.org/en
Category : Maritime