Development of an Unmanned Aerial System for Maritime Environmental Observation

We developed an octocopter-based unmanned aerial system (UAS) to monitor remote environments. The system contains various environmental monitoring sensors for measuring wind speed/direction, temperature, relative humidity, atmospheric pressure, fine dust, and multispectral and RGB data. The UAS consists of an unmanned aerial vehicle (UAV), a ground control station, and a server. We studied its electrical, mechanical, and software (SW) configurations. Specifically, we developed hardware (HW) and SW to control the yaw and gimbal pitch directions of the UAV conveying multispectral and RGB cameras. To prevent the obtained solar reflectance from affecting the multispectral and RGB data and to improve the quality of the obtained data, we maneuvered the yaw so that it would always deviate 135° from the solar azimuth angle, with the sun at its rear. Managing the yaw direction involves controlling the UAV based on a micro air vehicle link message-based robot operating system (ROS). To safely test the UAS performance before the maiden flight in an ocean area, we first evaluated PX4 and ROS-based SW through indoor software-in-the-loop simulation (SILS) via the Gazebo tool. Subsequently, we used an F450 UAV and an actual maritime UAV to sequentially perform pre-flight and flight experiments. This paper explains the system operation scenario, UAS component, and simulation and experimental results. The results reveal that the average yaw angle error during the mission, $\bar {\left |{ \theta _{y,des}-\theta _{y} }\right |}$ , is approximately 8°, and the average pitch angle of the gimbal during the mission, $\left |{ 50^{\circ }-\left |{ \bar {\theta _{c}} }\right | }\right |$ , is less than 5°.


I. INTRODUCTION
Satellites [1], [2], crewed aircraft [3], [4], and vessels [5] generally capture ocean color, which is a critical indicator of ocean conditions. However, obtaining images of desired locations at desired times may be difficult when using conventional remote sensing based on satellites, airplanes, and vessels owing to the inaccessibility of certain regions and inclement weather. Consequently, it may be challenging to capture anomalies, such as a red tide in the ocean [6], which requires immediate or time-based observations. Understanding phenomena such as the resuspension-precipitation process of suspended solids and the spread of low-salinity water originating in coastal areas is also difficult.
To measure the remote-sensing reflectance (Rrs) of seawater and estimate its constituents (such as phytoplankton The associate editor coordinating the review of this manuscript and approving it for publication was Stefania Bonafoni . chlorophyll concentration and marine inherent optical properties), researchers have widely used AWRs, either by transporting portable AWRs on a ship or by installing them at fixed observation points. Herein, we define Rrs as the waterleaving (Lw) radiance, normalized by the downwelling irradiance (Ed), and we measure both Lw and Ed above the sea surface. However, the use of the aforementioned method to measure the remote reflectance in various seawater environments multiple times is expensive. Furthermore, measuring it at the desired frequency in the desired sea area at the expected time is impossible during inclement weather. In addition, it is advantageous to monitor anomalies with continuous observation data for stationary observations. However, data from various sea areas cannot be obtained because the observation data are limited to stationary data.
Nevertheless, we can instantly capture various marine phenomena by using an unmanned aerial system (UAS), including vertical takeoff and landing (VTOL)-type unmanned aerial vehicles (UAVs). UAVs can move and hover over a location and obtain high-resolution data of the desired form in real-time. Specifically, using a micro-sized multi-copter UAV can measure real-time data for distances up to tens of kilometers, allowing sufficient monitoring of anomalies that occur near coastal areas or moving ships in the middle of a vast sea [7]- [9]. Because of the above-mentioned benefits, UAS can be used for marine fauna surveys in place of crewed aircraft [10]- [18].

II. RELATED WORK
For robust automatic UAS operation in a maritime environment subject to frequent gusts and strong winds, where a single mistake could result in catastrophic system destruction (due to ocean salinity) or loss into the sea, cuttingedge system development and data analysis techniques are essential to: -Develop an error-free and fully automatic UAS.
-Maneuver the unmanned system to maintain the desired attitude from any possible internal or external erroneous source. -Accomplish the environmental data acquisition mission without losing or damaging the incoming data. -Induce processed data that are as accurate as the direct measurement data. We performed a literature survey on the following four general topics related to the UAS for maritime environmental observations: -Automatic system development.
-Attitude-control algorithm development, considering the solar location. -Data measurement and analysis. -Latest technology.

