• Ei tuloksia

Developing a Framework for Trustworthy Autonomous Maritime Systems

3. AUTONOMOUS SHIP

Figure 4 and Figure 5 show real-time monitoring and failure rate after new application.

Figure 4: Realtime Monitoring in Facility Layout

Figure 5: Application of Failure Rate

A set of simulation engineering-based technologies, from the problem diagnosis stage to the development stage and the operational stage, are required to improve productivity, including hardware and software for all the process stages of shipbuilding.

In this project, we defined the field of simulation engineering-based virtualization production platform technology where we intend to create a foundation to eventually build a digital twin yard through the development of shipbuilding CPPS and to sustain the innovation system of the shipbuilding production system.

③ Efficiency

It is expected that the operating costs will be reduced through the use of system-wide energy efficiency technologies such as unmanned ships, eco-friendly fuel, increased fuel efficiency, real-time route optimization, and will also be more efficient in marireal-time affairs.

Autonomous vessels are not limited to fully autonomous vessels. We can set such vessels as remotely operated local vessel → remote controlled unmanned coastal vessel → autonomous unmanned ocean-going ship. Appropriate technologies are needed to verify each phase.

Thus, from level 3 of autonomous driving, autonomous vessels can be an extension of land-based services such as remote control, remote diagnostics, and remote maintenance. This is a step in which ICT provides efficiency and reliability in terms of commercialization. Furthermore, support services will be provided to handle various maritime services on land by reducing or unmanned personnel, and infrastructure that supports safe operation of ships such as buoy and lighthouse will be intelligent and converted.

In addition, connectivity of data is necessary to achieve level 3 of autonomous ship. E-navigation (S-100) can improve the connectivity of data and the communication environment between ship and land (diffusion of VSAT and speed of technology evolution), as well as standardization of in-vessel communication (NMEA, Modbus, ISO DIS19848, etc.).

3.2. Unmanned application of autonomous vehicle technology using artificial intelligence It will be possible to utilize autonomous technology for autonomous vessels by employing autonomous vehicle application technology. Level 4 or above autonomous cars can be defined as monitoring the driving environment and the fallback function for system errors that allows the system to respond autonomously without driver intervention.

According to the NHTSA autonomous vehicle introduction scenario, the early introduction of an autonomous vehicle into the market, even if it is not perfect, will help reduce fatalities in the long term. It is estimated that the introduction of a self-driving car, which is 10% safe compared with average human driving in 2020, will reduce the number of deaths from traffic accidents by approximately 520,000 compared with the introduction of a fully autonomous vehicle in the market in 2040. Therefore, the above implications of autonomous vehicle should be considered in reviewing the safety conditions of autonomous ships.

Sensor fusion technology for monitoring the driving environment is being developed from the rule-based approach level to deep learning-based level, and it is approaching or exceeding the human perception level. Artificial intelligence can improve the autonomous level through iterative learning on various driving environments; thus, it is essential to acquire driving data for learning.

According to the book "Deep Learning" by Ian Goodfellow, Yoshua Bengio, Aaron Courville (2016), there are approximately 5,000 learning data sets per category that yield acceptable performance, and at least one million learning examples are required to match or exceed human performance. In order to accurately achieve learning in artificial intelligence, it is necessary to tag the correct answer for each learning example, which is time consuming.

In addition to object recognition, autonomous navigation artificial intelligence technology requires a variety of context awareness, collision assessment, unexpected response, driving range extraction, and even end-to-end. Currently, artificial intelligence technology is mainly used in the field of cognition, and studies employing a deep learning model for object searching are increasingly conducted in cognitive fields using image sensors.

Artificial intelligence technology was first employed in autonomous driving object recognition, and in decision making based on a complex road situation. However, it is necessary to investigate different application methods for maritime situations in which autonomous ship are applied differently from autonomous vehicles.

Figure 6 shows the structure of the computing module for artificial intelligence of autonomous vehicle. The module is divided into a platform part composed of HW and OS, a deep learning part composed of a deep learning framework and model, and an application part composed of recognition, judgment, and control.

Figure 6: Structure of Computing Module for Autonomous Vehicle 4. CASE STUDY OF AUTONOMOUS SHIP

4.1. Unmanned ship development trend in Korea

Currently, the development of unmanned ships has continued to increase around the world.

South Korea is still in the initial stage of research in large unmanned commercial vessels, and technology for unmanned small ships is being developed mainly in laboratories and military facilities.

Figure 7: Examples of Unmanned Ship Development in Korea

The U-Tracer in Figure 7 is an outboard type unmanned water reactor developed by the Agency for Defense Development in 2015 and equipped with underwater acoustic target technology that can autonomously track obstacles and targets in water. Aragon is a multipurpose intelligent unmanned ship developed by a private research institute in 2015; it can operate up to 20 kilometers from the ground control system and was developed for marine research and surveillance purposes. Haigeum is an unmanned ship with maritime weapons system developed in 2015 for military operations such as surveillance reconnaissance, mine exploration, and other missions in coastal waters. M-Searcher is an unmanned water reactor developed by the Agency

for Defense Development in 2019 to carry out various tasks such as surveillance and underwater search and rescue.

In the future, Korea plans to conduct research on unmanned water vessels for ports/base areas. The technology presented in this paper was used on a platform mounted on the M-Searcher.

4.2. USV platform configuration

Figure 8 shows the composition of the USV platform used in this study. The USV platform is divided into the main control systems, engine control systems, waterjet control systems, and power control systems.

Figure 8: USV Platform Configuration

The main control system receives control commands from the operating system and transmits them to the engine, water jet, and power control systems. It interprets the commands received through Ethernet communication and transmits the corresponding operation to the control device to enable automatic control. Then, the status data of the system received from each device is collected and transmitted at 10 Hz cycle. The engine control unit is responsible for starting, RPM control, and switching functions.

The errors that may occur in this system have not been evaluated as it is still in the development stage. Therefore, the system was configured so that mode switching can be performed for manned control under such condition. In manned mode, the system was designed to ignore commands from the computer, even if such commands were transmitted to engine control.

Safety has been improved in the event of an error in the USV platform by allowing human control.

In remote starting, the system was designed to start only when there is no alarm from the ECU status information transmitted through J1983 protocol. The engine status information from the ECU is transmitted to the main control system.

The waterjet control system was also developed to enable the same stable operation as the engine by introducing the manned/unmanned control switchover function. The waterjet controls the nozzles responsible for steering the ship via hydraulic pressure and buckets, which is responsible for forward and backward movements.

In unmanned mode, the system was equipped with a mode function, which is the same as that of the engine, so that it cannot be operated even if an operator arbitrarily manipulates the steering control device. Information (such as the waterjet status and alarms) is transmitted to the waterjet control system at 10 Hz cycle through the I/O system, and the corresponding data is sent to the land control center via the main control system.

The power control unit distributes the power produced by the generator to the platform within the USV. The power supply should be controlled for stable operation of the system in the event of a malfunction because the USV platform is equipped with various devices. The power supply control system monitors the over-current and over-voltage of the system in the ship, automatically shuts down in the event of abnormalities, and sends monitoring data and alarms to the control center in real time.

4.3. Test environment

A test bed with the same engine and water jet was built for testing the USV platform. The engine and water jet used on the platform are the VGT450 from Marin Diesel, Sweden and the AJ285 from Alamari, Finland, as shown in Figure 9.

Figure 9: Propulsion equipment (a) Engine and (b) Water-jet Table 2 Specification of Platform Equipment

Equipment Specification

Engine (VGT450)

- Max. Power: 450 HP - Max. RPM: 3,600 RPM - Weight 510 kg (Dry) Water-Jet

(AJ285)

- Max. Power: 500 HP - Max. Shaft RPM: 3,700 RPM - Max. Impeller Dia.: 288 mm - Weight: 148 kg

Power (Self-Developed)

- Max. Capacity: approx. 13 kW - Channel: 6 (AC 220 V or DC 28 V) - Weight: 30 kg

The fabricated test bed was fixed on the test bed using a frame suitable for testing in the tank. Figure 10 shows the design of the frame to fix the test bed. Figure 11 shows a frame with the test bed; the test bed was fixed to a water tank, and long continuous operation tests were conducted in the basin.

Figure 10: Test Bed for the USV Platform and Frame for the Test Bed

Figure 11: Basin Test of the USV Platform

4.4. Test result

The remote control function of the system was tested using the test bed of the USV platform installed in the basin. Tests were conducted to verify the remote power control, remote engine control, and remote waterjet control functions of the platform.

Figure 12: Captured Images from Video during the Test and Test Result of Power Control Device Figure 12 shows the real-time control performance of the test bed obtained using the program for remote power control. The figure also shows photograph of the remote water jet controlling the engine speed at 1800 RPM, then remotely controlling the steering angle of the water jet in the range of 25° port and 25° starboard; the corresponding motion of the waterjet is also shown. This system is currently installed in the USV platform systems and various tests have been completed.