Autonomous navigation system, navigation control method, and autonomous navigation program

Abstract

This invention provides an autonomous navigation system, navigation control method, and program for unmanned vessels that perform safe evasive maneuvers while clearly indicating their intention to avoid collisions to surrounding vessels. [Solution] Information on objects to be avoided in the vicinity of the unmanned vessel is acquired, and if the proximity to such objects meets predetermined conditions, an avoidance target course is set to eliminate the proximity, and a significantly larger turning motion or speed change than during normal navigation is initiated. Furthermore, the course change is restricted until the conditions for passing the object are met, and coordinated avoidance actions are performed among multiple unmanned vessels as needed. This avoids unclear tracks caused by minute avoidance behaviors based on machine learning, and demonstrates an objectively recognizable avoidance intention early on, thereby achieving safe and predictable autonomous navigation.

Description

Technical Field

[0001] The present invention relates to a technology for controlling the autonomous navigation of an unmanned boat (ASV: Autonomous Surface Vehicle or USV: Unmanned Surface Vehicle) without crew on board. More specifically, in ocean observation, infrastructure inspection, maritime security, or cooperative transportation by a group of unmanned boats (swarm), while avoiding the risk of collision with manned boats or other unmanned boats navigating in the vicinity, it relates to an autonomous navigation system, a navigation control method, and an autonomous navigation program for performing objective and clear avoidance maneuvers that conform to international or regional maritime traffic rules.

Background Art

[0002] In recent years, with the development of various sensor technologies and artificial intelligence (AI), the development and sea trials of fully autonomous unmanned boats without crew on board have been progressing worldwide. When an unmanned boat navigates in general waters, just like a manned boat, safe navigation conforming to international or local maritime traffic rules is essential. Generally, when ships approach each other at sea, it is widely practiced in actual operations for the ship operators to confirm each other's intentions (such as "passing starboard to starboard" or "decelerating and passing on the stern side") through voice communication using an international VHF radio or the like. However, in the case of an autonomous unmanned boat without remote control, there are many cases where it is difficult to directly transmit the avoidance intention of the own ship to other surrounding ships through voice or visual signals, and it is necessary to clearly show the intention to other ships only by the physical behavior of the own ship (obvious changes in the course or increase or decrease in speed).

[0003] Regarding the intention transmission by such physical behavior, in actual operation, as an operation to avoid collision, it is important to make a course change or speed change large enough to be easily recognized by other ships, suppress continuous small course changes, and confirm the effect of the operation until the other ship passes safely. That is, it is required to take a clear action on a scale that can be immediately recognized by the operator of the other ship visually or by radar, etc., rather than making a minute course change.

[0004] Prior technologies for unmanned vessel avoidance control include techniques that use machine learning (reinforcement learning, etc.) to calculate cost functions in real time based on the difference between the vessel's bearing angle and the bearing to the destination (waypoint), and cost functions based on the approach distance to other vessels, and generate continuous avoidance maneuvers that minimize the total cost. There are also known systems that share information on the movements of the vessel and other vessels, calculate collision prediction areas (warning areas / attention areas) based on the vessel's velocity vector, and then display icons or messages on a terminal to notify the operator of a manned vessel, thereby assisting navigation. [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2021-18484 [Patent Document 2] Japanese Patent Publication No. 2024-009073 [Overview of the project] [Problems that the invention aims to solve]

[0006] However, conventional algorithms that generate real-time optimized avoidance trajectories using machine learning or model predictive control (MPC), such as those described in Patent Document 1, have significant safety issues in practical operation. Machine learning-based algorithms make minor adjustments to the course at control cycles (e.g., every second) based on changes in relative distance to other vessels and a cost function that evaluates the efficiency of the vessel reaching its destination. The resulting physical behavior tends to be either a continuous series of small changes in course and speed, like chattering, or conversely, an extremely smooth trajectory with tiny curves. Such behavior makes it extremely difficult for operators of nearby manned vessels to determine whether the unmanned vessel has recognized them and is intentionally avoiding them, or whether the vessel is simply being rocked by waves and wind, potentially leading to fatal confusion and unpredictable situations at sea.

[0007] Furthermore, if a quick reversal occurs after the initial evasive maneuver, there is a risk that the paths will intersect again before a safe passage can be achieved.

[0008] Furthermore, when multiple unmanned vessels sail in a swarm, if each vessel initiates evasive maneuvers independently based on its own machine learning model, the overall formation of the swarm collapses. This creates a situation where, from the perspective of other vessels, it appears as if "multiple unpredictable obstacles are scattering out," actually increasing the risk of collision—a challenge unique to cooperative control. Notification systems like the one described in Patent Document 2 are designed for manned vessels and do not solve the physical control of fully autonomous swarms of unmanned vessels.

[0009] This invention has been made in view of the complex problems of the prior art described above, and aims to provide an autonomous navigation system, a navigation control method, and an autonomous navigation program that can clearly and early indicate the ship's avoidance intention to surrounding vessels without relying on minute optimization behaviors of machine learning, and in which the avoidance intention is objectively expressed as behavior that can be observed from the outside. [Means for solving the problem]

[0010] To solve the above problems, an autonomous navigation system according to one aspect of the present invention comprises an acquisition unit that acquires information on objects to be avoided that are present around an autonomously navigating unmanned vessel, and a navigation control unit that executes an avoidance operation when the approach state satisfies predetermined approach conditions. In this system, the navigation control unit sets an avoidance target course to resolve the approach state at the start of the avoidance operation, and controls the change in course from the unmanned vessel's current course to the avoidance target course to be executed as a "turning operation or change in speed that is clearly larger than that of a predetermined period before the start of the operation (during normal navigation)." In other words, immediately after detecting a collision risk, it deliberately restricts the generation of a smooth internal trajectory and forces a hard and large initial operation.

[0011] Furthermore, an autonomous navigation system relating to another embodiment of controlling the navigation of a group of unmanned vessels including multiple unmanned vessels includes an action setting unit that sets actions to cause multiple unmanned vessels to perform evasive maneuvers in the same direction when at least one unmanned vessel initiates an evasive maneuver. This action setting unit is characterized in that, in order to show a clear intention to evade while maintaining the formation of the group, the control is set such that the steering amount or speed change amount differs between the inner vessel located on the inside of the turn and the outer vessel located on the outside during the evasive maneuver in the substantially same direction. [Effects of the Invention]

[0012] According to the present invention, at the start of an evasive maneuver, the smooth, continuous, minute course changes characteristic of machine learning are intentionally eliminated, and a "clearly larger turning motion or change in speed than before the start of the maneuver" is forced to be executed. As a result, the track vector of the unmanned vessel changes discontinuously and dramatically on radar and AIS, so that the evasive intention of the unmanned vessel (for example, to change course sharply to starboard to avoid the obstacle) can be communicated to the operators of surrounding manned vessels early and without doubt. [Brief explanation of the drawing]

[0013] [Figure 1] This is a schematic diagram showing an example of the overall configuration and applicable environment of an autonomous navigation system according to the first embodiment of the present invention. [Figure 2] This is a block diagram showing the hardware configuration and functional blocks related to autonomous navigation of an unmanned vessel according to the first embodiment. [Figure 3] This is a plan view comparing the trajectory of the evasion maneuver of the unmanned vessel in this embodiment with the trajectory of the conventional technology based on machine learning. [Figure 4] This flowchart shows the specific procedures for navigation control processing and avoidance actions in the first embodiment. [Figure 5] This is a schematic diagram showing the overall configuration of a cooperative control system for a swarm of multiple boats according to a second embodiment of the present invention. [Figure 6]This is an explanatory diagram showing the difference in steering amount and speed change amount between the inner boat located on the inside of a turn and the outer boat located on the outside during the evasive maneuver of the unmanned boat group in the second embodiment. [Figure 7] This flowchart shows the specific procedures for navigation control processing and cooperative avoidance operations for a group of multiple vessels in the second embodiment. [Figure 8] This sequence diagram shows the communication network configuration between each unmanned vessel, the land base station, and other vessels, as well as the procedure for sending and receiving intent information. [Modes for carrying out the invention]

[0014] Embodiments of the present invention will be described in detail below with reference to the drawings. In this specification and the drawings, components having substantially the same functional configuration are denoted by the same reference numerals, and redundant explanations will be omitted. The autonomous navigation system according to the present invention is a system for avoiding collisions with surrounding manned vessels, other unmanned vessels, or stationary obstacles when an unmanned vessel (ASV or USV) without a crew is navigating on water such as seas or lakes, and for clearly communicating its avoidance intention to those around it. The autonomous navigation system of this embodiment is broadly divided into a first embodiment applied to a single unmanned vessel and a second embodiment applied to an environment in which multiple unmanned vessels form a swarm and navigate in a coordinated manner. In the following description, the basic control mechanism and avoidance logic of a single vessel according to the first embodiment will be described in detail first, and then the specific coordinated avoidance operation in a group of multiple vessels (such as differences in behavior between inner and outer vessels) will be described as the second embodiment.

