1. Introduction
The ocean is a repository of diverse resources and includes a variety of ecosystems (Kang et al., 2019). Based on this environment, various leisure activities, such as scuba diving and snorkeling, are performed in the space (Cho, 2024). Recently, with increasing engagement in marine leisure activities and the expansion of related industries (Ministry of Oceans and Fisheries, 2019), concerns have been raised over safety. According to statistics on accidents related to marine leisure activities, there were seven deaths among 2,186 accidents in 2013 and three deaths among 1,463 accidents in 2014 on beaches. The main causes of the accidents were individual negligence, such as poor swimming skills and alcohol consumption. For the accidents by water leisure craft, a total of 36 accidents occurred in 2014, and most of them were also caused by individual negligence, such as negligence in navigation (Kwon and Bing, 2015). In the case of scuba diving, the types of the 310 diving accidents reported during the period from 2000 to 2013 were analyzed. More than 90% of them led to deaths, and the causes of the accidents were found to be safety accidents, mistakes, and inadequate training (Kwon and Pyo, 2015). Marine leisure-related accidents are caused by individual negligence in many cases. In the case of scuba diving, in particular, more attention is required because even a minor mistake may lead to death. Conversely, if individuals comply with safety regulations, they will be able to enjoy marine leisure activities more safely.
In scuba diving, a buddy refers to a diver paired within a group of two or more divers to ensure the safety of each other and provide assistance. Buddies continuously check each other’s equipment and status. Furthermore, they serve as safety supervisors who provide necessary support in emergencies, establish diving plans, and support and guide one another (Mišković et al., 2015). There are only a few cases of underwater robots that perform these roles. A representative study is the cognitive autonomous diving buddy (CADDY), which was conducted from 2014 to 2016. As part of the FP7 project, seven universities and research institutes participated to develop a diver behavior monitoring and mission performance support system and an underwater robot system that adapts to divers’ behavior and health conditions with high cognitive capabilities (Abreu et al., 2015). Various underwater robots, including BUDDY, have been developed to perform and verify the developed functions. They monitored divers’ missions from the optimal location by following them, guided the divers to their destination, and executed tasks (e.g., taking photos and supplying tools) based on the gesture commands of the divers (Egi et al., 2015). The roles of a diver safety support robot as a buddy are similar to those of CADDY, but more active safety support can be provided based on diving scenarios. It conducts preliminary inspections of diving points and provides guidance, such as ascent, descent, and observation, based on the scenarios. It also tows divers or functions as a scooter when necessary and can rescue divers in emergencies. Among similar studies, Wu et al. (2022) conducted research on a control method to replace the net inspection performed by divers with robots. They completed the task in a stable manner, but the robots do not assist divers’ direct activities. Yang et al. (2016) applied the Anchor Diver 5.2 system to the inspection of social infrastructure (e.g., pipes, dams, and bridges), which is performed by divers, to overcome the limitations of divers and improve the efficiency and accuracy of inspection. They demonstrated the potential of the system in the field, but it does not directly support divers’ activities; it only performs inspections on behalf of divers. Fattah et al. (2016) developed a remote robot diver system called R3Diver to perform underwater search and rescue operations. It exhibited satisfactory results at various depths but may cause safety problems (e.g., divers entangled with cables) during operation with divers because cables are essential for its operation. Kartal and Cantekin (2024) proposed a more efficient and practical alternative by replacing the method of inspecting natural gas pipelines, which heavily relies on divers, with a remotely controlled underwater robot. They verified the robot’s performance through tests in a water tank, but the robot requires remote control that utilizes cables and does not directly support divers. Most previous studies were focused on replacing divers, and it appears that there is no study on collaboration with divers except for research related to the CADDY project.
This study aimed to define the functions of the robot-based nautical advanced diving intelligent assistance system (NADIA) and derive conceptual design plans prior to its development, as shown in Fig. 1. Based on various literature reviews that include marine leisure activities, safety regulations, underwater environments, and safety accident analysis, consultations and surveys with marine activity-related experts, and direct diving experience were analyzed and summarized to derive the concept of supporting divers with robots. Robot-based diving scenarios were also defined. In addition to the definition and analysis of existing diving scenarios, based on the defined robot-based diving scenarios, robot functions were defined, including the information required for divers and robots, essential robot functions, and methods for displaying information and convey alerts. Specific design and production directions were determined by implementing the defined functions and reflecting the size and weight derived from consultations to improve the applicability of NADIA and enable successful task performance. As aforementioned, most previous studies aimed to replace divers with robots to perform certain tasks. Thus, the forms and functions of robots were limited and user requirements were not fully reflected. Due to these limitations, procedures such as the analysis of required robot functions and concept design were not performed systematically. Consequently, the developed robots stagnated in the research stage or were not commercialized. Meanwhile, this study aimed to assist underwater activities and support the safety of divers, with a focus on the diver–robot interaction. Therefore, a wide range of requirements were derived from divers in various fields, and an in-depth analysis of underwater environments and diving procedures was conducted considering the safety of divers. In addition, this study aimed to apply the robot to the actual marine leisure industry beyond simple scuba diving support. Against this backdrop, the process of deriving the functions and conceptual design plans for NADIA in accordance with the procedure presented in Fig. 1 is essentially required. This is expected to make it possible to develop a practical and user-centered diver support robot.
In Chapter 2 of this paper, existing literature and diver scenarios are analyzed. In Chapter 3, robot-based diving scenarios are set, and robot functions as well as conceptual design plans are derived based on the contents of Chapter 2. Finally, conclusions are presented in Chapter 4.
2. Analysis of Diving Scenarios: Research and Field Study
As aforementioned, diverse literature was investigated, including books related to marine leisure and underwater environments, as well as reports related to accident types and countermeasures, to identify major safety-related issues along with an understanding of underwater environments. Based on this, requirements for diving scenarios, diver safety, robot operation, and application extension were defined through consultations and surveys with experts related to marine leisure and safety, and diving was directly experienced to understand the actual underwater environment and diving procedure. This procedure is shown in Fig. 2. Existing diving scenarios were also defined and analyzed to define the most practical and effective method for the robot to support divers underwater.
2.1 Literature Survey
To investigate the procedure of scuba diving, safety guidelines, and marine environments, literature, including internet information, books, and reports, was surveyed and summarized as follows. In the survey of similar studies, the CADDY project was found to be most similar to this study. The project monitored divers’ mission from the optimal location by following them, guided the divers to their destination, and executed tasks (e.g., taking photos and supplying tools) based on the gesture commands of the divers (Egi et al., 2015). While the CADDY project aimed to develop functions such as diver tracking, monitoring, guide, and recognition, this study focused on supporting the overall scuba diving process as a buddy. There have also been studies by Wu et al. (2022), Yang et al. (2016), Fattah et al. (2016), and Kartal and Cantekin (2024), but these studies were mostly focused on the development of robots to replace divers, which differed from the purpose of this study.
2.1.1 Underwater environments
In general, underwater environments have higher density than air, and the transmission speed of light or sound is different from that on land. Thus, the senses of humans (e.g., vision and hearing) are amplified compared to those on land, and condition similar to zero gravity can be felt (SDI, 2019). In the case of vision, objects look approximately 33% larger and 25% closer than they really are. Vision or the sense of direction becomes dull, and dizziness may occur. At this point, it is necessary to stop, hold a fixed object, and wait until the conditions improve. In the case of light, underwater view is not good because light is not transmitted due to floating matter. Because the location of the diver or buddy can be lost, sufficient training is required. In the case of sound, it travels approximately four times faster than in air. Many sounds (e.g., breathing and engine sounds) can be heard, and a slight difference in the sound that reaches both human ears can make the sense of direction dull. Therefore, it is necessary to visually identify the buddy. If a ship’s engine sound is heard, it is required to stay in place until the sound is not heard and then ascend to the water surface along the surface marker buoy (SMB) or the anchor line of the boat (SDI, 2019).
2.1.2 Diving physics
Divers are exposed to various physical environments underwater. The main physical phenomena that affect divers underwater are buoyancy and pressure. When divers breathe underwater, they consume the same volume of air as on land. Because air with increased density underwater is consumed at the same volume, the air in the tank is consumed faster compared to breathing on land. Therefore, divers should dive considering the amount of air remaining in the tank underwater and the depth of water (SDI, 2019).
2.1.3 Diving physiology
Divers are affected by high pressure during diving, which has various impacts on the human body. When a diver descends underwater, the volume of the closed air space in the body of the diver contracts like the contraction of a balloon, resulting in compression. The human body is damaged by such compression, especially the ears, lungs, sinus, and mask (eyes). Such compression can be resolved through pressure equalizing; This method balances the pressure inside and outside by adding air to the closed space where compression occurs. It is effective to apply this method frequently before compression occurs, i.e., immediately before and during descent. If there is discomfort or pain in the ears during descent, it is necessary to stop the descent immediately, ascend until the ears become comfortable, and descend again while performing equalizing. If breathing is stopped during diving, lung compression occurs, resulting in an emergency situation. Thus, breathing should never be stopped during diving (SDI, 2019).
2.1.4 Major safety rules for diving
Major safety rules should be followed for safe diving. Divers should dive only when they are in good physical and psychological condition. Plans must be established before diving, and diving must be performed according to the plans. Diving should never be performed alone, and divers should descend and ascend at a safe speed based on the procedure. Finally, a constant breathing pattern should be maintained during diving, and breathing should never be stopped.
2.2 Consultations and Surveys with Experts
To obtain quality data from various fields of diving from open water to industrial diving, consultations and surveys were conducted with nine professional divers. The professional divers were representatives or senior-level experts, and they were engaged in associations and educational institutions, leisure resorts and shops, marine rescue, and industrial and professional diving fields. Various opinions related to diving scenarios, diving equipment, operation, and safety were collected and analyzed through meetings with the diving experts from each field. As shown in Table 1, the results were classified into four categories: scenario, safety, operation, and function. The contents of each category can be summarized as follows. The robot should be operated by the master diver for efficiency of operation, and it should be able to guide a beginner diver depending on the situation. In emergency situations, the robot should be able to inform divers with sound or light rather than direct actions. The robot should be able to tow divers when necessary and ensure the safety of divers by installing safety lines. For beginner divers, it is difficult to control speed during ascent and descent. The robot should be able to guide movement and induce breathing and equalizing. The robot should have a size and weight that make it easy to operate.
2.3 Direct Diving Experience
An attempt was made to directly acquire open water and advanced courses to obtain basic knowledge on underwater environments, a variety of diving equipment, and skills and improve the understanding of diving. The common items felt by each participant were summarized from the perspectives of the divers and robot, respectively, as shown in Tables 2 and 3. Regarding key items related to diving, (1) beginner divers need a descent line or the help of the master diver because it is difficult to control their posture underwater. (2) They should visually confirm the movement direction or the buddy and pay attention to their speed during ascent. Regarding considerations for the robot, (1) the robot should provide divers with diving information during ascent or descent. (2) To this end, the robot should be within the view of the divers. (3) It should also perform monitoring to maintain the distance between the divers. (4) In case of an emergency, it should share the situation through alarms, such as buzzers.
2.4 Diving Scenario Analysis
An analysis of diving scenarios was conducted to understand the diving procedure and derive diver support scenarios using the robot. In general, diving scenarios are mainly divided into seven steps, as shown in Fig. 3. Detailed procedures and preparations for each step are discussed below. Diving scenarios also include emergency situations and countermeasures. Because all situations in the marine environment could not be considered, two representative situations were derived through literature and expert consultations, and countermeasures were defined.
The derived emergency situations were (1) diver position deviation (increased distance from the buddy) and (2) abnormal diver condition (e.g., panic). Here, preventive measures for diver position deviation are to identify the position of the buddy at all times and to inform the master diver of the situation or buddy search, or return to the surface is performed in a predefined routine if an accident occurred. Preventive measures for abnormal diver condition are to continuously check the condition of oneself and the buddy and to inform the master diver of the situation and stop diving immediately if abnormal conditions are identified.
(1) Diving preparation: This refers to a briefing to share diving information and a procedure to check diver safety equipment. In the briefing, the location of the diving site, underwater terrain, environment, weather conditions, mutual hand signals, boat, and diving and exit methods are shared. A dividing schedule is also established, including the time in water, minimum amount of water at the time of exit, underwater conditions, and countermeasures in case of an emergency. A consent form is prepared to examine the health status (e.g., conditions and underlying diseases) of divers. Diver safety equipment is checked through the ABCDE process, where (E) is excluded from the preparation process.
(A)ir check (air pressure and valve opening status): check the residual pressure gauge during breathing through the primary regulator and octopus regulator.
(B)cd check (buoyancy regulator): check the status of the inflator and deflator as well as the status of the dump valve.
(C)omputer: check the dive computer operation and power.
(D)ive gear (putting on gear): check gear, including goggles and knives.
(E)ntry (diver entry): diver entry.
(2) Travel to dive site and pre-dive research: Travel to the dive site involves travel to the coast, but this study deals with travel by boat. Divers board the boat with prepared equipment. The equipment is well-fixed during boat movement for safety reasons, and divers sit on the prepared seats.
(3) Diver entry and descent: Diver entry methods include standing, jump, backward, seated, and walking entry. Entry is performed using the pre-defined method according to the support of the boat. Entry is performed using the ABCDE process. Here, the (E) process completes entry. After entry, divers descend after checking the buddy under the control of the diver master (leader) on the surface. During descent, it is necessary to maintain the equilibrium posture and alternately check the buddy (keeping a distance of 1 to 2 m from the buddy) and diver computer (check the depth). Beginners, however, may proceed in the vertical posture while the other items (e.g., view) remain the same. Divers descend at a speed of 0.34 m/s or less (0.08 m/s or less for beginners) while constantly performing equalizing. During descent, it is necessary to check the condition of oneself and the buddy to prevent accidents. In case of an emergency, it should be properly handled using the knowledge obtained in advance.
(4) Underwater tourism: Upon reaching the planned depth, it is necessary to check the pressure gauge of the air tank and the buddy under the control of the diver master. Underwater observation is then performed together with the buddy through objective instructions. During the observation, the equilibrium posture is maintained considering the protection of marine life, and it is necessary to check the residual pressure gauge, diver computer, and buddy at all times. In emergency situations, such as breakaway of the buddy and abnormal conditions, they should be properly handled using the procedures reviewed in advance.
(5) Travel to the tourist site: Movement to the observation site can be divided into movement after exit and underwater movement. Movement after exit follows the same procedure from the boat movement process after diver exit. In the case of underwater movement, divers stop activities and gather under the control of the diver master and then check the pressure gauge of the air tank and their buddies. After that, divers move to the next site with their buddies through objective instructions. During movement, it is necessary to maintain the equilibrium posture considering marine life protection and to check the residual pressure gauge, diver computer, and buddy at all times. In emergency situations, such as breakaway of the buddy and abnormal conditions, they should be properly handled using the procedures reviewed in advance.
(6) Diver ascent and exit: Divers stop activities and gather under the pre-determined diver schedule and control of the diver master, and then they check the pressure gauge of their air tank and their buddies. Then, they begin ascent through the ascent instruction. During ascent, it is necessary to maintain the equilibrium posture and alternately check the buddy (keeping a distance of 1 to 2 m from the buddy) and diver computer (check the depth). Beginners, however, may proceed in the vertical posture while the other items (e.g., view) remain the same. Ascent is performed at a speed of 0.3 m/s or less (0.15 m/s or less for beginners). After safety stop for 3 to 5 min at a depth of 3 to 5 m, final exit is performed. During ascent, it is necessary to check the condition of oneself and the buddy to prevent accidents. In case of an emergency, it should be properly handled using the knowledge obtained in advance.
(7) Return to the boat: Divers board the boat using the method discussed in advance. Devices are dismantled and then well-fixed for safety during boat movement. After simple washing, divers are seated for return.
3. Conceptual Design and Functional Definition of a Diver Support Robot
In this chapter, the concept of diver support and the functions of the robot are defined based on the analyzed data to derive conceptual design plans for the development of the robot.
3.1 Diver Support Concept Definition
In the summary of the contents mentioned above, (1) the robot has high degrees of motion freedom to improve the efficiency of operation, along with a lightweight and compact form for dexterous movement. (2) It should be able to estimate the diver’s position and posture and recognize obstacles in front based on artificial intelligence (AI). (3) Because the robot operates in close proximity to divers for guiding and towing, it should be equipped with precise autonomous control technology and self-localization technology. The position of underwater objects can be estimated using an ultra-short baseline and acoustic telemetry modem (USBL/ATM), which has limited precision. (4) The robot should ensure safety by providing diving information underwater. (5) The robot should convey information on emergency situations using sound or light. (6) The robot should be operated mainly by the master diver.
Based on this, the diver support concept is defined as follows:
(1) The robot should monitor diving activities and induce safe diving based on the established plan (e.g., movement of the boat, entry and communication confirmation, diving activity monitoring, and return). It should be able to display or alert the status.
(2) The onboard control system provides user-customized information and alerts based on overall information, including the diver status and diving plan.
(3) The robot supports one master diver by providing visual information, such as the depth of divers, dive time, and position of beginner divers.
(4) The robot should be equipped with an autonomous precision control function and self-localization technology.
(5) The robot should be equipped with a diver recognition system that utilizes cameras and sonar that is not affected by floating matter.
(6) The robot should have a compact and lightweight form for the user to easily carry and feature a modular structure for easy maintenance.
To ensure operational scalability and stability, the robot supports a hybrid operational method that can utilize both the remotely operated vehicle (ROV) and autonomous underwater vehicle (AUV) modes. It also facilitates wired/wireless conversion between the modes.
3.2 Robot-Based Diver Support Scenarios
No research has developed goals similar to those of NADIA except for the CADDY project, and the CADDY project does not provide separate diver support scenarios. It focuses on developing functions such as tracking and monitoring, guidance, and recognition, rather than supporting the overall process as a buddy in scuba diving activities. Therefore, in this study, the modes of NADIA on which diver activities will be supported were defined based on existing diver-centered scenarios, as shown in Fig. 4. In emergency situations, the situations of diver position deviation and abnormal diver condition derived from existing diving scenarios were defined. Based on the concept of supporting the master diver, in case of an emergency, NADIA informs the master diver of the situation and performs tasks such as standby and guidance according to the master diver’s instructions. In the case of diver position deviation, the onboard control system and NADIA constantly check the relative position between the divers and robot using USBL/ATM equipment. If a diver exceeds a certain distance, the first alarm and guidance are performed using light or sound. If there is no response, NADIA induces return by approaching the beginner diver according to the master diver’s instructions or remain at its current position while waiting for instructions from the master diver. In the case of abnormal diver condition, NADIA constantly monitors the conditions of the diver using the cameras and sonar installed in the recognition module. If abnormal conditions occur, NADIA notifies the master diver (using light or sound) or the onboard control system (using USBL/ATM) of the situation, induces emergency ascent, and performs towing according to the instructions.
NADIA can provide divers with underwater environmental information in advance. During diving, it shares the condition of divers and the diving situation with the master diver or the onboard control system. It can also tow divers or guide them to a specific location when necessary and can provide diving support, such as the scooter mode.
(1) Diving preparation: In addition to the briefing to share diving information and procedure to check diver safety equipment, NADIA is visually inspected and information such as the diving schedule is entered into NADIA and the control system.
(2) Travel to dive site and pre-dive research NADIA is fixed at the boat, and the boat travels to the dive site. Upon arrival at the dive site, the robot is remotely operated in the ROV mode to collect data, including the turbidity, temperature, and currents of the dive site, in real time. This can be omitted depending on the situation.
(3) Diver entry and descent: The performance of NADIA set to the AUV mode is verified in a simple manner. Descent is then performed according to the instructions and control of the master diver and onboard control system. The robot guides the divers according to the safety speed and speed of beginner divers so that they can reach the target point safely. Here, beginner divers may deviate or fall into abnormal conditions, which should be handled according to the predetermined scenarios.
(4) Underwater tourism: NADIA displays information such as the dive time, depth, diver position, and diver condition on the display screen during underwater tourism. It also supports taking photos and lighting by tracking the designated diver according to the instructions of the master diver and can operate with underwater scooters when necessary. Here, beginner divers may deviate or fall into abnormal conditions, which should be handled according to the predetermined scenarios.
(5) Travel to the tourist site: For travel to the dive site, the same functions as underwater tourism are provided. When necessary, divers are guided and towed to the designated destination. Here, beginner divers may deviate or fall into abnormal conditions, which should be handled according to the predetermined scenarios.
(6) Diver ascent and exit: When divers ascend, NADIA induces compliance with the ascent speed and safety stops for their safe return to the boat. After the exit, the robot is recovered and inspected. During ascent and safety stops, beginner divers may deviate or fall into abnormal conditions, which should be handled according to the predetermined scenarios.
(7) Return to the boat: After recovering NADIA, the robot is fixed at the designated position of the boat for return.
The five modes supported by NADIA are summarized in Table 4. They include the set-up mode for preparation of the robot before entry (e.g., setting), diving information mode for a preliminary survey of the dive site and delivering diving-related information to divers, diving guide mode to guide divers for activities in compliance with safety rules, safety monitoring mode for monitoring and responding to abnormal diver safety conditions, and diving support mode to support additional functions to enjoy diving. In addition, divers can call the robot using the call and release modes during diving. In the call mode, divers can change the robot into the command standby mode using diver equipment. In this instance, the robot can provide diving information to the diver. The release mode returns the robot in the call mode back to its mission.
The detailed operational scenarios for the defined modes were defined as follows; the set-up and diving support modes, which are related to the maintenance and support of the robot, were excluded from the scenario definitions:
(1) Diving info #1: NADIA operates in the ROV mode on the diver support boat. It acquires information, such as the temperature, current, and turbidity of the dive site by depth.
(2) Diving info #2: During diving, useful information, such as the dive time and depth, is provided to the divers through the display device installed in the robot.
(3) Diving info #3: When a diver calls the robot in the call mode, the robot moves to the front of the diver and provides diving information through the display device.
(4) Guide #1: After the entry of all divers, they begin to descend under the control and instruction of the master diver. In this instance, the robot descends at an appropriate speed and induces equalizing at predetermined intervals. During travel, it visually displays diving information for the divers to check the current status. Upon arrival at the destination, it delivers a completion notification and waits for the next instruction.
(5) Guide #2: When travel to another site is required, all divers gather in one place under the control of the master diver. According to the call and instructions of the master diver, the robot travels to the destination and displays current diving information on the screen. When the travel is complete, it switches to the release mode after a completion notification and stands by.
(6) Guide #3: The master diver changes the robot into the towing mode through a call or instruction. The user manipulates the robot as a scooter or travels along the predetermined path using the tow handle. Safety lines can be used when necessary.
(7) Guide #4: After all divers gather at one point, they begin to ascend under the control and instruction of the master diver. In this instance, the robot ascends at an appropriate speed and induces safety stops at predetermined intervals. During travel, it visually displays diving information for the divers to check the current status. Upon arrival at the destination, it delivers a completion notification and waits for the next instruction.
(8) Monitoring #1: It monitors the distance between the divers (distance between the buddies) and conveys alerts using light or sound to maintain a constant level.
(9) Monitoring #2: When a diver deviates from the group, the robot visually informs the situation (e.g., the location of the diver, current situation, and diving information) to the master diver, conveys alerts constantly using light or sound, and waits for the next instruction.
(10) Monitoring #3: When signs of panic are detected from a diver (they can be transmitted from the diver equipment), the robot visually informs the situation (e.g., the location of the diver, current situation, and diving information) to the master diver, conveys alerts constantly using light or sound, and waits for the next instruction.
(11) Monitoring #4: When signs of non-breathing are detected from a diver, the robot visually informs the situation (e.g., the location of the diver, current situation, and diving information) to the master diver, conveys alerts constantly using light or sound, and waits for the next instruction.
3.3 Summary of Required Functions for NADIA
When the functions required for NADIA are summarized, they can be classified into the mechanism, algorithm, communication, and operation of the robot, as shown in Table 5.
Mechanism-related functions are as follows. The size and weight of NADIA must be minimized for its operation by one or two divers, and some parts must be modularized to facilitate maintenance. In implementing five degrees of freedom (surge, sway, heave, pitch, and yaw), a minimum number of thrusters should be used for shape optimization. NADIA requires a tow handle to tow or guide divers, and it should be able to install tow lines for safety. It also requires lights to secure the view of the divers and light emitting diodes (LEDs) to notify the situation. It should be equipped with a display device to show diving information (e.g., dive time) and the current status (e.g., diver position). A recognition module, including two cameras, one sonar sensor, and computing devices, is installed to estimate the status, position, and posture of divers by recognizing them.
Because NADIA supports diver activities in close proximity, control, navigation, and recognition algorithms are essentially required. The robust control technique algorithm should be used for precise control up to the set location. To this end, a navigation algorithm, an accurate self-localization system, must be implemented in the underwater environment. In addition, an inertial measurement unit (IMU), Doppler velocity log (DVL), depth sensor, digital compass system (DCS), and navigation sensor that includes the global positioning system (GPS) are required. For interaction with the divers, the robot accurately tracks their status, position, and posture through a recognition algorithm.
NADIA’s communication system has absolute position tracking systems (e.g., GPS and USBL) in addition to navigation, making it possible to respond to long-term utilization or loss. It also implements the ATM function for wireless communication between the robot and boat or between the divers and boat underwater. A wireless communication function is installed to back up and remotely control the robot’s data on the boat or on land. This communication function is modularized for geometry minimization, thereby increasing its applicability to various applications.
NADIA is a hybrid underwater robot capable of both ROV and AUV modes. The basic mode of NADIA is the AUV mode with no cable because cables may cause safety problems, such as entangling divers, when they support divers underwater. This requires batteries to provide sufficient power during the dive time. In this instance, NADIA supports one master diver to maximize the efficiency of robot operation. For the preliminary survey of the dive site, however, it operates in the ROV mode without communication restrictions because it is necessary to receive camera and sonar images as well as information from various sensors in real time. The change of modes is set in advance by the user using set-up#4 or set-up#5 in the robot support mode setting before the mission begins.
3.4 Concept Design for NADIA
The conceptual design plans of the robot were derived by reflecting the required functions of NADIA summarized above, as shown in Fig. 5 and Table 6. NADIA modularizes hardware by function (e.g., communication and recognition modules) to facilitate maintenance and achieve a compact and lightweight form. It also applies vector arrangement of six thrusters to obtain the maximum degree of freedom with a minimum number of thrusters and to implement surge, sway, heave, pitch, and yaw motions. Diver towing bars are installed at the back of the robot, and a button is installed for a diver to manipulate the robot. In front of it, a holder is reflected to install safety lines. The display device is installed at an upward diagonal position between the safety bars so that divers can see it well when they hold the towing bars and when they travel. The absolute position of the robot is measured using GPS on the water, and the internal computing devices can be remotely accessed using wireless fidelity (wifi) when necessary. A communication module is installed to deliver simple remote control commands using radio frequency (RF) communication. This includes LEDs for alarms.
The recognition module for AI-based estimation of the diver’s position, posture, and status was designed and implemented. It consists of a high-performance embedded AI computing module (NVIDIA Jetson AGX Xavier), two cameras, one sonar sensor, and two lights. It is designed to detect and track divers in real time in various underwater environments. To accurately estimate the diver’s position, posture, and status in the underwater environment, a training dataset similar to the actual environment was constructed. The diver’s movement and specific postures were captured by installing a structure that includes background noise in the water tank environment, and more than 100,000 image data were collected using cameras and sonar. In addition, synthetic data in consideration of various turbidity and lighting conditions as well as additional diver postures were created using Unreal Engine 5-based Holoocean underwater simulation, and they were used for training. To estimate the diver’s position in continuous frames in a stable manner, three-dimensional position information was calculated by applying a stereo camera-based distance estimation technique. The Damo-YOLO model was applied for object detection and bounding box extraction, and the diver’s movement was tracked in real time using a Deep body tracker or BoostTrack as a tracking technique. The estimated diver recognition results were then used as input into the 6DRepNet model to construct a pipeline for predicting the diver’s posture. In addition, air bubbles that occur during breathing were detected to examine the abnormal conditions of the diver, and abnormal conditions were estimated through frequency analysis. A long short-term memory (LSTM)-based anomaly detection model was applied to detect abnormal behavior that occurs when a diver falls into abnormal conditions and classify normal and abnormal states. In situations with low visibility (e.g., high turbidity), an auxiliary position estimation system is constructed using sonar to enable robust estimation. Finally, the algorithms are enhanced by collecting additional data from sea areas to develop AI-based robust diver recognition and estimation algorithms.
Because the robot operates in close proximity to divers, high-precision autonomous control technology is required. A robust control algorithm was applied based on the time delay control (TDC) technique to construct a simple controller that is strong against disturbances (Cho et al, 2023). To this end, a control system was constructed, including a high-performance embedded AI computing module (NVIDIA Jetson AGX Xavier) and embedded support board. It runs control, path management, and mission management algorithms.
USBL/ATM information can be used underwater to estimate the diver’s position, but this method has limitations in terms of precision and signal stability. Therefore, a navigation system was developed separately for the robot to accurately estimate its position based on IMU, DVL, depth sensor, DCS, and GPS. Finally, the robot was set to operate in the hybrid mode and equipped with a battery to support the AUV mode. The battery was designed to be easily detachable for the convenience of charging and maintenance. Lastly, sensors for acquiring information (e.g., water temperature and turbidity) were also installed to investigate the underwater environment.
3.5 NADIA Performance Verification Plan
The proposed functions must be verified according to appropriate procedures after implementation. Various functions of the robot will be implemented in water tanks or in sea areas during development to achieve the target values listed in Table 7. The robot has a weight of 60 kg or less to reduce the weight of the platform geometry and minimize its size. To tow divers through tow bars, NADIA has a traction force of over 150 N using vector arrangement of six thrusters. In addition, NADIA should have a battery capacity that allows its operation for more than two hours at a speed of 0.51 m/s or higher. To verify precision autonomous control performance and diver support, divers are guided according to the safety guide scenario for performance verification with a target success rate of 90% or higher. The safety guide scenario for performance verification should be newly defined according to the implemented functions. The self-localization system achieves a navigation accuracy of 0.5% or less based on GPS. Here, the navigation accuracy is the value obtained by dividing the navigation error compared to GPS at the final location by the total distance traveled. The recognition module achieves a diver recognition accuracy of 80% or higher, diver position estimation accuracy of 0.2 m or less, and diver posture estimation accuracy of 15 degrees or less to verify diver position and posture estimation accuracy. It achieves a consciousness detection accuracy of 90% or higher to verify diver status estimation performance.
4. Conclusions
With increased engagement in marine leisure activities and the expansion of related industries, concerns have been raised over safety. In particular, most accidents in scuba diving are caused by individual negligence, and more than 90% of them lead to deaths. This means that scuba diving can be enjoyed safely if established safety procedures are well followed. To ensure the safety of divers, projects have been performed to construct underwater diver safety support robots and integrated operating systems. In these projects, autonomous control technology that enables support in close proximity to divers; AI-based technology to recognize the diver’s position, posture, and status; and onboard and onshore control systems have been developed using lightweight robot platforms with high degrees of freedom.
Prior to the development of NADIA, this study aimed to derive robot functions and conceptual design plans for mission performance. Highly reliable data were collected and summarized through the literature related to scuba diving and the leisure industry, consultations with professional divers, and direct experience. The concept of how to operate the robot was also identified through diver scenario analysis. Robot-based diver support scenarios were constructed through this concept. Based on this, the functions of the robot were defined and conceptual design plans were derived. The conceptual design plans will be used in specific robot design and production to improve the applicability of the robot and facilitate successful task performance.
PDF Links
PubReader
ePub Link
Full text via DOI
Download Citation
Print