A. AUTOMATIC UAS DEVELOPMENT
The fully automatic mission operation of the UAS requires system development, which can be divided into three main categories: software (SW), hardware (HW), and system integration.
-  [21]. -System integration involves a vision-guided UAV for the automatic landing mission using an in-house developed FC and avionics hub board, gimbal, camera, laser rangefinder, MC, and wireless module [22].

B. UAV ATTITUDE-CONTROL ALGORITHM
To obtain environmental data from the desired location and position, the attitude of the UAV must be maneuvered as quickly and precisely as possible.
-Haus et al. presented and applied the centroid vectoring technique-based attitude-control method -which uses the concept of a robot dynamic model and ubiquitous Jacobian matrix -to both aerial and underwater vehicles [23]. -Segales et al. combined a stability augmentation system (SAS) and a control augmentation system (CAS), which are PID controller-based, to maneuver the attitude of the tricopter UAV [24].

C. SOLAR-TRACKING ALGORITHM
Specifically, to achieve efficient optical and multispectral data of the sea surface, the sensor direction must be controlled by referring to the solar location so that the amount of sunlight on the obtained optical and multispectral data is minimized. For this purpose, a solar-tracking algorithm should be considered along with the UAV attitude-control algorithm.
-Calculating the solar location and building a low-cost solar-tracking mechanism involves solar-tracking mathematical models, along with active and chronological algorithms 25]. -Developing a low-cost closed-loop solar-tracking system involves using the solar-positioning algorithm to estimate solar azimuth and elevation angles, which is eventually used to position the solar panel toward the sun [26]. -Jamroen et al. developed a pseudo-azimuthal system model for calculating the solar altitude and azimuth angles for a solar tracking system [27]. SW-based small UAV equipped with a mechanism that provides easily interchangeable measurement sensors and tested it in a lake [35].
E. MARITIME ENVIRONMENTAL DATA ANALYSIS Unmanned system development been researched for safe, rapid, and low-cost (but robust) data collection; active maritime data analysis research is also in progress to overcome the limitations of remote data.
-The characteristics of a coastal environment based on both coastline change and marine litter concentration in shoreline zones have been determined using a UAS based on the geoinformation produced using the UAS-structure from motion (UAV-SfM) pipeline method [36]. -The Rrs of seawater is the most fundamental optical property for remotely estimating components in seawater, including chlorophyll concentration, suspended solid concentration, and colored dissolved organic matter. To estimate seawater components and phenomena from Rrs, it is necessary to develop algorithms using the accumulated seawater reflectivity and constituent data under various seawater and atmospheric conditions, as discussed in a previous work [37]. -A method incorporating both superpixel segmentation and a convolutional neural network (CNN) was proposed for extracting accurate Ulva prolifera (a green seaweed species) region from superpixel UAV image classification [38]. -The UAS has been used to detect marine litter, such as floating plastics [39], [40]. -The methodology and practical information of UAV-based photography for oceanographers to lower barriers to entry reduced the cost of developing robust unmanned systems and obtained high-quality and map-ready data [41]. -A protocol to enhance the reliability of environmental data acquisition using UAVs is presented by considering the morphology, environmental conditions, and flight parameters 42]. -For satellite image classification, high-resolution images observed by a UAV with a resolution below 5 cm were used for support vector machine data training [43].

F. LITERATURE REVIEW
Several studies have reviewed various applications of UASs for environmental monitoring [44]- [46]. Specifically, these reviews dealt with environmental data measurement; processing; assessment; and limitations of the land, air, and sea; as well as maritime archeology, pollutants, physical and biological oceanography, wildlife, behavioral ecology, habitat assessments, coastal geomorphology, etc.

G. MAIN CONTRIBUTION
The novelty of our study is the development and verification of real-time data broadcasting UAS to retrieve threedimensional (3D) maritime environmental observation data by reducing the reflected light with automatic UAV yaw and gimbal pitch angle control.
In this study, we primarily focused on developing an environmentally monitored UAS. Specifically, we designed the UAV to automatically control both the pitch and yaw directions of multispectral and RGB cameras, based on the solar incident angle, to calculate the reflected light shown on the captured image and eliminate sunglint. Simultaneously, to remotely estimate the seawater constituents by calculating the reflected light from the surface of the sea, we needed to install a multispectral sensor with a gimbal to obtain two-directional data (obtained at an inclination of 40 • from the top and bottom) [47]. Furthermore, to construct a 3D environmental map, we developed a UAV to obtain additional data, including wind speed/direction, temperature, RH, atmospheric pressure, fine dust, multispectral, and RGB data. Finally, we synchronized the data in real-time using a server located in a remote area through long-term evolution (LTE) and Wi-Fi communication.
The rest of this paper is organized as follows: -Section III describes the system design of the UAS operation. -Sections IV, V, and VI present overviews of the electrical system, mechanical system, and SW configurations of the UAV, ground control station (GCS), and server. -Section VII details the indoor and outdoor flight experiments with the corresponding flight logs and their analyses. -Section VIII presents 3D environmental maps generated from data obtained from outdoor flight experiments. -Section IX presents our conclusions and directions for future work.

III. SYSTEM DESIGN A. SYSTEM REQUIREMENTS
For non-professional use in maritime environmental observation, we needed to develop a UAS that incorporates a low-altitude VTOL-capable UAV that: -Can be produced and repaired at a relatively low cost.

B. OPERATION OVERVIEW
For automatic 3D data acquisition using a UAV, Figs. 1 and 2 provide an overview of the UAS operation and an operational scenario. Several mandatory restrictions must be considered.
-The UAV starts its operations from a stationary ship (moving slowly over time) or on the ground. -The UAV is at an altitude of 30 m and backs the sun at approximately 135 • , and the camera headed at 40 • from the bottom rotates to a heading angle of 40 • from the top over a pre-assigned period [48]. The attitude of the UAV -specifically, the yaw angle -should be calculated based on the solar position from the global positioning system (GPS) location and current time. -After the mission operation, the UAV should return to the launch location, which may be different from the original position. -All mission operations must be controlled using a user-friendly graphical user interface (GUI) to enable use by non-experts. -The UAV should not only perform basic operations, including circular flight, ascending, descending, and flying to a specific location and hovering during mission operations but also all functions provided by the QGroundControl (QGC) software [49]. In Fig. 2, the yellow lines represent the path of the UAV, and the green lines represent the path of the image center of both the multispectral camera and RGB camera. The red marks along the yellow path represent the waypoints of the UAV.

C. OPERATION MANUAL
For non-expert use, we minimized and fully automated the UAS operation processes, as presented in the UAS operation manual in Table 1. In total, the UAS required seven steps, including connecting the operator's smartphone LTE communication with the GCS to disarm the UAV motors after mission completion.

D. OPERATION STEPS
To commence the UAS operation, we placed the UAV on the ground, as shown in Fig. 3(a), and then installed two tripods, one for Here+ real-time kinematic (RTK) GPS and another Step-by-step procedure, from preparing UAS to retrieving the UAV.
for Ubiquity Nanostation NSM5, as shown in Fig. 3(b). For safety purposes, we placed the UAV at a sufficient distance from the GCS operator. As soon as we performed a thorough system check, the operator started the mission and we removed the UAV, as shown in Fig. 3(b). During the mission, the operator monitored incoming data from both the UAV and sensors for emergencies, as shown in Figs. 3(c) -(e).      Table 2 lists the UAV and sensor data necessary for systematic data management. Although a slight time difference occurs between the data measured by each sensor, the data transmission rate was sufficiently low to prevent data transmission loss.

H. MATHEMATICAL APPROACH
Because the UAV heading angle was 135 • from the sun and the gimbal attached to the end of the heading arm was tilted 40 • from the bottom, we preplanned the route of the UAV so that the field of view (FOV) of the camera attached to the gimbal covered the region of interest (ROI), as shown in Fig. 6. Here, the positive x-axis and y-axis represent the north and west directions, respectively.
For simplicity, we assumed that the center of the FOV (P f ) should go over the polygon corners (S i ), which were selected VOLUME 9, 2021  points on the ROI; i.e., the P f should be located inside the polygon drawn by the selected points, S i .
Based on this, we can calculate the desired waypoints (W i ) for use as a list of input waypoints in the QGC SW, as follows: where P u,i is the UAV position, i.e., [(x, y, z) u and i = 1, 2, . . . , n], which is in agreement with the desired waypoints (W i ); R y (θ h ), is the rotation matrix of θ h around the y-axis; θ h is the heading angle of the UAV (θ h = θ a ± 135 • ·π 180 • ), which is the same as the yaw angle of the UAV (θ y ); θ a is the solar azimuth angle; V fc is the vector pointing to the center of the camera frame from the gimbal ( V fc = l a cos θ c ); l a is the UAV altitude; R z (θ c ) is the rotation matrix of θ c around the z-axis; θ c is the angle measured from the XY plane to V fc (θ c = 40 • ·π 180 • ), which is the same as the pitch angle of the gimbal; V h is the vector pointing to the end of the heading arm from the center of the UAV (|V h | = l r ); l r is the length of the UAV arm; and P c represents the camera position [x, y, zc].

IV. UNMANNED AERIAL VEHICLE A. SYSTEM REQUIREMENTS
The UAV needed to transport various environmental sensors; specifically, a gimbal carrying a multispectral camera and an RGB camera, which were installed such that they would face the same direction. In addition, the UAV needed to refer to the predefined waypoints as it flowed over the mission area.
The main functions of the UAV can be summarized as follows: -The flight operation commands, such as takeoff, waypoint flight, attitude control, return to launch (RTL), and landing. -The yaw angle that the UAV had to head to and maintain a 135 • bearing from the solar azimuth angle, with the sun at its rear. -The pitch angle of the gimbal was maintained to allow both a multispectral camera and an RGB camera inclined at an angle of 40 • from the exact vertical, which periodically changes its bearing to 40 • from the top to 40 • from the bottom, as shown in Fig. 7. -Carry environmental sensors to collect data and send them to the GCS.  Fig. 8 shows that the UAV carries battery packs with a normal voltage of 24 V. We used several battery eliminator circuit (BEC) modules for the buck converter function to convert 24 V to either 12 V or 5 V. However, we used a power over ethernet (PoE) injector for the boost converter function to convert 12 V to 24 V.

C. MECHANICAL SYSTEM CONFIGURATION
The mechanical configuration of the octocopter multi-copter type was based on an environmental monitoring UAV, because of its stable flight performance, to maximize noisefree sensor measurements. Figs. 9 and 10 show the detailed UAV appearance with accompanying sensors and Table 3 lists its specifications. Tables 4 -6 list the specifications of the battery pack. We used battery packs #1 and #2 for the UAV thrust and #3 for the sensors.

D. SOFTWARE CONFIGURATION
To maneuver the directional angle of both the multispectral camera and RGB camera mounted on the gimbal, we controlled the yaw angle of the UAV and pitch angle of the gimbal based on the ROS, which was installed on    drivers, for communication among various autopilots using the MAVLink communication protocol. Alternatively, we could use the MAVROS to set the user datagram protocol (UDP) MAVLink bridge between the UAV FC and GCS.
Specifically, we used the desired yaw angle of the UAV, calculated using Pysolar Python libraries [50], to manipulate the bearing of the UAV by overlapping the yaw angle of the waypoint_node, as shown in Fig. 11. We used the Pysolar Python libraries to calculate the azimuth angle: Here, the angle of the Pysolar vector (θ p ) was measured from the east (Fig. 12). As seen from Table 7, the UAV used Python-based SW running on the Ubuntu OS.

E. COST ANALYSIS
The required cost to develop the UAV is given from the bill of materials (BOM), listed in Table 8.

B. ELECTRICAL SYSTEM CONFIGURATION
We installed four 14-in monitors for operator convenience. For stable Wi-Fi communication within a straight-line distance of 15 km, we installed a Ubiquity Nanostation NSM5 to provide a robust Wi-Fi communication connection between the UAV and GCS. As presented in Table 9, we installed a 15 Ah power bank that simultaneously supplied 12 V and 220 V power to achieve a long operation time. We installed a CARPRIE switch panel to serve as the power switch of the monitors and Kimpok AV receivers. The power switches of the desktop PC and power bank were installed in a separate area to avoid power cuts because of possible misinterpretation of directions by a non-professional operator. Fig. 13 provides a detailed description of the configuration of the remaining electrical system.

C. MECHANICAL SYSTEM CONFIGURATION
We designed the GCS to be compact, light, and sufficiently portable for a single user, as shown in Fig. 14 and 15. We equipped the exterior with a waterproof function for extreme environments and long indoor and outdoor storage. We designed the GCS interior so that it would be easily attachable and detachable for frequent performance upgrades and maintenance. Table 9 and 10 provides detailed specifications of the GCS mechanical system.

D. SOFTWARE CONFIGURATION
The desktop PC ran Python, QGC, and Qt, as listed in Table 11. We developed a user-intuitive GUI for UAV maneuver and sensor measurement data analysis, as shown in Fig. 16.  Table 12 presents the BOM to show the cost required to develop the GCS.

VI. SERVER A. SYSTEM REQUIREMENTS
A server can be installed anywhere on the Internet. In this study, we placed our server in a university laboratory with a network switch. Because of the university's Internet firewall policy, we needed a network switch to identify the connected server and forward data to the port connected to the server with an assigned media access control (MAC) address. The university information security team constantly monitored the MAC address to prevent hacking attacks. External users who did not have access to the university Internet had to use the assigned MAC address with a secure password to connect to the server and retrieve the sensor measurement dataset.

B. ELECTRICAL SYSTEM CONFIGURATION
The network switch connection was the only special connection, as shown in Fig. 17. A customized server PC was wired through an Ethernet line with a network switch.

C. MECHANICAL SYSTEM CONFIGURATION
For robust server management, we installed a customized server PC with high computing power in the laboratory at Chosun University, as shown in Fig. 18. Table 13 lists the specifications.

D. SOFTWARE CONFIGURATION
In addition to Python, no specific SW was required for the server, as shown in Table 14.  Table 15 presents the BOM to show the costs required to develop the server.

VII. EXPERIMENT
This section presents the operational scenario and corresponding flight experiment with flight logs and their analysis. The experiment comprised four steps, ranging from an  indoor experiment based on a software-in-the-loop simulation (SILS) to an outdoor experiment to convey all the necessary sensors to the mission area, as presented in Table 16. The level of difficulty corresponds to the number of steps.
To compare the flight logs, we applied a similar squareshaped trajectory pattern for each experimental step. We also cloned the entire system for the same version of SW into each FC and MC used for each experimental step. We used the same SW GUI for the indoor and outdoor experiments, except for Gazebo (which we used only for the indoor experiment), as shown in Fig. 19.
In Tables 17 to 20, θ a is the real-time solar azimuth angle, calculated by referring to Hoffmann's ''SunCalc'' [51], θ y,des is the desired yaw angle of the UAV, θ y is the yaw angle of the UAV, and θ c is the pitch angle of the gimbal. In addition, we present the exact experiment date, time, and location because the data are used to calculate the real-time solar azimuth angle θ a . Specifically, we need to pay attention to using the exact latitude and longitude GPS data by considering the offset between the real location and the location based on the induced GPS data. If the offset is too large, it will affect the home position and can lead the UAV to land at an undesired location after the mission.

A. SILS EXPERIMENT (SIMULATED UAV)
We performed an indoor experiment based on SILS. In the Gazebo SW environment, we used a simulated UAV to determine the proper yawing control functionality based on the solar azimuth angle before the actual UAV flight, as shown in Fig. 20. Table 17 details the experimental data.

B. PRE-FLIGHT EXPERIMENT (F450 UAV)
We performed an outdoor pre-flight experiment on an F450 UAV carrying only a Jetson TX2, assembled with an  Orbitty Carrier board but without a sensor to ensure the proper yawing control functionality based on the solar azimuth angle before actual UAV flight, as shown in Fig. 21. Table 18 provides detailed information on the experiments.

C. PRE-FLIGHT EXPERIMENT (MARITIME UAV)
We performed an outdoor pre-flight experiment on a maritime UAV carrying all sensors to ensure proper functionality of the yawing control based on the solar azimuth angle, the pitching control of the gimbal, and the functionality of the sub-components before the actual UAV flight at the mission area (see Fig. 22). Table 19 provides detailed information on the experiment.

D. REAL FLIGHT EXPERIMENT (MARITIME UAV)
We performed our final outdoor experiment with a maritime UAV conveying all the sensors to evaluate the functionality of the UAV, as shown in Fig. 23. Table 20 provides detailed information on the experiment.

VIII. EXPERIMENT RESULTS
Based on the results of a real flight experiment with a maritime UAV, we verified the following, as shown in Fig. 24.
-Manipulation of the yaw angle of the UAV and pitch angle of the gimbal. -The sensor data acquisition. In Fig. 24(a), red represents the heading direction of the UAV, and blue represents the heading direction of the gimbal. It should be noted that both the heading directions of the UAV and the gimbal were well maintained in the pre-assigned direction.
In Fig. 24(b), red indicates the magnitude and directional angle of the wind. As expected, the wind magnitude fluctuates at takeoff and landing because turbulence occurs when the UAV is located near the ground and the generated turbulence disturbs the data measurement of the wind sensor.
From Figs. 24(c) to (g), yellow, green, and blue indicate sensor data including temperature, RH, pressure, PM2.5, and PM10, respectively, depending on the magnitude. If the sensor data magnitude increases, the color becomes yellow, and vice versa.
In Fig. 24(h), purple, orange, yellow, and red represent the UAV trajectory, sun, solar azimuth, and home location, respectively. In addition, white boxes indicate the directions of the north, south, east, and west. Multiple images from the multispectral and RGB cameras were stitched on the ground map. Although not depicted in Fig. 24(h), multiple multispectral and RGB images of the sky were subsequently used to monitor the light conditions at the moment the images were captured.
In Fig. 24(h), we can see that the UAV takeoffs at the home position (red empty circle) and flies along with the predefined waypoints (purple line), while the heading direction of the UAV is 135 • away from the solar azimuth (orange circle with white boundary) and finally lands at the preplanned home position again. From the stitched images on the ground   in Fig. 24(a), we can conclude that the heading directions of both the UAV and gimbal worked as expected.
We tested a short flight time of approximately 2 min because we only wanted to focus on determining the right attitude control of the UAV and gimbal.
Note that there is no relationship between the average yaw angle error, θ y,des −θ y , and the experiment times listed in . This is because the desired yaw angle required by the UAV was calculated based on the Pysolar library [50], which was primarily developed to model solar radiation on Earth. The Pysolar open-source library uses the user's current location, date, and time to calculate solar altitude, azimuth, clear-sky radiation, etc. We considered the weather variable to calculate the solar azimuth angle (the main focus of this study) to control the UAV to back the sun at 135 • and thus minimize sunglint on the captured images.

IX. CONCLUSION
In this study, we developed a UAS composed of a UAV, GCS, and a server to measure wind speed and direction, temperature, RH, atmospheric pressure, fine dust, and multispectral and RGB data of the maritime environment. It can control the yaw and gimbal's pitch direction based on the solar incident angle, calculate the amount of reflected light on the captured image, eliminate sun, and, ultimately, estimate seawater constituents.
The main application of the proposed maritime monitoring UAV is to observe ocean color (for example, a red tide), which requires immediate and time-based observations. The MAVROS package controlled the yaw angle of the UAV and the pitch angle of the gimbal, which enabled MAVLink communication among computers running ROS.
Our experimental results indicate that the yaw angle of the maritime UAV maneuvered to 135 • from the solar azimuth angle, with the sun at the rear. The average yaw angle error during the mission,¯ θ y,des − θ y , was approximately 8 • . Similarly, the average pitch angle of the gimbal during the mission, 50 • − θ c , was less than 5 • . With respect to the environmental monitoring sensors, we analyzed the wind speed obtained with direction, temperature, RH, atmospheric pressure, fine dust, and multispectral and RGB data by visualizing 3D graphs along with the UAV trajectory for intuitive analysis.
Although some previous studies yielded relevant results, these have limitations because the experiments were conducted near a reservoir, rather than over an ocean, due to the short flight time that arises from the incapacity of the maritime UAV. In this study, because of concerns about damage arising from a crash into the ocean, we postponed experiments over the ocean area until we could verify the robustness of the system through multiple flight experiments. The short flight time is mostly due to the non-optimized, bulky, and unnecessarily heavy HW and the low energy density of the Li-Po battery pack compared to that of the Li-Ion battery pack. Although we attempted to increase the energy density by hybridizing a 22 Ah Li-Po battery pack, a 17.5 Ah Li-Ion battery pack, and a 3.5 Ah Li-Ion battery pack in this manuscript, we further needed to increase the capacity of the Li-Ion battery pack.
Unfortunately, we did not test the flight time endurance in this project because of fall and breakage risks. To safely test the flight time, our research group is currently designing a three-axis jig-stand system with a three-axis load cell. With this system, we can safely perform an indoor flight endurance test before the actual outdoor flight time endurance system.
The maritime UAV that we developed has an autonomy level between 2 and 3; this is because a pilot needs to activate the UAV with predefined waypoints and activate sensors with predefined measuring time, the UAV automatically flies under limited conditions, etc.
In the future, we expect the maritime UAV to be redesigned to weigh less than 8 kg (with frame, sensors, and batteries); approximately 60% of the weight of the current UAV prototype. This is achieved by replacing the following: -Octocopter frame with a coaxial quadrotor frame, with smaller motors and propellers. -Three-axis gimbal with a one-axis gimbal.
-Electric speed controllers (ESCs) with ESCs having a lower maximum permissible current, 3D printed zig (for the internal mounting of sensors) made of polylactic acid material with a zig made of light carbon material. -Power cables with higher American wire gauge cables. -USB hubs with a self-made multiple-layered printed circuit board with signal converters for RS-485, Ethernet, USB 2.0, and USB 3.0. The main purpose of the redesign is to achieve greater stability, maneuverability, and endurance. The flight time of the redesigned maritime UAV is expected to exceed 40 min.
The reason for postponing the flight experiment over the sea area was predominantly due to the unknown flight endurance capability. Notice that the BOM information in Tables 8, 12, and 15 demonstrates an excessively high system cost and we could not ensure the robustness of the UAS for flying over the sea, where a single minor mistake could result in catastrophic mission failure (for example, the loss of the UAV in the seawater). Therefore, we postponed flight tests in proximity to the sea for future work.
In the future, we will mainly focus on reducing the overall UAV weight to less than 8 kg and increasing the flight time to over 40 min. We will also perform thorough robustness tests by focusing mainly on the possibilities and limitations of flight endurance and wind resistance capabilities. We expect to perform dozens of flight tests before initiating the flight experiment over the sea area.
Our primary purpose in publishing this manuscript is to allow remote-sensing researchers, particularly those studying ocean color, and to easily follow up on how to develop an unmanned system-based data acquisition tool. With this in mind, we dedicated ourselves to introducing unmanned system technologies as clearly and as detailed as possible so that non-experience remote-sensing researchers can easily understand it and follow it for their remote-sensing research purposes.
With the systematic information presented in this manuscript, future researchers will be able to develop various types of unmanned systems for maritime data acquisition. We hope that researchers investigating ocean color can reduce their research expenses from direct seawater collection, as well as reduce unexpected human casualties from boarding research vessels in dangerous weather, despite the limitation to coastal seawater.
In a transitional period when unmanned systems are being expanded to various fields of study, particularly remote sensing for the maritime environment, we hope that this manuscript serves as a fundamental reference paper for the development of unmanned remote-sensing systems.