[0015] [First Embodiment: Overall System Configuration] FIG. 1 is a schematic diagram showing an example of the overall configuration and application environment of the autonomous navigation system according to the first embodiment of the present invention. The autonomous navigation system 1 of the present embodiment is configured centering on an unmanned boat 10 that autonomously navigates on water. The unmanned boat 10 has a function of navigating toward a preset destination (waypoint) and executes various missions such as ocean observation, infrastructure inspection, maritime security, and material transportation. There may be an avoidance target 50 in the navigation area of the unmanned boat 10. The avoidance target 50 is assumed to be a manned boat on which an operator is on board, an unmanned boat operated by another operator, a drifting object, or a stationary obstacle such as a reef. The unmanned boat 10 senses the situation around itself and constantly monitors the approach state (collision risk) with the avoidance target 50. Further, the unmanned boat 10 is configured to be able to communicate with a control center installed on land and other ships via a wireless communication network and has a function of transmitting intention information indicating its own navigation state and avoidance intention to the outside. Thereby, surrounding ships can easily predict the behavior of the unmanned boat 10.

[0016] The unmanned boat 10 uses various sensor groups mounted on the hull to accurately acquire own-ship state information such as the current position of the own ship, the bow azimuth, the ground speed and water speed, and the attitude (roll, pitch, yaw). At the same time, it acquires other-ship information such as the position, relative distance, relative azimuth, and movement vector (course and speed) of the surrounding avoidance target 50. These pieces of information are input into an information processing device (computer) provided inside the unmanned boat 10, and the collision risk level is determined based on a predetermined algorithm. Generally, when ships are in a meeting relationship, the necessity and priority of avoiding navigation according to the situation are determined in operation. The unmanned boat 10 is designed to realize a clear avoidance action that is easy for humans to visually recognize and understand by conforming to these general navigation rules and intentionally excluding "fine and minute course changes" by optimization of machine learning and forcibly executing "a turning operation or a speed change that is clearly larger than before the start operation". This is one of the most important technical features of this system.

[0017] [Hardware Configuration and Functional Blocks] FIG. 2 is a block diagram showing the hardware configuration of the unmanned boat 10 according to the first embodiment and functional blocks related to autonomous navigation. The unmanned boat 10 is roughly divided into a sensor unit 110 that measures the surrounding environment and the state of the boat itself, an actuator unit 120 that is responsible for propulsion and steering, a communication unit 130 that communicates with the outside, and a control device 100 that integrally controls these. The sensor unit 110 includes, for example, a global navigation satellite system (GNSS) receiver, an inertial measurement unit (IMU), an automatic identification system (AIS) for ships, a marine radar, optical sensors such as a stereo camera and LiDAR (Light Detection and Ranging), and a wind direction and speed meter. Thereby, the unmanned boat 10 can surely capture the avoidance target object 50 using both radio wave information and optical information, and track its movement. In particular, information from AIS and radar is indispensable for vector analysis of the target object, and is continuously used as the basic data for determining the approaching state in the information acquisition unit 101.

[0018] The actuator unit 120 includes a main engine (for example, an electric motor or an internal combustion engine) that applies propulsion force to the unmanned boat 10 and a propulsion mechanism such as a propeller, and a steering mechanism such as a rudder or a swiveling thruster for changing the bow direction. The actuator unit 120 receives a control command (target rotation speed, target rudder angle, etc.) from the control device 100 and changes the physical behavior of the unmanned boat 10 accordingly. The communication unit 130 includes a VHF band radio, a satellite communication device, or a mobile communication module such as 4G / 5G, and transmits and receives data to and from an onshore base station or other ships. The control device 100 is a computer (information processing device) equipped with a storage device such as a CPU, a GPU, a ROM, a RAM, and a flash memory. An autonomous navigation program according to the present invention is stored in the storage device, and by the CPU reading and executing this program, various functional blocks described later are logically realized.

[0019] The control device 100 comprises, as functional blocks, an information acquisition unit 101, a proximity state determination unit 102, a navigation control unit 103, and an intention information transmission unit 104. The information acquisition unit 101 comprehensively collects data output from the aforementioned sensor unit 110 and communication unit 130 to acquire information about the vessel and the object to be avoided 50. At this time, the information acquisition unit 101 also calculates and acquires the range of deviation in the course of the unmanned vessel 10 caused by external disturbances such as waves, wind, or currents. Based on the information acquired by the information acquisition unit 101, the proximity state determination unit 102 periodically calculates the relative distance (DCPA: Distance at Closest Point of Approach) and the time to Closest Point of Approach (TCPA: Time to Closest Point of Approach) between the unmanned vessel 10 and the object to be avoided 50, and determines whether these meet predetermined proximity conditions (i.e., whether the risk of collision exceeds a threshold). These proximity conditions are dynamically set so as to be changeable according to the degree of congestion in the sea area and the maneuverability of the unmanned vessel 10. Furthermore, the communication unit 130 is a communication interface for sending and receiving information with a land base station or mother ship, another ship, or an external network, and may use, for example, AIS, VDES, a mobile communication network, or a wireless LAN. In addition, the communication unit 130 may be configured to send and receive the information via satellite communication (SATCOM) via the artificial satellite 140 in order to realize beyond line of sight communication or wide-area communication.

[0020] The navigation control unit 103 has a core function of controlling the propulsion mechanism and steering mechanism to execute an avoidance maneuver when the proximity condition determination unit 102 determines that the proximity condition has been met. In this embodiment, the navigation control unit 103 sets a new "avoidance target course" to resolve the proximity condition at the start of the avoidance maneuver. Then, it controls the change of course from the current course of the unmanned vessel 10 to the avoidance target course to be executed all at once as a "larger turning maneuver or change in speed than the predetermined period before the start of the maneuver (during normal navigation)". Here, a "large turning maneuver" means, for example, that while the maximum rudder angle change allowed during normal waypoint following is 5 degrees, at the start of the avoidance maneuver a large rudder angle of 15 degrees or more is temporarily commanded, and the heading of the ship is changed significantly (for example, by 30 degrees or more) in a short time. As a result, the track remaining on the radar screen of other ships is clearly bent, intentionally creating a situation in which the ship's intention to avoid the obstacle is clearly observed and recognized from the outside. Here, "a predetermined period before the start of the operation" refers to the observation period set immediately before the start of the avoidance operation (change of course or change of speed), and can be set as, for example, the preceding T seconds (e.g., 1 second to 10 seconds) or the most recent N control cycles (e.g., 10 cycles to 200 cycles). Note that T and N may be fixed values, or they may be variable depending on the navigation conditions (speed, disturbances, sensor update cycle, etc.).

[0021] Furthermore, the intention information transmission unit 104 transmits intention information to the outside, based on either the actual trajectory, trajectory history, or predicted trajectory of the unmanned vessel 10, immediately before the navigation control unit 103 initiates the avoidance operation described above, or while the turning operation is being performed. Specifically, it broadcasts updated data on the avoidance target course and status information such as "changing course to starboard" to surrounding vessels via maritime communication protocols such as VDES (VHF Data Exchange System). This allows operators of other vessels to confirm that the unmanned vessel 10 is indeed taking avoidance action, not only through primary information such as visual observation or physical sharp turns on radar, but also through secondary information such as digital data transmitted via communication, dramatically improving safety at sea. This synchronization of physical behavior and data communication brings about superior effects not found in conventional autonomous navigation algorithms.

[0022] Figure 3 is a plan view comparing the trajectory of the evasive maneuver of the unmanned vessel 10 in this embodiment with the trajectory of a conventional technique based on machine learning. In the figure, the dotted line represents an example of an evasive trajectory generated by conventional reinforcement learning or model predictive control (MPC). This conventional trajectory places extreme emphasis on efficiency in reaching the destination, and therefore attempts to smoothly return to the original course by drawing an extremely gentle curve while maintaining a constant clearance from the object to be avoided 50. However, as mentioned above, such minute, continuous changes in course are difficult to distinguish from noise or hull movement caused by waves on the radar screen of other vessels, making it difficult to give the operator of another vessel the assurance that "the unmanned vessel is intentionally avoiding the object." As a result, other vessels may take excessive evasive action, which can actually create a secondary risk of collision where the two vessels' lanes intersect.

[0023] In contrast, the solid line in Figure 3 shows the avoidance trajectory executed by the navigation control unit 103 of this embodiment. At the point where the approach conditions are met (the avoidance start point), the unmanned vessel 10 executes a sudden turning maneuver with a significantly larger rudder angle than during the navigation period before the start of the maneuver. As a result, the track appears as a clear "broken line" or "arc with a sharp curvature," as shown in the figure. The operators of other vessels can immediately see this dramatic change in heading on their radar or AIS vector display and intuitively understand that "the unmanned vessel has recognized their vessel and is attempting to fulfill its avoidance obligation." Even if the target trajectory calculated within the system is inherently smooth, this system intentionally restricts the gradual course changes at the start of the avoidance maneuver and prioritizes executing a series of clear, continuous turning maneuvers, thereby creating an objectively recognizable indication of intent from surrounding vessels.

[0024] Furthermore, after such a turning maneuver or change in speed, the navigation control unit 103 implements control that strictly restricts the unmanned vessel 10 from changing its course back toward the object 50 (so-called reverse maneuver) until the relative position of the object 50 to the unmanned vessel 10 satisfies the "predetermined conditions for completion of passage". Specifically, it maintains the state of being on the avoidance target course after turning (hold state) and prevents dangerous behavior such as inadvertently trying to return to the original waypoint and crossing the bow of another vessel. During this hold period, minor course corrections are permitted, but steering commands that would point the bow toward a predetermined prohibited sector centered on the current position of the object 50 are forcibly filtered or blocked within the control algorithm.

[0025] This "completion condition" is set as a condition that includes at least one of the following: the relative distance of the object to be avoided 50 as seen from the unmanned vessel 10 changes from decreasing to increasing (i.e., the closest point has been safely passed), or the relative azimuth angle of the object to be avoided 50 relative to the direction of travel of the unmanned vessel 10 exceeds a predetermined angular threshold (for example, beyond the direct side to 100 degrees aft). When the information acquisition unit 101 and the proximity state determination unit 102 confirm that this condition has been met, the navigation control unit 103 releases the aforementioned restriction on course change. After the restriction is released, the unmanned vessel 10 recalculates the target trajectory to return to its original destination and transitions to normal autonomous navigation mode by passing through the sea area where safety has been confirmed. In this way, the program mechanically and reliably ensures the safety operation requirement of confirming the effectiveness of the avoidance action until the other vessel has completely passed a safe distance.

[0026] Furthermore, the navigation control unit 103 also performs flexible speed control according to the size of the surrounding navigable water area and the estimated time (relative distance) to reach the object to be avoided 50. The information acquisition unit 101 acquires the size of the navigable water area around the unmanned vessel 10 from nautical chart data and radar information, and if it determines that the size is greater than or equal to a predetermined standard value (i.e., the open ocean with sufficient space to avoid obstacles), the navigation control unit 103 performs the aforementioned single large course change while maintaining a speed of at least the "minimum speed" set based on the speed of the unmanned vessel 10 before the start of the avoidance operation. Maintaining speed while making a large course change leads to a greater emphasis on the vector change relative to other vessels and is also effective in maintaining good rudder effectiveness (steering responsiveness) of the unmanned vessel 10. In wide water areas, avoiding unnecessary deceleration also has the effect of preventing delays in mission execution.

[0027] On the other hand, in narrow waterways where reefs are approaching or in areas where other vessels are densely packed, if the estimated time or relative distance for the unmanned vessel 10 to reach the object 50 to be avoided (the distance at which it can safely turn) falls below a threshold even shorter than the predetermined approach conditions, the navigation control unit 103 switches to a different control mode. In these highly urgent situations, simply turning with a large rudder angle is insufficient to avoid a collision, or a secondary risk of colliding with another obstacle arises due to the turn. Therefore, in the avoidance operation, the navigation control unit 103 performs a speed change that forcibly reduces the speed of the unmanned vessel 10 to below the aforementioned "minimum speed." Specifically, the design minimizes the collision energy with the object while securing time to observe the situation by rapidly reducing the output of the propulsion engine or performing a sudden stop (crash astern) by reversing. Here, "processing range" refers to the range that ensures the necessary margin to safely complete the avoidance maneuver, and is determined, for example, based on the unmanned vessel's minimum turning radius, maximum deceleration (or stopping distance), steering response delay, and a predetermined safety margin. The processing range can be expressed, for example, as (1) a safety area set around the object to be avoided, (2) a prohibited area that the future trajectory must not enter based on the vessel's kinematic model, or (3) a determination area representing the minimum margin at which an avoidance maneuver should be initiated. Furthermore, "estimated arrival time" and "relative distance" are indicators representing the proximity between the object to be avoided and the unmanned vessel, and at least one of them can be used to determine collision risk. For example, estimated arrival time can be calculated based on the relative position vector and relative velocity vector of the unmanned vessel and the object to be avoided, as the time when they are closest together (TCPA) or the estimated time until they reach a predetermined distance. Relative distance can be calculated as the current relative distance, the closest approach distance (DCPA), or the minimum distance on the future trajectory. In terms of implementation, a collision risk may be determined if at least one of the following conditions is met: (1) estimated arrival time is below a threshold, or (2) relative distance is below a threshold.

[0028] Furthermore, the navigation control unit 103 of this embodiment implements a robust avoidance control algorithm that takes into account the presence of disturbances. The unmanned vessel 10 at sea is constantly affected by disturbances such as waves, wind, or currents, and especially in the case of small unmanned vessels, it is unavoidable that the bow heading will momentarily swing by several to more than ten degrees (yawing) when hit by waves. The information acquisition unit 101 calculates and acquires the steady-state fluctuation range of the course (variance, standard deviation, etc.) caused by these disturbances in real time from time-series data of the IMU and compass. When the navigation control unit 103 performs an avoidance operation under the influence of disturbances, the larger this fluctuation range is (rougher the sea conditions), the larger the amount of course change or speed change in the initial turning operation of the avoidance operation will be set and controlled accordingly.

[0029] The technical significance of this disturbance-linked control correction is extremely great. For example, in a situation where the hull is constantly being tossed ±5 degrees by waves, even if a ship changes course by 10 degrees to avoid a disturbance, it is highly likely that other ships in the vicinity will perceive this as "behavior within the range of being tossed by waves" and will not recognize it as an attempt to avoid a disturbance. Therefore, when the navigation control unit 103 detects that the fluctuation range is ±5 degrees, it forcibly outputs a course change command of, for example, 20 degrees or more, so as to clearly exceed the noise floor. In other words, by generating a "physical signal (behavior)" that can ensure a sufficient S / N ratio (signal-to-noise ratio) relative to the baseline disturbance noise level, it is possible to reliably and clearly communicate the ship's attempt to avoid a disturbance to other ships, even in rough weather.

[0030] In addition, the fail-safe function for unexpected situations occurring during the "course change restriction period (hold period)" for avoidance target 50 will also be explained. The navigation control unit 103 will execute exceptional processing if, during the period in which the unmanned vessel 10 is restricted from returning its course towards the avoidance target 50, the information acquisition unit 101 detects a third vessel or obstacle and determines that a "new approach state" with it meets predetermined conditions (i.e., another danger has appeared beyond the avoidance point). Specifically, the restriction on changing course towards the initial avoidance target 50 is temporarily lifted or relaxed, and a secondary avoidance action is performed to resolve the new approach state. This prevents the system from becoming trapped due to excessive adherence to mechanical rules and ensures the safety of the entire system.

[0031] Figure 4 is a flowchart showing the specific procedures for navigation control processing and avoidance operations in the first embodiment. The series of processes shown in this flowchart are realized by the CPU of the control device 100 mounted on the unmanned vessel 10 repeatedly executing an autonomous navigation program stored in the memory device at a predetermined control cycle (for example, a 100 millisecond or 1-second cycle). This process begins when the unmanned vessel 10 departs from a port or mother ship and enters autonomous navigation mode. First, the control device 100 calculates the target trajectory to waypoints based on pre-provided mission data, generates baseline control command values ​​(normal course and speed) for the propulsion and steering mechanisms, and starts navigation (step S10).

[0032] After commencing navigation, the information acquisition unit 101 acquires environmental data regarding the situation around the vessel and information on the movements of other vessels via the sensor unit 110 and the communication unit 130 (step S11). Specifically, it acquires the vessel's highly accurate position coordinates, azimuth angle, and ground speed from the GNSS and IMU. Simultaneously, it receives static information (such as ship names) and dynamic information (position, ground course, ground speed) broadcast from surrounding vessels via the AIS receiver. Furthermore, it uses the ARPA function of the maritime radar to extract the position and relative velocity vectors of unregistered avoidance targets 50 such as small vessels not equipped with AIS, buoys, and floating objects. Through the acquisition of this combined data, the unmanned vessel 10 grasps the surrounding situation in three dimensions.

[0033] Next, the proximity determination unit 102 determines the proximity state to quantitatively evaluate the future collision risk based on the acquired vector information of the own vessel and the object to be avoided 50 (step S12). Specifically, it extends the relative velocity vectors of the current own vessel and the other vessel and geometrically calculates the point where they are closest (CPA). Then, it calculates the relative distance at that closest point (DCPA) and the time from the current time to reach that closest point (TCPA). The proximity determination unit 102 compares these values ​​with a preset safety threshold (for example, a threshold of 0.5 nautical miles for DCPA and a threshold of 15 minutes for TCPA) and determines whether the unmanned vessel 10 and the object to be avoided 50 are in a relative position and whether there is a risk of collision.

[0034] If, as a result of the determination in step S12, neither DCPA nor TCPA falls below the safety threshold (i.e., the predetermined approach conditions are not met), the control device 100 determines that it is safe to maintain the current target trajectory and continues normal navigation from step S10. On the other hand, if the approach conditions are met, the navigation control unit 103 determines, based on the relative position relationship based on the facing relationship, whether the vessel should prioritize evasive action and transitions to a mode for executing evasive action. The most distinctive feature of this embodiment is that, in the initial response immediately after transitioning to this evasive mode, smooth minor adjustments by machine learning are eliminated, and a large, clear change in behavior is deliberately made.

[0035] When transitioning to avoidance mode, the navigation control unit 103 sets a new "avoidance target course" to resolve the proximity situation (step S14). For example, if it is determined that the unmanned vessel 10 and the object to be avoided 50 are approaching from the front, the navigation control unit 103 calculates an avoidance target course that reduces mutual course interference while ensuring a safe distance from the other vessel. Therefore, the navigation control unit 103 calculates an avoidance target course that is changed by, for example, 30 degrees or more to the starboard side from the current course. At this time, it refers to the size of the surrounding navigable water area (presence or absence of shoals and other vessels) acquired by the information acquisition unit 101, confirms that there is enough space to safely make a large turn, and then determines an avoidance target course of a size that can be clearly recognized by other vessels.

[0036] When an avoidance target course is set, the navigation control unit 103 outputs a control command to the actuator unit 120 to execute the course change from the current course to the avoidance target course in one go as a "turning operation larger than the predetermined period before the start operation" (step S15). Specifically, the rudder angle limit (e.g., a maximum of 5 degrees) that is used when following a waypoint during normal navigation is temporarily released, and a large rudder angle close to the maximum rudder angle (e.g., 35 degrees) is commanded (an operation similar to a hardover). As a result, the heading of the unmanned vessel 10 changes dramatically in a short time, drawing a clear wake on the sea. Operators of other vessels can visually perceive the unmanned vessel 10's intention to avoid obstacles by observing the occurrence of this wake and the change in the radar vector.

[0037] Furthermore, the navigation control unit 103 may perform a clear change in speed simultaneously with or instead of a turning operation (step S16). As mentioned above, if the surrounding navigable waters are wide and safe, the course change is performed while maintaining a speed above the "minimum speed" in order to maintain rudder effectiveness. However, in an urgent situation where the estimated time to reach the object to be avoided 50 (TCPA) is extremely short and emergency collision avoidance is required, the navigation control unit 103 applies the brakes by rapidly reducing or reversing the propeller rotation speed (crash astern). This rapid deceleration operation is also displayed on other ships' AIS and radar as a "dramatic reduction in the speed vector," thus having the effect of strongly communicating to other ships that the unmanned vessel 10 is aware of the crisis.

[0038] In parallel with performing such large turning maneuvers and speed changes, the intent information transmission unit 104 transmits digital data indicating the unmanned vessel 10's avoidance intention to external vessels or land-based base stations (step S17). Specifically, it broadcasts status information such as "currently changing course 30 degrees to starboard" and "decelerating by reversing the engine," as well as updated predicted trajectory data, via VHF data communication (VDES), etc. This allows surrounding vessels to cross-reference the physical behavior of the unmanned vessel 10 (primary information) with the digital intent information received via the communication network (secondary information). This dual intent transmission mechanism dispels the "eerie" and "unpredictability" unique to unmanned vessels that other vessel operators may feel, and supports safe passage.

[0039] After performing clear evasive maneuvers (steps S15-S16) and the bow of the unmanned vessel 10 reaches the evasive target course, the navigation control unit 103 transitions to a "hold period" (step S18). During this period, the unmanned vessel 10 is strictly restricted from changing course again toward the evasive object 50 until its relative position with the evasive object 50 satisfies the "passage completion condition". Conventional algorithms using machine learning tend to immediately try to return to the original destination by making small adjustments to the rudder as soon as the relative distance increases even slightly, which gives other vessels a sense of fear that "it's coming towards us". In this embodiment, by providing this hold period, the course is fixed, forming a consistent and stable evasive trajectory.

[0040] However, even during this hold period, the unmanned vessel 10 continues to monitor its surroundings. If, for example, an unforeseen event occurs while the vessel is holding a large course to starboard, such as a third vessel (a new object to avoid) approaching from ahead of its direction of travel, the proximity determination unit 102 detects a new collision risk. If such a "new proximity condition" meets predetermined conditions, the navigation control unit 103 exceptionally releases or relaxes the restriction on changing course toward the original object to avoid 50 (step S19). Then, while maintaining a safe distance from the original object to avoid 50, it allows secondary evasive maneuvers to avoid the third vessel (for example, a slight change of course to port), thereby preventing the system from becoming rigid and causing a secondary accident.

[0041] Normally, during the hold period, the process of the unmanned vessel 10 and the object to be avoided 50 passing each other (meeting or overtaking each other) proceeds. The navigation control unit 103 continuously determines whether the "passage completion condition" has been met based on the latest relative position information acquired by the information acquisition unit 101 (step S20). The passage completion condition is a condition to objectively ensure that the other vessel has passed to a position that is sufficiently safe from the perspective of the own vessel. It is not sufficient for the relative distance to be a certain extent; the first condition is that the relative distance (DCPA) of the object to be avoided 50 as seen from the unmanned vessel 10 has clearly shifted from decreasing to "expanding" (the point of closest approach has been passed).

[0042] Furthermore, as a second condition to ensure more certainty of successful passage, the relative azimuth angle of the object to be avoided 50 with respect to the direction of travel of the unmanned vessel 10 is referenced. For example, it is confirmed that the object to be avoided 50 has passed directly beside the vessel (90 degrees) and moved behind it (for example, at an angle of 100 degrees or 110 degrees or more). In this embodiment, the conditions for successful passage are considered to have been met when both the relative distance begins to increase and the relative azimuth angle exceeds a predetermined angular threshold, or when at least one of these conditions is clearly met and it is predicted that the course of the other vessel will not intersect with the course of the vessel. This strict determination criterion ensures that the safety operation requirement of confirming the effectiveness of the avoidance maneuver until the other vessel has completely passed is met.

[0043] If it is determined that the conditions for completion of passage have been met (YES in step S20), the navigation control unit 103 terminates the avoidance mode and begins processing to return the unmanned vessel 10 to its original destination (step S21). Specifically, it recalculates a new target trajectory to go from the current position to the next waypoint (or final destination). In generating this return trajectory, since there is no longer an immediate collision risk in the surroundings, it is not necessary to make a dramatic change of course using a large rudder angle as was done when the avoidance started. Rather, when returning, it is desirable to avoid abrupt movements and draw a smooth curve that prioritizes fuel efficiency and hull stability to rejoin the original route. Therefore, in return mode, it is possible to resume optimization control using machine learning or MPC.

[0044] When tracking the recovery trajectory begins, the navigation control unit 103 returns to the normal navigation mode (step S10) and continues autonomous navigation along the set route. Thus, the autonomous navigation system 1 of this embodiment has a hybrid control architecture that clearly switches the quality of the unmanned vessel 10's behavior (size of rudder angle and sharpness of course change) between normal navigation and the start of avoidance maneuvers. This achieves a high level of balance between two inherently conflicting requirements: "smoothness" for efficient navigation across vast ocean spaces and "clarity (dynamics)" for communicating intentions to people in the vicinity when necessary.

[0045] Here, we would like to add a note about the redundancy and derivative configurations of sensors specific to ASVs, which are important when operating a single unmanned vessel 10 in actual sea areas. In the harsh environment of the sea (salt damage, direct sunlight, sea spray, etc.), there is always a risk that some sensors may suddenly malfunction. For this reason, the sensor unit 110 of this embodiment is redundantly equipped with multiple sensors based on different principles. For example, it uses both an image recognition module using an optical camera and a three-dimensional point cloud measurement module of an object using millimeter-wave radar or LiDAR. This ensures that even if the camera's view is obscured by backlight or dense fog, the system can continue to track the object to be avoided 50 by measurement using radio waves or laser pulses, and the data supply to the information acquisition unit 101 is not interrupted.

[0046] The information acquisition unit 101 implements "data fusion (sensor fusion)" technology, which integrates and processes heterogeneous data obtained from multiple sensors. For example, AIS data accurately conveys the absolute position and name of an object, but its update cycle can be slow, ranging from several seconds to tens of seconds. On the other hand, millimeter-wave radar and cameras have short update cycles (tens of milliseconds) and can capture changes in relative distance to an object in real time. The information acquisition unit 101 integrates these using estimation algorithms such as Kalman filters and particle filters to estimate the current position and velocity vector of the object to be avoided 50 with high accuracy and low latency. This high-precision vector information provides a robust foundation for the navigation control unit 103 to activate a "clear, one-time, large course change" at the appropriate timing.

[0047] Furthermore, it is conceivable that satellite or mobile communications may suddenly be lost while the unmanned vessel 10 is navigating the open ocean far from the land base station. The control device 100 of this embodiment is equipped with an autonomous fail-safe function to ensure safety even when communications are lost. If the communication unit 130 is unable to receive external control signals or intent information from other vessels (such as VDES), the navigation control unit 103 automatically switches to a standalone avoidance mode that relies solely on self-contained information (radar and camera data) obtained from the vessel's sensor unit 110. In this state, the aforementioned approach conditions (such as the DCPA threshold) are readjusted to be stricter than normal (for example, to 1.0 nautical mile) in order to provide a larger safety margin against unknown obstacles.

[0048] Furthermore, the introduction of infrared (IR) cameras is also effective as a complement to optical sensors during nighttime navigation or in rough weather. At night at sea, there are cases where small fishing vessels and other vessels do not have their legally mandated lights on, or are hidden in the waves and are extremely difficult to see (so-called unlit vessels or drifting containers). The infrared camera mounted on the sensor unit 110 can detect heat sources emitted by objects (engine exhaust heat or human body heat), making it possible to detect even highly stealthy obstacles that are difficult to see with visible light cameras or radar at an early stage. The proximity condition determination unit 102 takes this heat source tracking data into account, and can determine proximity conditions at night with the same accuracy as during the day, enabling the navigation control unit 103 to execute clear avoidance actions.

[0049] Furthermore, when the unmanned vessel 10 navigates environments where the terrain itself is the object to be avoided, such as coral reefs or complex island areas, the control device 100 may also use Simultaneous Localization and Mapping (SLAM) technology. A pathfinding algorithm using AI, such as reinforcement learning, is appropriately applied to the real-time local map constructed using SLAM. However, the principle of "clarifying the behavior during avoidance," which is fundamental to this embodiment, is never compromised even in such complex environments. That is, no matter how detailed the obstacle avoidance route calculated by the AI, the navigation control unit 103 applies filtering processing before outputting the final steering command, suppressing small chattering behavior and forcibly intervening in a process that shapes the trajectory into an easily discernible one that combines straight-line movement and large turns as much as possible.

[0050] As described above, according to the autonomous navigation system of the first embodiment, when a single unmanned vessel 10 encounters surrounding vessels or obstacles, it intentionally eliminates the ambiguous and fragmented trajectory that is common in machine learning, and performs a "large and clear change of course or speed" that is immediately obvious from the perspective of a human (another vessel's operator) or radar. This enables early and unambiguous communication of the unmanned vessel's intentions, preventing unnecessary confusion and secondary collisions at sea. Next, a second embodiment of the present invention will be described in detail, which is applied to an environment in which multiple unmanned vessels equipped with such excellent avoidance logic gather and navigate in cooperation as a swarm.

[0051] [Second Embodiment: Cooperative Control System for Multiple Boat Groups (Swarm)] The autonomous navigation system according to the second embodiment of the present invention will now be described. This embodiment is applied to environments in which multiple unmanned vessels, rather than a single unmanned vessel, form a swarm and navigate in coordination in the same sea area. In recent years, there has been a growing need for swarm navigation, which involves the unified control of multiple unmanned vessels while maintaining a certain formation, in applications such as wide-area ocean observation, large-scale offshore infrastructure inspection, or towing and escorting large cargo by multiple vessels. However, if an external object to be avoided (such as a manned vessel) is encountered during such swarm navigation, and each unmanned vessel performs evasive actions independently based on its own judgment, not only will the formation of the entire swarm collapse, but it may also give surrounding vessels the impression that multiple unpredictable obstacles are scattering in all directions, potentially inducing secondary collisions.

[0052] To solve the challenges specific to multiple vessels, the autonomous navigation system according to this embodiment is equipped with a cooperative control algorithm that, when at least one unmanned vessel in a group of unmanned vessels needs to initiate an avoidance maneuver based on its proximity to an object to be avoided, causes all unmanned vessels belonging to that group to perform a clear avoidance maneuver in substantially the same direction while maintaining the group's cohesion. This makes it possible to perceive from the outside as if a single large vessel is performing avoidance maneuvers in a regular manner, and to communicate a clear and consistent intention to avoid obstacles to the operators of other vessels. Figure 5 is a schematic diagram showing the overall configuration of a cooperative control system for a group of multiple vessels according to the second embodiment of the present invention. As shown in the figure, the system may consist of a group of multiple vessels (swarm) 10G composed of multiple unmanned vessels 10A, 10B, 10C (hereinafter simply referred to as unmanned vessel 10 when referred to collectively) interconnected via a communication network, and a central control device 200 installed on a land base station or mother ship that controls them. The central control device 200 includes at least an operation setting unit 200A that sets target trajectories for the normal navigation and avoidance maneuvers of a group of multiple vessels 10G, and an avoidance scenario planning and distribution processing unit 200B that plans an avoidance scenario for the group based on the avoidance target course set by the operation setting unit 200A, and distributes control parameters to each unmanned vessel 10, including steering amount and speed change amount according to the role of the inner vessel / outer vessel, etc.

[0053] Here, the multiple unmanned vessels 10 are equipped with substantially the same hardware configuration (sensor unit 110, actuator unit 120, communication unit 130, etc.) as described in the first embodiment. In this embodiment, either a distributed control method is adopted in which the control device 100 of each unmanned vessel 10 autonomously calculates cooperative operation, or a centralized control method is adopted in which the overall control device 200 calculates and distributes target trajectories and control commands to each unmanned vessel. In the following description, for the sake of explanation, the overall control device 200 (or the control device 100 of a specific unmanned vessel that is the leader of the group) is described as having the function of an operation setting unit 200A that determines the operation of the entire group, and the function of an avoidance scenario planning and distribution processing unit 200B that distributes control parameters to each unmanned vessel 10 based on the settings of the operation setting unit 200A. During normal navigation, the operation setting unit 200A has the function of calculating and setting the target trajectory of each vessel so that each unmanned vessel 10 can navigate while maintaining a preset relative distance, relative speed, and relative angle.

[0054] In determining the proximity state, the closest approach distance (DCPA) and closest approach time (TCPA) are calculated based on the closest unmanned vessel (closest unmanned vessel) among the unmanned vessels 10 constituting the group to the object to be avoided 50 (other vessels, etc.), or on a virtual center coordinate that treats the entire group as a single virtual bounding box. Based on the information acquired by the information acquisition unit, the operation setting unit 200A determines that if at least one unmanned vessel in the group meets the predetermined proximity conditions, the entire group should transition to avoidance mode. Once this determination is made, the navigation control unit of each unmanned vessel 10 is configured to temporarily suspend its individual optimization control and start cooperative control according to the avoidance target course for the entire group set by the operation setting unit 200A and the control parameters calculated and allocated by the avoidance scenario planning and allocation processing unit 200B.

[0055] The core of the group-wide avoidance scenario set by the operation setting unit 200A is to instruct multiple unmanned vessels 10 to perform avoidance maneuvers in the same direction (or nearly the same direction). For example, if a collision occurs where another vessel passes on the starboard side, the operation setting unit 200A commands all unmanned vessels belonging to the group to simultaneously change course to the starboard side. However, if all vessels turn with exactly the same rudder angle and speed, differences in turning radii could cause the trajectories to intersect between the vessel on the inside and the vessel on the outside of the turn, or the relative distance to rapidly decrease, creating a risk of collision within the group (mutual interference). To prevent this, the avoidance scenario planning and distribution processing unit 200B calculates the geometric relationship between the current position of each unmanned vessel and the turning center, and performs a distribution process to assign individual control parameters according to the role of each vessel.

[0056] Specifically, as shown in Figure 6, when an evasive maneuver is set to cause the entire group to make a large turn to the starboard side, the control is set so that the steering amount or speed change amount differs between the unmanned boat located on the inside of the turn (inner boat 10A) and the unmanned boat located on the outside of the turn (outer boat 10C). In Figure 6, P indicates the pivot point of the turn, and RA, RB, and RC indicate the turning radii of the inner boat 10A, the intermediate boat 10B, and the outer boat 10C, respectively. Also, αA, αB, and αC indicate the steering angles (or heading change angles) assigned to the inner boat 10A, the intermediate boat 10B, and the outer boat 10C, respectively, and VA, VB, and VC indicate the speed vectors (magnitude of speed and direction of travel) of the inner boat 10A, the intermediate boat 10B, and the outer boat 10C, respectively. The navigation control unit, based on the control parameters (calculation results of the avoidance scenario planning and distribution processing unit 200b) distributed from the central control unit 200, causes each unmanned vessel to perform a change of course or speed change from its current course as a turning motion or speed change greater than the predetermined period before the start of the operation. This maintains the feature of the first embodiment, which clearly indicates the group's intention to avoid obstacles to other vessels, while simultaneously solving the difficult problem of preventing the collapse of swarm navigation.

[0057] During evasive maneuvers, control that considers the physical relationship between angular velocity and linear velocity is essential to maintain the relative positions of multiple unmanned vessels 10 within a predetermined tolerance range. When the entire group turns in a large arc while maintaining a constant formation, geometrically, the outer vessels 10C located on the outside of the turn must travel a longer distance in the same amount of time compared to the inner vessels 10A located on the inside. Therefore, the operation setting unit generates and outputs a control signal that sets the target rotational speed (i.e., set speed) for the propulsion mechanism of the outer vessels 10C higher than the speed of the inner vessels 10A. This makes it possible for each vessel in the formation to maintain a state of being spread out in a fan shape around the turning axis, while drawing a consistent evasive trajectory as seen from the perspective of surrounding vessels.

[0058] For example, suppose the cruising speed of all boats during normal navigation is 10 knots. When the entire group performs an evasive turn of 45 degrees to starboard, the operation setting unit commands the navigation control units of each boat to reduce the speed of the inner boat 10A, which is closest to the center of the turn, to 8 knots, maintain the speed of the unmanned boat 10B, which is located in the middle of the center, at 10 knots, and increase the speed of the outermost boat 10C, which is making the outermost turn, to 12 knots. In this way, by dynamically increasing or decreasing the linear velocity of each boat in proportion to the distance from the geometric center of the group (turning radius), the angular velocity of the entire group can be kept constant. By setting this speed difference, the relative distance and relative angle between each boat are strictly maintained within the set allowable range (for example, a distance of 50 meters ± 5 meters from an adjacent boat) even during a turn, completely preventing distortion or collapse of the formation.

[0059] In addition to speed settings, the operation setting unit intentionally introduces differences in steering amount (rudder angle setting) between each boat. In order to complete the turn while safely maintaining the relative distance between each boat during the process of the entire group moving simultaneously toward the same avoidance target course, it is necessary to appropriately distribute the turning radius between the inner boat 10A and the outer boat 10C. Generally, if the same speed and rudder angle are taken, the turning trajectory will only move in parallel, but as mentioned above, if there are differences in speed, or if centrifugal force and sideslip during the turn are taken into consideration, the optimal rudder angle required for each boat will differ. The operation setting unit executes a logic that sets the steering amount of the inner boat 10A to be larger than the steering amount of the outer boat 10C.

[0060] Specifically, the inner boat 10A is commanded to use a large rudder angle, for example, 25 degrees, in order to complete the turn sharply and precisely on the inner circumference. On the other hand, the outer boat 10C is commanded to use a rudder angle of about 15 degrees in order to increase speed while smoothly tracing the outermost trajectory. By setting the rudder angle larger for the inner boats and smaller for the outer boats in this way, the turning pivot point for the entire group is fixed in space, and a beautiful concentric turning trajectory is formed. This concentric avoidance behavior sends an extremely clear visual message to surrounding manned vessels that a single, large, controlled system is intentionally avoiding obstacles, and is nothing less than the group-level realization of the "clear turning motion" described in the first embodiment.

[0061] In addition to the pre-configuration by the operation setting unit as described above, external disturbances such as waves and currents exist on the actual sea, making it difficult to perfectly follow the calculated trajectory. Therefore, the navigation control unit of each unmanned vessel 10 in this embodiment simultaneously performs feedback control (local steering control) to maintain the relative distance, relative speed, and relative angle with adjacent unmanned vessels within a predetermined range, even during evasive turns. The sensor unit 110 of each unmanned vessel 10 (particularly millimeter-wave radar, cameras, and RTK-GNSS data sharing via inter-vessel communication) continuously acquires the position and velocity vectors of other unmanned vessels in the group with high precision in milliseconds.

[0062] Each boat's navigation control unit superimposes this local feedback correction value onto the baseline command values ​​for the target avoidance course, target speed, and target rudder angle provided by the operation setting unit. For example, if the inner boat 10A drifts further outward than expected due to the effects of waves, and the relative distance to the adjacent unmanned boat 10B is about to fall below the acceptable lower limit, the navigation control unit of the inner boat 10A instantly corrects by cutting the rudder angle a few more degrees inward or slightly reducing the speed. At the same time, the navigation control unit of the unmanned boat 10B also detects this approach and corrects by slightly avoiding it outward. Such fine adjustments on the order of microseconds to milliseconds are performed autonomously between each boat, thus achieving both the macro-level intention of a large and clear change of course for the entire group and the micro-level collision avoidance within the group.

[0063] Furthermore, the operation setting unit of this embodiment also performs a unique process in determining the completion of passage to decide when to end the avoidance operation as a group. The operation setting unit continuously monitors the relative positional relationship between the "closest approaching unmanned vessel" among the multiple unmanned vessels 10 that is closest to the object to be avoided 50 and the object to be avoided 50. Then, using this closest approaching unmanned vessel as a reference, it determines whether a predetermined "completion condition" indicating the completion of passage of the object to be avoided 50 has been met. This is to prevent situations in which other vessels located at the rear or outer edge of the group are still exposed to danger by making it a condition that the safety of the vessel with the highest risk in the entire group is confirmed.

[0064] The conditions for completing the passage are defined, as in the first embodiment, to include at least one of the following: the relative distance (DCPA) of the object to be avoided 50 as seen from the nearest unmanned vessel clearly shifts from decreasing to increasing, or the relative azimuth angle of the object to be avoided 50 relative to the direction of travel of the nearest unmanned vessel exceeds a predetermined angular threshold (e.g., 100 degrees backward). The operation setting unit receives sensor data and the determination result of the proximity state from the nearest unmanned vessel, and only when it determines that this condition is fully met does it set an operation command to release the avoidance operation (i.e., hold state) in the same direction for the multiple unmanned vessels 10 and distribute it to the entire group.

[0065] Upon receiving the command to cancel the evasion maneuver, the navigation control unit of each unmanned vessel 10 terminates the evasion mode and begins processing to return to the group's original destination (waypoint). Even during the return maneuver, instead of each vessel heading separately back to its original course, the operation setting unit assigns each vessel a smooth target trajectory for the return. During this return, since there is no longer an immediate collision risk in the surrounding area, there is no need to make a sudden change of course using a large rudder angle as at the start of the evasion maneuver. The operation setting unit prioritizes fuel efficiency and hull stability, and restarts optimization algorithms using machine learning and model predictive control (MPC) to execute control that gradually brings each vessel into a predetermined formation.

[0066] Figure 7 is a flowchart illustrating the specific procedures for navigation control processing and cooperative avoidance operations of a group of multiple boats in the second embodiment. This process is mainly executed in conjunction with the operation setting unit 200A and avoidance scenario planning / distribution processing unit 200B of the central control device 200 (or leader boat) and the navigation control unit of each unmanned boat 10. First, in step S30, the operation setting unit 200A sets a target trajectory for normal navigation based on a predetermined formation for the group of unmanned boats, and each boat begins to navigate in accordance with this trajectory. In step S31, environmental data regarding the boat itself and the surrounding objects to be avoided 50 is collected in real time from the sensor groups of each unmanned boat constituting the group and aggregated to the operation setting unit 200A.

[0067] In step S32, the operation setting unit 200A calculates the proximity status (DCPA and TCPA) between the nearest unmanned vessel in the group (or the virtual center of the group) and the object to be avoided 50, and determines the collision risk. If the proximity conditions are not met, the normal navigation in step S30 is continued, but if it is determined that the proximity conditions are met (YES in step S33), the process proceeds to step S34. In step S34, the operation setting unit 200A sets an "avoidance target course in the same direction" for the entire group to resolve the proximity condition. Then, in step S35, the avoidance scenario planning and distribution processing unit 200B calculates control parameters such that the steering amount and speed change amount differ between the inner vessel 10A located on the inside of the turn and the outer vessel 10C located on the outside, in order to maintain the group's formation when turning to the avoidance target course, and distributes these parameters to each vessel.

[0068] The navigation control unit of each unmanned vessel 10 simultaneously executes a "turning motion or speed change larger than the predetermined period before the start of the operation" in step S36, based on the received control parameters. At this time, in step S37, the navigation control unit of each vessel monitors the relative distance and relative speed between vessels using local sensors and superimposes micro-feedback corrections to keep them within an acceptable range. Simultaneously with this large turning behavior, in step S38, the representative vessel of the unmanned vessel group or the intention information transmission unit of the central control device 200 broadcasts digital data (VDES, etc.) indicating the group's intention to avoid the object to be avoided 50 and other surrounding vessels.

[0069] Next, we will explain the process by which a group of boats (swarm) 10G communicates its intention to avoid an object 50 (another boat, etc.) via communication. Figure 8 is a communication sequence diagram showing an example of the transmission, reception, and display of intention information in this embodiment.

[0070] In Figure 8, 210A and 210B represent other vessels (including manned and unmanned vessels) navigating around the group of vessels 10G. Each of these other vessels, 210A and 210B, is equipped with communication devices and capable of receiving at least intent information. Note that 210A and 210B are specific examples of the object to be avoided 50; if the object to be avoided 50 is something other than other vessels, the communication partner corresponding to that object (e.g., air traffic control) may be substituted.

[0071] Figure 8 shows labels 301, 303, 304, and 305, which indicate the types of messages transmitted and received via the communication network. For example, 301 is a status information message containing navigation status information (position, course, speed, etc.) transmitted from other vessels 210A and 210B. 303 is an intention information message transmitted from a group of multiple vessels 10G (or a central control unit 200) to other vessels 210A and 210B, and this intention information message may include at least the avoidance direction (e.g., change of course to starboard), the avoidance target course or predicted trajectory data (e.g., a sequence of points), and the start time (or start conditions) of the avoidance operation. 304 is a response message indicating confirmation of receipt or whether the intention information message 303 was received. 305 indicates an internal notification (or display control) for display updates or warnings from the other vessels based on the received intention information. These messages can be transmitted and received via VDES, AIS, or equivalent maritime communication methods.

[0072] In Figure 8, 401 indicates a display device that presents navigation conditions and received intent information to the operators or monitors of other vessels 210A and 210B. The display device 401 may be an electronic chart display device, a radar display device, a screen of an integrated navigation system, or a user interface equivalent thereto. Furthermore, 402A and 402B may indicate the display state on the other vessels 210A and 210B before receiving intent information (for example, an estimated track display when the operational intent of the group of vessels 10G is unclear), 403A and 403B may indicate the display state after receiving intent information message 303 (for example, a state with added avoidance direction, avoidance target course, or predicted trajectory data for the group of vessels 10G), and 404A and 404B may indicate a display state that emphasizes attention based on intent information (for example, a warning display, highlighting, or indication of a collision risk area).

[0073] Figure 8 is a sequence diagram showing the communication network configuration between each unmanned vessel, a land-based base station or mother ship (central control unit 200), and other ships, as well as the procedure for sending and receiving intention information. As shown in the figure, immediately before the group of vessels 10G begins evasive maneuvers, the central control unit 200 uses AIS or VDES to transmit a message indicating the group's evasive intention (e.g., "The entire group is changing course 30 degrees to starboard") and predicted trajectory data to other ships. The other ships' steering equipment (ARPA radar, etc.) or display device 401 receives this and overlays the predicted evasive route of the group of vessels 10G on the screen. This allows other ships to understand the evasive intention of the group of vessels 10G from both the physical track changes and the digital message, thereby reducing uncertainty regarding the behavior of the unmanned vessel group. Note that "T1" to "T4" in Figure 8 are procedure numbers used to explain the communication procedure and are not drawing symbols (reference symbols).

[0074] Subsequently, in step S39 of the flowchart (Figure 7), the operation setting unit 200A continuously determines whether the passage completion condition based on the nearest unmanned vessel has been met. As long as the passage completion condition is not met, the group as a whole is restricted from returning its course towards the object to be avoided and holds the set avoidance target course. If it is determined that the passage completion condition has been met (YES in step S39), the process moves to step S40, and the operation setting unit 200A issues a command to cancel the avoidance operation. Finally, in step S41, each unmanned vessel 10 returns to the normal navigation mode (step S30) with smooth control, heading towards its original destination according to the return target trajectory newly distributed by the operation setting unit 200A.

[0075] Thus, according to the autonomous navigation system of the second embodiment, even in an environment where multiple unmanned vessels are navigating in a swarm, the entire group can perform large and clear, controlled course changes. By appropriately distributing steering input and speed between the inner and outer vessels, it is possible to suppress the collapse of the formation and collisions between vessels while communicating avoidance intentions to surrounding vessels in accordance with general maritime traffic rules.

[0076] [Variations and other application forms] Although the first and second embodiments of the present invention have been described in detail above, the present invention is not limited to these specific embodiments, and various modifications are possible within the scope of the technical idea described in the claims. Those skilled in the art will readily conceive of alternative configurations and derivative applications, such as those exemplified below, based on the disclosure herein. These modifications are also included within the technical scope of the present invention.

[0077] For example, in the above-described embodiment, the propulsion mechanism of the unmanned vessel 10 was assumed to be a typical ship configuration with a screw propeller and a rudder, but it is not limited to this. The present invention is equally applicable to unmanned vessels equipped with an azimuth thruster capable of moving the hull in any direction as the actuator unit 120, a waterjet thruster, or multiple swivel outboard motors. In this case, the "swivel movement larger than a predetermined period before the start operation" may be embodied not only as a sudden change in the heading (yawing), but also as a clear avoidance vector accompanied by a sudden lateral slide movement (sway) while maintaining the heading. What is important is that the change in behavior is objectively noticeable from the perspective of other ships.

[0078] Furthermore, the parameters used to determine the proximity state are not limited to the closest approach distance (DCPA) and closest approach time (TCPA). The information acquisition unit 101 may define the proximity conditions using "Time To Collision (TTC)" acquired from point cloud data of a stereo camera or LiDAR, or a spatial overlap rate based on a ship domain model. In addition, an image recognition model such as deep learning may be used to directly estimate the course and speed of the other ship from the size of its bow wave and the color of its lights (masthead lights and side lights), and this information may be input to a state estimator such as a Kalman filter to determine the collision risk with higher accuracy and earlier.

[0079] Furthermore, the switching between "minor optimization control using machine learning" and "hard control with a clear large rudder angle" in this system may be dynamically managed not only by a rule-based algorithm but also by a higher-level artificial intelligence agent. For example, a special reward function that evaluates "visibility from other ships (magnitude of trajectory curvature)" could be incorporated into the reinforcement learning model, and the model could be trained to deliberately choose a larger, more costly (higher energy consumption) trajectory in emergencies.

[0080] Furthermore, the arrangement of the operation setting unit in the second embodiment is not limited to the aforementioned central control device 200 of the land base station. For example, an edge computing (or fog computing) configuration may be adopted in which one of the multiple unmanned vessels 10 constituting the unmanned vessel group, the one with the highest communication capability and computing processing capability, is dynamically selected as the leader vessel, and the operation setting unit function is installed in the control device 100 of that leader vessel. In this case, even in an open ocean environment where communication with the land base station is interrupted, it becomes possible to autonomously and self-containedly perform cooperative avoidance operations using only the local communication network within the group (for example, an ad-hoc network such as VHF band or Wi-Fi), dramatically improving the redundancy and fault tolerance of the system.

[0081] Furthermore, the autonomous navigation system of the present invention is applicable not only to swarms consisting solely of unmanned vessels, but also to hybrid fleets that include both manned and unmanned vessels (for example, a configuration in which one manned mother ship leads multiple unmanned workboats). In this case, the system can detect from the sensor unit or steering system signal that the operator of the manned mother ship has manually performed an evasive maneuver (such as a large rudder angle turn), and using this as a trigger, the operation setting unit of the accompanying group of unmanned vessels can automatically calculate an evasive maneuver in the same direction that synchronizes with the evasive trajectory of the manned mother ship, and apply the allocation logic for inner and outer vessels to make them follow. As a result, the operator can concentrate on steering their own vessel without worrying about the risk of collision with unmanned vessels, and the operational burden is greatly reduced.

[0082] The various processes (information acquisition, proximity status determination, navigation control, operation settings, etc.) in the control device 100 and the central control device 200 of the unmanned vessel 10 described above may be implemented by dedicated hardware circuits (such as ASICs or FPGAs), or they may be logically implemented by a general-purpose processor (such as a CPU, MPU, or GPU) executing a software program stored in a storage medium. When the processor executes the program, the program is deployed in memory as a group of modules with a hybrid structure that combines a machine learning inference model (a trained model) and conditional branching processing to realize the "override of large rudder angle commands at the start of operation (rule-based intervention)" which is unique to the present invention, and is periodically calculated at high speed.

[0083] The autonomous navigation program for performing these processes may be pre-installed and provided on a non-transitory computer-readable medium such as ROM or flash memory during the manufacturing stage of the unmanned vessel 10, or it may be configured so that the latest algorithms are distributed and updated as needed via OTA (Over The Air) updates over telecommunication lines such as the internet. By providing this distribution and update function as a program, it is possible to flexibly adapt the threshold values ​​of the control parameters to changes in maritime traffic rules and the emergence of new objects to avoid (for example, new types of high-speed hydrofoils).

[0084] Based on the specific hardware configurations, algorithmic processing flows, and resulting physical behavioral changes of each embodiment described above, the following paragraphs systematically summarize how each invention described in the claims of this specification is supported by the scope described in the detailed description of the invention in a way that a person skilled in the art can recognize that the problem can be solved (support requirement) and by the scope described clearly and sufficiently to enable a person skilled in the art to implement it (enablement requirement). This system offers a solution to the problem of clearly indicating to the outside what the unmanned vessel's avoidance intention is through the coordination of specific functional blocks and hardware.

[0085] [Other variations and derivative embodiments] The first and second embodiments, which form the basis of the present invention, have been described above, but the technical concept of the present invention is not limited to those described above. The following paragraphs detail various modifications and derivative embodiments that can be incorporated into this system or that can function as independent autonomous navigation systems. These provide technical options for future system expansion and adaptation to specific mission requirements, given the diversification of the operating environment of unmanned vessels.

[0086] (Dynamic modification of avoidance margins based on object type recognition) The information acquisition unit 101 may have a function to apply a trained image recognition model, such as deep learning, to image data acquired by the camera to identify the "ship type (attributes of the object)" of the object to be avoided 50. For example, it may identify whether the object is a "large merchant ship" that is likely to maintain a constant course and speed, a "fishing boat" that exhibits irregular behavior while fishing, or a "yacht (sailing ship)" or "pleasure boat" that is easily carried away by the wind. The proximity state determination unit 102 performs control to dynamically change the proximity conditions (DCPA and TCPA thresholds, i.e., avoidance margins) for determining collision risk according to the identified ship type.

[0087] Specifically, if the target object is identified as a "fishing vessel" or "small pleasure boat" that is likely to exhibit irregular behavior, the navigation control unit 103 expands the threshold for approach conditions (for example, from 0.5 nautical miles to 1.0 nautical miles) compared to avoiding a merchant vessel, and initiates the aforementioned "clear turning maneuver" at an earlier stage. Furthermore, in preparation for unpredictable changes in the other vessel's course, it recalculates the target trajectory to maintain a wider clearance than usual even during the hold period after turning. This reduces the risk of accidents that depend on the characteristics of the other vessel and enables safer autonomous navigation that is more in line with the actual conditions of the sea.

[0088] (Dynamic leader transfer and degradation control during communication interruption) In coordinated navigation by a swarm of multiple vessels, as in the second embodiment, it is conceivable that communication between the central control unit 200 (land base station or leader vessel) and some of the unmanned vessels 10 may be suddenly interrupted (lost) due to radio interference or topographical factors (such as island shadows). In such a case, an isolated vessel that can no longer receive control commands from the overall operation setting unit of the swarm will immediately activate a local operation setting function within its own control unit 100 and seamlessly transition to an autonomous, distributed, degenerate mode that relies solely on short-range communication (such as VHF or ad-hoc Wi-Fi) with other vessels in the vicinity.

[0089] During the transition to this degenerate mode, isolated unmanned vessels dynamically decide whether to act as a temporary "sub-leader" or a "follower" that tracks other communicable vessels, based on pre-shared target trajectory and formation information for the entire swarm. If communication is lost during evasive maneuvers, each vessel's navigation control unit switches to a fail-safe algorithm that prioritizes collision avoidance (ensuring local safety) over maintaining the swarm's formation, and continues to avoid large rudder angles based on individual sensor information. This makes it possible to avoid a fatal system failure of the entire swarm due to vulnerabilities in the communication infrastructure.

[0090] (Dynamic reconstruction of group formation after evasion maneuvers are completed) Once the unmanned vessel group has completed the aforementioned coordinated avoidance maneuver (distribution of steering amount and speed on the inside and outside of the turn) and the conditions for passing the object to be avoided (50) have been met, the group will return to its original destination. At this time, the operation setting unit may not simply return to the formation before the avoidance (for example, a single line or a V-formation), but may comprehensively evaluate the current sea conditions, the size of the remaining navigable water area, and the remaining distance to the destination, and dynamically reconfigure the group's formation into an optimal new shape.

[0091] For example, if a group of unmanned vessels that were in a V-formation in the open ocean are pushed to the vicinity of a narrow coastal channel or reef area as a result of avoiding a large ship, the operation setting unit will determine that the safe navigable width has narrowed. Using this change in environment as a trigger, the operation setting unit calculates the return trajectory of each vessel to transition the formation of the entire group from a V-formation to a "line-ahead" formation suitable for follow navigation. In this way, by extending the avoidance operation beyond simply avoiding obstacles to include environmentally adaptive formation reconstruction immediately after avoidance, the safety of group control and mission continuity are further enhanced.

[0092] (Synchronization of visual and auditory intent communication linked to physical behavior) The system may be configured to automatically activate external notification equipment (searchlight, directional LED display, whistle, or directional speaker, etc.) mounted on the unmanned vessel 10 in perfect synchronization with the execution of a "turning operation larger than a predetermined period before the start operation" by the navigation control unit 103. For example, at night or in dense fog, it may be difficult to instantly visually confirm the wake generated by a large rudder angle turn of the unmanned vessel using only the radar of other vessels. In such circumstances, the intention information transmission unit 104 triggers the transmission of hardware-based visual and auditory signals at the same time as the command to start a physical turning operation is output.

[0093] Specifically, at the moment a large evasive turn to starboard begins, a powerful flashing light mounted on the mast flashes in a specific pattern, or an arrow (animation) indicating the direction of evasion is displayed on a high-brightness LED display mounted on the side of the bow. Furthermore, a predetermined maneuvering signal (e.g., a short warning sound) is automatically sounded via the electronic whistle. In this way, by completely linking and synchronizing the "physical and dramatic behavioral changes" of the hull with "visual and auditory signals" within the system, it becomes possible to communicate intentions strongly that directly appeal to the five senses of humans, maximizing the psychological sense of security given to other vessels.

[0094] (Evasion action selection logic based on remaining energy) If the unmanned vessel 10 is a battery-powered electric vessel, or if there is a limit to the fuel capacity, the navigation control unit 103 may be equipped with a function to constantly monitor the vessel's energy level (SoC: State of Charge) and dynamically select the means of evasive action (prioritizing changing course or prioritizing changing speed) based on that level. Generally, turning maneuvers that involve taking a large rudder angle while maintaining high speed increase the load on the propeller and consume a large amount of energy. Similarly, sudden reverse (crash astern) also requires a large amount of power.

[0095] If the battery level falls below a predetermined warning threshold and an object to be avoided (50) is encountered, the navigation control unit (103) selects an energy-saving avoidance trajectory that avoids energy-intensive large rudder angle turns and sudden stops, and instead safely passes the stern of the other vessel by only a slight reduction in power output (gradual deceleration) as early as possible. In the case of a group of multiple vessels, the operation setting unit can integrate group energy management control and avoidance control to maximize the cruising range and survivability of the swarm as a whole, such as by instructing the formation to rearrange positions in advance so that the vessel with the least remaining energy in the group is positioned on the inside of the turn (shortening the distance traveled).

[0096] (Updating the learning model of avoidance data through edge-cloud integration) The unmanned vessel 10 may also be equipped with a function to package various data (raw sensor data, its own behavior log, the trajectory of the other vessel, wave conditions, etc.) from a predetermined period before and after performing an avoidance maneuver as an "avoidance event dataset" and upload it to a cloud server at a land base station via the communication unit 130. The machine learning-based route optimization algorithm used during normal navigation is continuously retrained (fine-tuned) on the cloud side using data from such "edge cases (near misses) in which large rudder angles were actually required to avoid obstacles."

[0097] New inference models and control parameters updated on the cloud server are periodically delivered to each unmanned vessel 10 via OTA (Over The Air). This allows the entire system to accumulate empirical knowledge about "under what circumstances large mechanical rudder angle interventions are necessary," and in the long term, the optimization model for normal navigation itself will evolve (through federative learning, etc.) to generate safer and more marginal trajectories before emergency hard control is activated. This data pipeline linking edge and cloud will be a powerful means of ensuring the continuous performance improvement of the autonomous navigation system.

[0098] (System implementation hardware and software configuration) Each functional block of the control device 100 and the central control device 200 described above may be implemented as hardware using dedicated logic circuits (ASIC, FPGA, etc.), or it may be implemented in software by a processor such as a CPU (Central Processing Unit) or GPU (Graphics Processing Unit) that loads an autonomous navigation program stored in ROM or flash memory into a work area such as RAM and executes it. There may be one or more processors, and some heavy computational processing (such as image recognition) may be offloaded to a dedicated accelerator chip.

[0099] The autonomous navigation program consists of a set of instructions for the computer to execute the various navigation control methods described above. This program is provided by being recorded on a non-transient, computer-readable storage medium (e.g., solid-state drive (SSD), EEPROM, optical disc, etc.) installed when the unmanned vessel is shipped, or it may be provided in a format that can be downloaded via a communication network, such as via OTA. Software implementation has the advantage of allowing for flexible and rapid system updates without hardware changes in response to future revisions of maritime laws and the development of new international standard protocols.

[0100] In this specification, "unmanned vessels (ASV / USV)" are not limited to vessels that navigate on the surface of the sea. For example, the technical concept of "clear communication of intent through clear behavior" of the present invention can be directly applied to small autonomous boats that monitor water quality in lakes and dams, surface drones that inspect infrastructure and collect garbage in rivers, and even to hybrid autonomous underwater vehicles (AUVs) that navigate in a semi-submersible state and surface as needed, even when they are navigating near the water surface.

[0101] Furthermore, this system's avoidance targets are not limited to physical vessels and obstacles. For example, it can acquire "virtual geofences" (no-entry areas) such as marine protected areas (MPAs) or fishing rights areas where fixed nets are installed from electronic chart data, and configure the system to perform a clear turning maneuver to leave the area when approaching such a geofence, similar to avoidance maneuvers for other vessels. This allows unmanned vessel operators to clearly demonstrate their compliance with legally restricted airspace and sea areas to external observers through a physical track.

[0102] Furthermore, when unmanned vessels are used for special purposes such as mine sweeping missions or suspicious vessel surveillance missions, the cooperative control algorithm of the present invention can also be applied to "encirclement formations," where multiple vessels are positioned to surround a suspicious vessel, and "coverage formations," which comprehensively scan a specific sea area. Even during these missions, in the event of an unexpected encounter with a civilian vessel, it is expected that the system can be developed into advanced tactical control, such as temporarily releasing the encirclement state as a group, applying the steering amount distribution between the internal and external vessels as described in this embodiment to safely retreat, and then quickly resuming the mission.

[0103] As described above, the present invention is not merely an optimization algorithm, but an architecture that defines fundamental "behavior" rules for unmanned vessels to be accepted into human (manned vessel) social systems in diverse environments. The elemental technologies disclosed in each embodiment and modification (such as forcing large rudder angles, group steering distribution, linkage with object recognition, and formation reconstruction) may be applied individually or in any combination, making it possible to construct countless system variations according to the operational purpose and the resources of the onboard platform.

[0104] According to the autonomous navigation system, navigation control method, and autonomous navigation program of the present invention, unclear avoidance trajectories caused by the characteristics of machine learning can be eliminated, and even an unmanned platform without a crew can early demonstrate its avoidance intention to surrounding vessels as "objective and clear physical behavior." This dramatically reduces the risk of unexpected intersections and collisions in the sea and enables both safe formation maintenance and intention communication in the coordinated control of multiple vessels. [Explanation of Symbols]

[0105] 1. Autonomous Navigation System 10 Unmanned boat (ASV / USV) 10A inner boat 10B Unmanned boat (intermediate) 10C outer boat 10G Multiple boat swarm 50 Objects to avoid (other ships, etc.) 100 Control device (information processing device) 101 Information acquisition section (acquisition section) 102 Proximity Status Determination Unit 103 Navigation Control Unit 104 Intent Information Transmission Unit 110 Sensor section (camera, radar, LiDAR, IMU, etc.) 120 Actuator section (propulsion mechanism, steering mechanism) 130 Communications Department 140 Satellite 200 Integrated Control System 200A Operation Setting Unit 200B Avoidance Scenario Planning and Allocation Processing Unit 210A Other vessels (an example of an object to avoid) 210B Other ships (an example of an object to avoid) 301 Status Information Message 303 Intent Information Message 304 response message 305 Display update / warning notification (or display control) 401 Display device Display status before receiving 402A (other ship 210A) Display status before receiving 402B (other ship 210B) Display status after receiving 403A (other ship 210A) