Evolution of Algorithms and Applications for Unmanned Surface Vehicles in the Context of Small Craft: A Systematic Review

Abstract

The development of autonomous vessels and unmanned surface vehicles (USVs) has generated great interest in the scientific community due to their potential and advantages for use in various environments and applications. Several literature review papers have been produced from different perspectives, contributing to a better understanding of the topic and to the analysis of advances, challenges, and trends. We hypothesize that the greatest attention has been focused on the development of high-impact applications in the maritime sector. Additionally, we depart from the need to investigate the potential and advances of USVs in fluvial environments, which involve particular operating conditions, where there are different socio-environmental conditions and restrictions in terms of access to conventional energy sources and communication systems. In this sense, the main objective of this work is to study USVs in the particular context of small craft. The search for records was conducted in Scopus and Web of Science databases, covering studies published from 2000 to 16 May 2024. The methodology employed was based on the PRISMA 2020 guidelines, which is a widely recognized protocol that ensures quality and rigor in systematic reviews and bibliometric analyses. To optimize the data collection and selection process, the semaphore technique was additionally implemented, allowing for an efficient categorization of the studies found. This combined methodological approach facilitated a systematic and transparent evaluation of the literature. This study was developed based on three research questions about the evolution of research topics, areas of application, and types of algorithms related to USVs. The study of the evolution of works on USVs was carried out based on the results of the meta-analysis generated with the Bibliometrix tool. The study of applications and developments was carried out based on information obtained from the papers for six study categories: application environment, level of autonomy, application area, algorithm typology, methods, and electronic devices used. For each of the 387 papers identified in the databases, labeling was performed for the 359 screened papers with six study categories according to the availability of information in the title and abstract. In the categories application sector, autonomy level, application area and algorithm type/task, it was identified that most studies are oriented toward the maritime sector, the developments to achieve full autonomy for USVs, the development of designs or algorithms at the modeling and simulation level, and the development and implementation of algorithms for the GNC subsystems. Nevertheless, this research has revealed a much wider range of environments and applications beyond maritime, military, and commercial sectors. In addition, from the mapping of the types of algorithms used in the GNC architecture, the study provides information that can be used to guide the design of the subsystems that enable USV autonomy for civilian use in restricted environments.

1. Introduction

The development of autonomous vessels and unmanned surface vehicles (USVs) has generated a significant amount of research due to their potential applications in the naval sector and the maritime, fluvial, and lacustrine industries as well as in studies associated with environmental monitoring. Additionally, the development of these types of systems involves addressing aspects related to route planning, navigation systems, automatic control, object and environment recognition, communications, safety, and regulation. According to [1], the growth period of studies related to autonomous and unmanned surface vehicles began in 2010 with a substantial increase in recent years.

Although previous advances in aerial and ground vehicles have provided a fundamental basis for the development of USVs, there are various challenges linked to the particularities of the different types of vehicles as well as to the complexity of the environments, contexts, and applications [2]. Numerous studies have contributed to the field of USVs, covering topics including literature reviews, bibliometric analysis, applications, vessel design, GNC system development, communications, risk management, cooperative systems, and regulatory and normative frameworks.

This work has been developed within the framework of the project “BERCO—Development of a boat focused on remote-controlled electromobility for the transport of logistic supplies from TRL 3 to TRL 5, as a strategy to validate the functionality of charging stations that use second-life batteries”, which is part of the program “TULATO—Technologies for the adoption of efficient energy and mobility systems that promote sustainable development oriented towards regions with high biosocial and energy potential such as Tumaco, Nariño”. Our hypothesis is that most of the research and industry efforts on USVs are concentrated on maritime, military, and commercial applications. This study seeks to explore the development of USVs for civilian use in confined environments, especially rivers. The main objective of this work is to identify the evolution of studies related to USVs, focusing on advancements in algorithms for achieving USV autonomy and the applications related to small craft, to guide the development of future USV technologies for the specific context of interest. The autonomy capabilities of the USV are highly dependent on the guidance, navigation, and control (GNC) architecture. While several literature reviews on USVs focus on specific subsystems or task, a comprehensive mapping of the entire GNC architecture remains lacking. Our review aims to address this gap. The review works presented in [1,3] are mainly focused on the bibliometric analysis of the identified documents. The authors present the main topics of the state of the art and establish future research directions based on the analysis performed. In our work, in addition to bibliometric analysis, the collected information is organized in such a way that it is possible to map the selected works within the framework of the GNC architecture. In addition, while the work presented in [4] focuses on civilian USV applications in disaster management, our study presents a wider range of both civilian and military applications.

To achieve the goals of this research, the study is developed based on the following research questions:

RQ1: What has been the evolution of studies on unmanned surface vehicles regarding the applications and the achievement of autonomous capabilities?

RQ2: What are the general areas in the civil and military fields in which unmanned surface vehicles are used?

RQ3: What types of algorithms have been the subject of research for achieving unmanned surface vehicles autonomy?

The main contributions made in this work, achieved through the development of these three research questions, are the following:

• A perspective of the evolution of USV studies is presented based on the information obtained with the bibliometrix analysis tool. The bibliometric analysis shows the evolution over time of academic production from 2004. Trends are identified in terms of the number of papers published per year, academic production in the five most relevant journals in which the topic has been published, and the most relevant topics.
• A mapping of the different areas of application in the civil and military fields is carried out.
• A mapping of the different task and types of algorithms used to achieve USV autonomy is carried out.
• A labeled dataset, comprising papers indexed in Scopus and Web of Science, that features six study categories: application sector, autonomy level, application case or area, algorithm type or task, methods developed or implemented, and electronic devices used in the system implementation. This dataset enables a variety of analyses of USV literature.

This document is organized as follows. Section 2 describes relevant works of literature review and bibliometric analysis, published in the last six years. Section 3 describes the methodology used for the development of the systematic literature review (SLR). Section 4 presents the results of the SLR through the development of each research question. Section 5 presents the discussion of the results obtained. Section 6 presents some study limitations ans draw the lines of future work. Finally, Section 7 presents the conclusions.

2. Previous Works of Bibliometric Analysis and Literature Review

The volume of publications in the literature on USV is quite extensive, and the topics are very varied. Several authors have contributed valuable literature review and bibliometric analysis works, which have allowed the establishment of concepts and the identification of trends and taxonomies, facilitating the study and understanding of the subject. Two types of literature review works can be identified. One group of works develops the current state of developments and trends in ASV and USV. The other group consists of studies that address specific topics related to USV. A summary of some literature reviews conducted in recent years for USV-relevant topics is provided in

Table 1. Literature reviews for USV-relevant topics.

Review Topics	References	Year
Current state and trends of autonomous vessels and USV	[1,3]	2024
	[2]	2023
	[5]	2022
Path planning	[6]	2024
	[7]	2023
	[8]	2021
Path-following control systems	[9]	2023
Adaptive control	[10]	2024
Autonomous docking	[11]	2024
Deep learning in Maritime Autonomous Surface Ships (MASSs)	[12]	2023
Decision making in MASS operations	[13]	2024
Regulation of remotely controlled and autonomous commercial vessels	[14]	2023

, outlining the respective topic and publication year.

2.1. Literature Reviews on the Current State and Trends of USV

Several literature reviews address the developments and trends of USV in general, as in [1,2,3,5]. In [1], the authors present an analysis of the advances in the development of autonomous and unmanned vessels based on a search of papers conducted in the Scopus database and using the bibliometric analysis tools VOSviewer and CiteSpace. For each of these tools, the data analysis is presented, describing the divisions of the field obtained in each case. In [2], the topic of USVs is developed within the framework of a broader review on unmanned maritime vehicles (UMVs), which also includes unmanned underwater vehicles (UUVs). The authors present for each type of UMV the fields of application, developments carried out in different countries, and technologies used in the navigation system. In addition to addressing the past and present challenges of UMVs, the authors present an analysis of the current state and trends focused on the use of artificial intelligence for the development of autonomy and cooperation in multi-vehicle systems. The work presented in [3] describes the results of a systematic literature review on USVs with papers published between 2000 and 2023 found in the Web of Science (WoS) database. The authors used the VOSviewer tool to perform the corresponding bibliometric analysis. As part of the results, six future research lines were identified. In [5], the topic of USVs is approached by presenting a context of the needs and levels of automation, the applications, the advantages and implications of their use, and some of the main technologies used in the development of USVs and their applications.

2.2. Literature Reviews on Specific Topics Related to USV

Several review papers focus on specific aspects associated with different tasks or components of the vessel’s control architecture, such as path planning [6,7,8], path-following control systems [9], adaptive control [10], autonomous docking [11], deep learning in Maritime Autonomous Surface Ships (MASSs) [12], decision making in MASS operations [13], and the regulation of remotely controlled and autonomous commercial vessels [14].

Path planning is fundamental in the context of USVs and has been widely studied. Various review papers on path planning can be found in the literature. In [6], the authors provide a chronological overview of eight previous literature review papers related to path planning, which were published from 2006 to 2023. They also present a taxonomy of path planning, categorizing algorithms into global path planning and local path planning. This article also presents a state-of-the-art review of path-planning algorithms in chronological order, which were divided into four time intervals starting from the year 2000, specifying whether the implementation was in simulated conditions or validated in real-world scenarios. The work presented in [7] conducts a state-of-the-art review of global and local planning algorithms using the same taxonomy found in [6]. Additionally, it includes the state of the art of algorithms for proximity risk avoidance and cluster path planning. The authors state that their contribution lies mainly in including aspects related to complex maritime environmental conditions in their analysis. In [8], the authors present various aspects related to autonomous surface vehicles (ASVs). This article provides a context for ASVs, the conceptual elements associated with path planning, and a timeline of literature reviews on path planning and collision avoidance. Furthermore, sections are dedicated to the regulatory framework, the architecture of the guidance, navigation, and control (GNC) system, and the path planning algorithms classified into classical, advanced, and hybrid approaches.

Path following is related to the component that various authors define as the central element for the autonomy of USVs and ASVs, which is known as the guidance, navigation, and control system [8,9]. In the literature review presented in [9], the authors explain the trajectory-tracking problem, focusing on the vessel’s guidance and control subsystems. They also present the state of the art of guidance laws and motion control systems. An important contribution of this article is the comparison of algorithms found in the literature based on the architecture structure, simulation or experimental results, guidance law used, consideration of environmental disturbances, type of controller, consideration of vessel dynamics, type of stability analysis, and consideration of degrees of freedom for the control problem.

Among the many challenges associated with the development and operation of USVs, selecting an appropriate control strategy to respond to the complexities of the environment in real-world applications is crucial. Among the approaches found in the literature, ref. [10] presents a review of Model-Free Adaptive Control (MFAC). The article provides a context for MFAC, highlighting its characteristics and advantages over conventional control schemes. The authors describe and present the state of the art of three data-driven approaches, which use neural networks, reinforcement learning, and fuzzy logic.

Autonomous docking is essential for the development of fully autonomous USVs. The developments and challenges of this specific aspect of USV operation are studied in [11]. The authors provide a description of the process and components of the architecture for autonomous docking. Additionally, the article presents a state-of-the-art review of existing methods for autonomous docking, considering aspects such as sensors, decision making, path planning and collision avoidance, and control. Based on the study conducted, the authors present the opportunities and challenges of the technology for autonomous docking, including the development of control algorithms, operation in complex environments, hardware/software integration, vessel modeling, multi-USV cooperative work, and efficient use of the existing knowledge and development framework for advancing future developments.

Artificial intelligence is a field that plays a significant role in developing autonomy for vessels and surface vehicles. Among its most notable contributions is the implementation of GNC systems that do not require complex models for vessel control and adapt to operation in dynamic environments. In [12], a literature review on the use of deep learning in MASS is presented. The authors develop a comprehensive state-of-the-art review of the works that have used deep learning in control and navigation systems, identifying the advantages and types of applications. Additionally, reference is made to works related to transport and logistics, a comparison with traditional methods developed for autonomous vessels is made, and the trends of deep learning in autonomous vessels are analyzed.

The analysis of factors involved in decision making and the role of the human component in MASS operations is addressed in the systematic literature review presented in [13]. The authors use the PRISMA methodology to select papers obtained from the Google Scholar, Research Gate, Scopus, and Web of Science databases, which is related to decision making in teams composed of autonomous systems and humans. The authors identify seven themes associated with decision making, for each of which they present the state of the art. Based on these themes, a series of design recommendations for MASSs is made, and a decision-making model based on the interaction of these seven factors is proposed. This model is validated through its application to a UAV accident situation as a case study.

One of the issues that has received less attention in the context of autonomous vessels is the regulatory and normative framework for the commercial operation of such systems. In [14], the authors present a systematic quantitative literature review on the regulatory framework applicable to autonomous vessels, using a hybrid methodology that combines traditional narrative review and the PRISMA method. The selection of papers used in this work is based on searching scientific literature in the Google Scholar, HeinOnline, and Scopus databases. The authors develop this work based on three research questions aimed at investigating the regulatory framework for autonomous vessels in Australia and other countries. Additionally, the article outlines the main lessons learned and open topics as input for future research developments.

3. Methodology

We followed the PRISMA 2020 guidelines for conducting this systematic review. PRISMA is a reporting guideline for systematic reviews and meta-analyses that is designed to ensure transparency, integrity, and accuracy in literature review or bibliometric studies [15]. The steps taken for the development of this methodology are shown in the flow chart in Figure 1.

3.1. Databases

For the search, the indexed academic databases Scopus and Web of Science (WOS) were chosen due to their ease of generating search equations and being processed jointly through analysis tools such as Bibliometrix. The databases were consulted on 16 May 2024. The databases and access information are shown in

Table 2. Databases consulted for the development of the literature review.

Data Source	URL
Scopus	www.scopus.com (accesed on 16 May 2024)
Web of Science	access.clarivate.com (accesed on 16 May 2024)

.

3.2. Search Strategy

During this phase, the keyword groups to be used in the search equations were defined. In line with the research objective, keywords were classified into 2 groups, which are described in

Table 3. Groups of keywords for the search.

Group	Keywords
Group 1	boats, boat, small boats, riverboat, small boat, small craft
Group 2	Unmanned Surface Vehicles, Unmanned Surface Vehicle, Autonomous Vehicles, Autonomous Surface Vehicles, Unmanned Surface Craft, Unmanned Maritime Vehicle, Remotely Operated Surface Vehicles, Unmanned Surface Vessels, USV

. The first group consists of keywords that refer to the product: small craft. The second group describes the characteristics and types of use expected in this product: unmanned surface vehicles. In this way, the search is aimed at retrieving documents that contain information about unmanned surface vessels, leaving the search open to the sectors of application of these vessels and the different algorithms and technologies employed.

Once the keyword groups were defined, search equations were constructed using the Boolean operators available in the databases. Given the research objective of focusing on unmanned surface vessels within the context of small craft, the decision was made to combine both groups using the “AND” operator. This operator narrows down the search results as it moves through the keyword clusters, as detailed in

Table 4. Database search equations.

Database	Algorithm Search
Scopus	( TITLE-ABS-KEY ( “boats” OR “boat” OR “small boats” OR “riverboat” OR “small boat” OR “small craft” ) AND TITLE-ABS-KEY ( “Unmanned Surface Vehicles” OR “Unmanned Surface Vehicle” OR “Autonomous Surface Vehicles” OR “Unmanned Surface Craft” OR “Unmanned Maritime Vehicle” OR “Remotely Operated Surface Vehicles” OR “Unmanned Surface Vessels” OR “USV” ) )
WOS	“boats” OR “boat” OR “small boats” OR “rivercat” OR “small boat” OR “small craft” (All Fields) and “Unmanned Surface Vehicles” OR “Unmanned Surface Vehicle” OR “Autonomous Surface Vehicles” OR “Unmanned Surface Craft” OR “Unmanned Maritime Vehicle” OR “Remotely Operated Surface Vehicles” OR “Unmanned Surface Vessels” OR “USV” (All Fields)

for each database. Search terms were applied to the “title, keywords, and abstract” fields in Scopus and “all fields” in WOS.

The bibliographic data were consolidated and analyzed using Bibliometrix 4.3.3 in the R environment, supplemented by Microsoft Excel. The database of the papers is in spreadsheet format, containing information on authors, abstract, article type, DOI, journal, title and year of publication. Duplicate records were eliminated to maintain data integrity.

Table 5. Number of papers found in the databases with the equation “group 1 AND group 2”.

Data Source	Number of Papers
Scopus	377
Web of Science	85
Total after duplicates removing	387

presents the number of papers retrieved from each database and the total number of academic papers obtained after duplicate removal.

3.3. Selection Criteria

In order to select the appropriate papers to address the questions and contribute to the objective, inclusion and exclusion criteria were established for the documents to be reviewed. The main inclusion criterion pertains to academic documents (journal papers, book chapters, and conference papers).

Two main approaches were considered for defining the inclusion/exclusion criteria: classification based on hull length and the primary architecture components required for autonomy and the USV application. The distinction criterion between small craft and larger vessels/ships is primarily based on the length of hull (LH), as established by the ISO 8666 standard [16]. This standard defines small craft as those with a hull length not exceeding 24 m. Consequently, if a craft exceeds this measurement, it is considered a vessel or ship, according to its length. Through this differentiation, applicable regulations, safety requirements, and technical specifications that various vessels must comply with are determined. In most cases, information regarding hull length is unavailable, and the identified studies are considered consistent with the context of small craft based on the search criteria.

To establish the components of interest, we relied on the architecture description presented in [17], as shown in Figure 2. The guidance, navigation, and control (GNC) subsystem is a critical component for autonomous maritime vehicles [8,9]. Moreover, the specific application of a USV is heavily influenced by the modules and equipment used for data acquisition in a given environment, upon which various types of studies are conducted or specific sector-based problems are solved. In this regard, the GNC subsystem and data acquisition are considered the key components for this study on the evolution of USV development.

Given the aforementioned considerations, the following exclusion criteria were applied according to the corresponding stage of the methodology:

• Identification: duplicate records.
• Screening: literature reviews, conference reviews, short surveys and weekly magazines.
• Eligibility: items with the following conditions.

  -

  Labeled for the design of USV, hull, propulsion or power supply.

  -

  Works in the field of major vessels.

  -

  Conference or journal papers that are published in other conference or journal with updated or extended versions.

  -

  Items that do not address the focus of the study or do not provide relevant information. This criterion includes works in the preconceptual phase with no clear indication of technological maturity level scaling.

3.4. Data Collection Process

This work studies the thematic areas and algorithms employed in the development of unmanned surface vehicles, aiming to consolidate existing literature, identify research gaps, and propose future research directions. The data collection process is performed in the eligibility stage, after having excluded duplicate items in the identification stage and items corresponding to conference reviews, short surveys and weekly magazines in the screening stage. To explore environments, application areas and algorithm types, six columns were added to the database spreadsheet to extract relevant information from papers titles and abstracts. These columns were used to label each of the papers considering the following six categories: application sector (maritime, riverine, or lacustrine), IMO autonomy level, application case or area, algorithm type or task, methods developed or implemented, and electronic devices used in the system implementation. The labeling was carried out by two members of the work team. This process also included a review and validation stage by one of the labeling managers. In cases where doubts arose, the validation was carried out with all the members of the team considering the selection criteria established for this work. Due to insufficient information in most cases, complete categorization was challenging. Nevertheless, the extracted data enabled the formulation of the proposed research questions.

Table 6. Number of papers labeled by category.

Category	Number of Labeled Papers
Application Environment	239
Level of Autonomy	198
Application Area	359
Algorithm Type	338
Method Used	174
Electronic Devices	127

summarizes the number of papers categorized by study category.

Some of the defined study categories do not directly contribute to the development of the research questions. However, they provide relevant information about the context of the reviewed works and can contribute to the identification of approaches for the development of future works.

3.5. Selection Process

As part of the paper labeling process, a color-coded system was employed to categorize papers based on their perceived contribution to the current study, specifically regarding USV application fields and the guidance, navigation, and control architecture. Papers were assigned colors as follows: green for clear and direct contributions, yellow for potential contributions, and red for those with no apparent relevance. This categorization was determined by analyzing the title and abstract of the papers. A summary of the number of papers labeled according to the contribution to the object of study is shown in the

Table 7. Number of items prioritized using the semaphore technique.

Color	Number of Papers	% Over Screened Papers
Green	238	66.30
Yellow	28	7.80
Red	93	25.91

.

The 238 papers assigned with the green color are selected for inclusion in the literature review. The distribution of papers across the categories, within this subset of the database, is presented in the

Table 8. Number of papers included in the review labeled by category.

Category	Number of Labeled Papers	% Over Included Papers
Application Environment	160	67.23
Level of Autonomy	141	59.24
Application Area	238	100.00
Algorithm Type	225	94.54
Method Used	147	61.76
Electronic Devices	107	44.96

. In light of research questions 2 and 3, it is noteworthy that the categories related to application area and algorithm type exhibit labeling rates of 100% and 94.54%, respectively.

4. Results

This section presents the results of the systematic literature review (SLR) to answer the research questions posed. Initially, Section 4.1 presents the results obtained for the study categories “Application Environments”, “Autonomy Levels”, “Application Areas”, and “Algorithm Typology”. Subsequently, Section 4.2 shows the results of the review in terms of the three research questions.

4.1. Classification of Papers According to Study Categories

This subsection presents the results of five categories considered to be the most generic, corresponding to application environments, level of autonomy, application area, algorithm typology, and methods. For each of these categories, the number of labeled items and the different subcategories identified in the study are presented for the 238 selected papers.

4.1.1. Application Environments

In the “Application Environment” category, 191 papers were labeled from the selected papers. Most papers refer to the marine environment with 89 papers. Applications for lacustrine and coastal environments are in second and third place with 22 and 20 papers, respectively. The fluvial environment occupies the fourth place with 18. Finally, there is a group of papers that refers to applications in environments such as ponds, reservoirs, and wetlands, which contributes 42 papers. In 28 papers, two or more application environments were identified. On the other hand, a total of 78 papers were not tagged, as the environment information was not found explicitly. In this case, it is considered that most of these works are carried out on topics applicable to any of the environments. However, further in-depth study is required to perform an adequate classification. The distribution of the number of categorized papers according to the application environment is shown in

Table 9. Number of papers labeled in the category “Application Environments”.

Application Environments	Number of Papers
Maritime	89
Lacustrine	22
Coastal	20
Fluvial	18
Other (ponds, pools, reservoirs)	42

.

4.1.2. Autonomy Levels

In the “Autonomy Levels” category, 141 papers were tagged, of which 126 were classified as “IMO: fully autonomous vessel” and 23 as “IMO: remotely controlled uncrewed vessel”. The remaining 97 papers lacked explicit information on autonomy levels. It is assumed that most of these studies contribute to developing fully autonomous USVs, but further research is needed for a more accurate classification.

Table 10. Number of papers labeled in the category “Autonomy Levels”.

Level of Autonomy (IMO)	Number of Papers
Fully autonomous vessel	126
Remotely controlled uncrewed vessel	23

presents the distribution of papers by autonomy level.

4.1.3. Application Areas

In the “Application Areas” category, the 238 selected papers were tagged based on information extracted from their titles and abstracts. This category directly provides the data needed to address the research question concerning the general civil and military application areas of unmanned surface vehicles. The identified areas were classified into 10 groups, as presented in

Table 11. Number of papers labeled in the “Application Areas” category.

Application Area	Number of Papers
Academic	128
Environmental Monitoring	30
Naval/Security	24
Bathymetry/Cartography	22
Risk and Disaster Management	13
Aquaculture/Fishing	8
Oceanography	8
Hydrography/Hydrology	5
Transportation/Tourism/Ports	3

. Firstly, there is a group of 128 papers categorized as purely academic works. These papers focus on designs or algorithms at the modeling and simulation level without specifying a particular application field. Furthermore, the lower construction and operational costs of USVs, coupled with their ability to access restricted areas, have driven the development of studies with practical real-world applications, such as data acquisition for environmental monitoring, bathymetry, oceanography, and hydrography. Additional practical applications are directed toward the naval, fishing, disaster management, and transportation sectors.

4.1.4. Algorithm Typologies

The “Algorithm Typologies” category was assigned to 194 of the selected papers based on information extracted from their titles and abstracts. This category directly provides the data needed to address the research question concerning the types of algorithms used in the development and operation of unmanned surface vehicles. The works are mostly related to path-planning algorithms. In total, 24 types of algorithms were identified, as shown in

Table 12. Types of algorithms identified.

Algorithm Type	Number of Papers
Data collection	40
Path planning	32
Cooperative robotics systems	26
Obstacle avoidance	17
Environment perception	16
Trajectory tracking	16
Collision avoidance	15
Obstacle detection	15
Path following	14
USV state estimation	13
Control	11
Object detection	9
Position control	8
Target tracking	8
Heading control	6
Motion control	6
Course control	5
Autonomous docking	4
Trajectory planning	3
Target detection	3
Sensor fusion	2
Remote control	2
Heading and speed control	2
Mission planning	2
Target localization	2
Application-specific task	8

. The information obtained indicates that the tagged papers can be grouped based on the USV architecture. This grouping is developed in Section 4.2.3, which addresses the research question about the types of algorithms used in USVs.

4.2. Research Questions Results

In this subsection, the results for the development of the three research questions are presented, which are based on the information obtained from the papers labeled in the database according to the study categories. To address research questions 1 and 2, we analyzed all labeled articles. For research question 3, given the specific selection criteria related to USV autonomy algorithms, the analysis was centered on the selected articles. This focused approach allowed us to identify the different approaches used in studies within the GNC architecture framework.

4.2.1. What Has Been the Evolution of Studies on USV Regarding the Applications and the Achievement of Autonomous Capabilities?

The results obtained show that the topic is relatively recent. The first documents in the databases are recorded from the year 2004. Starting in the year 2015, they began to increase significantly, and most of the production has occurred in the last 5 years (2019–2024) with 54.43% of the total as shown in Figure 3. The trend is expected to continue in 2024.

The produced documents have been published through various sources, as shown in Figure 4. Initially, the journal dedicated to strategic and military security issues, Jane’s Defence Weekly, was the only one; in 2007, production on this topic ceased. Subsequently, in 2010, The IEEE International Conference of Intelligent Robots and Systems, a journal on robotics, began publishing and has since maintained relevance, with a higher cumulative total compared to the rest, along with Ocean Engineering, which is dedicated to research and development in naval engineering. Based on this, it is pertinent to appreciate that unmanned vehicles encompass a broad scope of interest that includes the naval industry, intelligent systems, and strategic military purposes.

The thematic evolution confirms the aforementioned. Figure 5 shows that initially, the trending topic was military operations. Subsequently, hulls, navigation systems, and unmanned vehicles gained relevance along with the introduction of the term “ocean engineering” to categorize the research carried out in this field. Since 2008, remotely operated vehicles have maintained their presence in the thematic, alluding to the importance of these vehicles being able to operate without an onboard crew. For the years 2011–2015, robotics in this area began to be discussed as well as oceanography and autonomous navigation. Finally, for the years 2015–2017, the term “unmanned surface vehicles/vessels” was introduced, and with it, terms such as “obstacle avoidance”, “maneuverability”, “object detection”, “controllers”, and others, indicating evidence of the growing interest in achieving autonomy in these vehicles through navigation and control technologies.

The evolution of academic production, considering the document type, shows an increasing trend in the number of articles, both journal and conference papers, with a more pronounced growth starting from 2010. Figure 6 shows the chronological evolution of the quantity of journal and conference articles. It can be observed that in most years, more conference papers have been published than journal articles. Additionally, fluctuations in both cases are evident. In this systematic review, starting from the screening stage, according to the document type, 206 conference papers, 150 journal articles, 2 book chapters, and 1 editorial were used.

4.2.2. What Are the General Areas in the Civil and Military Environments Where Unmanned Surface Vehicles Are Used?

Hydrodynamic modeling technology and marine vehicle control systems have progressed significantly in recent years. In particular, USVs have found a number of fields where their application is highly useful [18]. In fact, they have gone from being considered as heavy and expensive equipment to viable instruments for multiple scientific and commercial applications [19].

In this sense, based on the bibliographic review carried out, it is identified that civil applications are more widely disseminated in the scientific field; in fact, more than 80% of the reviewed works, which corresponds to approximately 334 references, present applications of unmanned vehicles in the civil environment, while only 51 works describe applications in the military or defense field. Figure 7 illustrates the percentages of the identified applications.

When reviewing applications in the civil field, it was identified that the largest proportion of works related to USVs has been developed in the academic field, which is followed by works in the areas of environmental monitoring and bathymetry. Some minor applications correspond to the areas of disaster management [4,20], fishing [21] and hydrography [22]. Figure 8 presents the percentage distribution of the main application areas of USV technologies in the civil field.

Within the applications in the academic area, works related to target tracking [23,24] and trajectory tracking [25], obstacle avoidance [26], path planning [27,28,29] and control techniques [30,31] stand out.

On the other hand, in the environmental area, works stand out in the lines of water sampling/water quality [32,33], nuclear/oceanic radiation [34,35], current studies [36] and ecological protection [37], among others.

Likewise, as part of the main applications of bathymetry, some works developed in the Arctic environment [38], in lakes [39,40] and in the marine/coastal environment [41,42,43] stand out.

In the military field, the largest proportion of works developed are related to applications focused on security and defense, surveillance and reconnaissance, interception/interdiction, and intruder detection. Figure 9 presents a summary of the main application areas of USVs in the military environment.

In the area of security and defense, works related to maritime border security [44], port security [45] and merchant escort [46] stand out. As part of the works in the area of surveillance and reconnaissance, some are focused on patrolling hostile environments with civilian traffic [47] and coastal surveillance [48], among others.

The distribution of applications identified in the selected articles is presented in Figure 10.

For applications related to monitoring and data collection, the studies shown in

Table 13. References for the papers labeled in the ”Application Areas“ category related to data acquisition for specific applications.

Application Area	Number of Papers	References
Environmental Monitoring	30	[12,32,35,37,41,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73]
Bathymetry/Cartography	22	[38,39,40,41,42,43,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89]
Risk and Disaster Management	13	[20,34,52,88,90,91,92,93,94,95,96,97]
Aquaculture/Fishing	8	[21,98,99,100,101,102,103,104]
Oceanography	8	[36,105,106,107,108,109,110,111]
Hydrography/Hydrology	5	[112,113,114,115,116]
Transportation/Tourism/Ports	3	[117,118,119,120,121]

were identified, according to their application areas in the civil environment.

4.2.3. What Types of Algorithms Have Been the Subject of Research for Achieving USV Autonomy?

In the process of labeling the papers according to the algorithm type, 22 subcategories were identified. However, some of them can be grouped considering the relationship between them in terms of the USV development stage, the type of task, or the component of the USV GNC system architecture.

Table 14. References of the identified types of algorithms grouped according to the GNC architecture subsystems.

Algorithm Type	Papers	References	Subsystem
Environment perception	16	[25,49,82,118,122,123,124,125,126,127,128,129,130,131,132,133]
Obstacle detection	15	[26,49,122,134,135,136,137,138,139,140,141,142,143,144,145]
USV state estimation	13	[56,70,119,146,147,148,149,150,151,152,153,154,155]
Object detection	9	[44,150,156,157,158,159,160,161,162]	Navigation
Target tracking	8	[23,24,160,163,164,165,166,167]	(69)
Target detection	3	[24,168,169]
Sensor fusion	2	[170,171]
Target localization	2	[99,172]
Path planning	32	[20,28,29,36,37,45,47,76,118,128,129,132,141,150,164,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189]
Obstacle avoidance	17	[25,31,45,121,139,140,156,160,167,181,182,183,186,190,191,192,193]	Guidance
Collision avoidance	15	[27,164,174,179,180,187,190,194,195,196,197,198,199,200,201]	(71)
Trajectory planning	3	[165,193,202]
Mission planning	2	[203,204]
Trajectory tracking	16	[25,31,62,98,118,165,193,203,204,205,206,207,208,209,210,211]
Path following	14	[56,76,77,84,187,189,201,212,213,214,215,216,217,218]
Control	11	[42,97,104,116,128,150,219,220,221,222,223]
Position control	8	[23,31,39,112,221,224,225,226]
Heading control	6	[36,62,227,228,229,230]	Control
Motion control	6	[119,204,212,231,232,233]	(74)
Course control	5	[66,173,217,234,235]
Autonomous docking	4	[156,236,237,238]
Remote control	2	[146,239]
Heading and speed control	2	[43,240]

presents the references related to GNC architecture systems, which are categorized by subsystem.

The control subsystem emerged as the most prevalent research topic with 74 papers primarily focused on trajectory tracking. Guidance systems constituted the second category, encompassing 71 studies. Path-planning algorithms were the core focus of nearly half of these papers. While obstacle and collision avoidance was addressed in 32 papers, it is inherently linked to path planning. Trajectory planning, considering both spatial and temporal dimensions, was explored in a smaller subset of four papers. Navigation subsystem research comprised the third category with 69 studies predominantly focused on environment perception.

The developed database also provides information about the methods used to address the corresponding tasks. This information was identified for 147 articles.

Table 15. Methods implemented for environment perception and obstacle detection.

Task	Method/Algorithm	Reference
Waterline detection/Obstacle detection	Image segmentation	[49]
Wind speed and direction estimation	Neural networks–Perceptron	[124]
Estimation of meander parameters	Gaussian filters/Restricted interval Kalman filter	[125]
Navigable waterway detection	Deep learning-based semantic segmentation/Planar homography/Line detection	[128]
Background segmentation and change detection	Background subtraction	[130]
Coastline-water detection and recognition	Line segment detection/coarse-to-fine strategy	[133]
Obstacle detection	Sensor fusion/Weighted ELM binary classifier	[134]
Miltimodal perception for obstacle detection	CNN–YOLO V7	[135]
LiDAR-based ambient detection	Sensor data fusion/Voxel filtering	[136]
Hallucinating hidden obstacles	Compositional model	[137]
Temporal context for obstacle detection	Temporal context extraction from image sequences for ambiguity reduction	[138]
Obstacle avoidance system	CNN–YOLO V4/Vector Field Histogram (VFH)	[139]
Obstacle detection/Obstacle distance ranging	Fuzzy Kohonen Network (FKN)	[140]
Obstacle detection	Segmentation	[141]
Stereo obstacle detection	Semantic segmentation	[26]
Real-time stationary obstacle detection and localization	Robust two-step outlier rejection method	[143]
Real-time obstacle detection	SKIP-ENET segmentation model	[144]
Small obstacle segmentation/Obstacle map estimation	Efficient semantic segmentation networks/Efficient Multi-Feature Aggregation (MFA) module/Gaussian mixture model-based Feature Separation (FS) loss function/FASNET	[145]

presents information on articles related to environment perception and obstacle detection. The remaining information is available in the database and is not included in this article due to its length and quantity.

5. Discussion

This literature review examines the evolution and development of two key aspects of USVs: applications and algorithms used for achieving USV autonomy. While the work presented in [4] is focused on civilian USV applications in disaster management, our study presents a wider range of both civilian and military applications. Other review works, presented in [1,3], provide a comprehensive overview of the main topics of the state of the art and the future research directions based on the bibliometric analysis. In our work, the collected information is organized in such a way that it is possible to map the selected works within the framework of the applications and the GNC architecture. The state of the art is further enriched by analyzing data collected from the review based on our three research questions and six study categories. The results of this systematic literature review on USVs in the context of small craft allow us to identify several relevant aspects for defining guidelines oriented toward the development of USVs and applications for the specific context of fluvial environments in areas with technological and socio-environmental constraints. The methodology used, in addition to obtaining relevant bibliometric information, allowed us to identify key information to support needs in terms of research and development. To answer the three research questions, six study categories were proposed, of which three were used as the main input. The remaining three categories provided complementary information or can be used in future work to expand the present study and identify potential lines of future work. In general terms, based on these categories, it was identified that most of the reviewed works are oriented toward developments for marine environments and systems to achieve the full autonomy of vessels. In addition, applications in the civilian field are identified, which are focused mainly on data acquisition for environmental scientific studies. Furthermore, there is a marked interest in developing path-planning algorithms, particularly those related to obstacle and collision avoidance. The discussion related to each of the research questions follows.

5.1. Evolution of USV Studies

Research in the field of USV is a relatively recent development, gaining significant interest in the last 10 years. In review papers such as the one presented in [1], the authors identify a period of growth in academic production related to autonomous and unmanned vessels starting in 2010. In our study, this growth stage is clearly evident from 2014 onwards. This trend is more similar to the phase of rapid growth between 2014 and 2019 reported in [3]. There is a coincidence of these years with some critical points observed in the graph of production of the most relevant sources over time shown in Figure 4. According to the trends shown in Figure 5, it is identified that in the early years, there are mainly reports of applications in the military field. Subsequently, in 2016, there are applications in the civil field related to oceanography and three years later with bathymetry. On the other hand, the trend in terms of focus allows us to identify the following order of trend topics in terms of the GNC architecture: navigation system between 2011 and 2018, control system in 2019 and guidance systems from 2021 onwards. With respect to the level of autonomy, remotely controlled unmanned vessels appear in 2009 and 2021, while the term autonomous navigation appears from 2016 onwards.

5.2. Areas of Application in the Military and Civilian Fields

The majority of the reviewed papers focus on designing or implementing algorithms to achieve vessel autonomy. Consequently, these papers were classified as primarily academic. Such works can contribute to both naval and civilian applications. While these studies have the potential to significantly advance real-world applications, many remain at the simulation or early prototype stage. Conversely, this study reveals a substantial body of literature where the specific application of the USV is explicitly identified. Civilian applications significantly outnumber military ones, with a primary focus on data collection for environmental monitoring and aquatic ecosystem studies. Bathymetry has emerged as a recent trend, as illustrated in Figure 5 around 2019. The extensive use of USVs in bathymetric surveys, as well as other environmental and oceanographic studies, is primarily due to their cost-effectiveness, operational safety, and ability to access restricted areas.

5.3. Types of Algorithms Used for USV Development and Operation

The review results show the navigation system addresses different challenges associated with data acquisition and environmental perception, having a considerable participation among the studies found. On the other hand, works related to guidance system and control systems have a greater development related to path-planning and collision avoidance algorithms, which is followed by control strategies associated with path or trajectory tracking. In addition to the works that involve the different components of the GNC architecture, two more groups related to applications stand out. On the one hand, there are works related to missions that address general aspects such as planning and others more specific such as target detection and tracking. Due to their characteristics, these developments are associated with the military field. On the other hand, cooperative systems constitute the other group, where systems with multiple unmanned vehicles are proposed that can be homogeneous or heterogeneous. In these collaborative robotics works, algorithms for formation control are found and have applicability in both the civilian and military fields.

6. Limitations and Future Work

This systematic literature review has resulted in a database of scientific papers on USVs, focusing on applications and algorithms for small craft. In addition to the corresponding bibliometric information, the elaborated database contains information on environments and application areas, level of autonomy, algorithms, and electronic devices. In the present article, some of these categories are analyzed to answer the research questions, but the different relationships that may exist between them are not explored. Furthermore, the database could be enriched by a more in-depth review of the selected papers due to limitations in abstract information. Given the significant amount of papers in the database, automated tools can be valuable for extracting and validating information. In addition, the scope of this study is limited to the Scopus and Web of Science databases. The database and analysis could be expanded by applying the methodology to other scientific databases.

The volume of research dedicated to maritime applications is substantial compared to other domains. A promising avenue for future research involves a more in-depth analysis of works focusing on riverine, lacustrine, and other confined aquatic environments. This analysis aims to identify the unique needs, challenges, and distinct requirements of these settings compared to maritime environments.

7. Conclusions

This systematic literature review investigates the evolution of Unmanned Surface Vehicle (USV) developments and applications in small craft. Employing the PRISMA methodology, the study analyzes scientific publications retrieved from Scopus and Web of Science. Three research questions guide the investigation: the evolution of USV research, civil and military applications, and the algorithms used in USV development and operation. Bibliometric analysis and manual categorization were used to answer these questions. The results identify trends in USV research, map applications, and analyze the algorithms used in USV guidance, navigation, and control (GNC) systems. Numerous civil applications, particularly data acquisition for environmental and oceanographic studies, were identified. Additionally, the study highlights the significant development of path-planning and collision avoidance algorithms. This research contributes to the state-of-the-art in autonomous and unmanned vessels, providing a baseline for researchers seeking to develop USVs for applications in technologically and socio-environmentally constrained contexts.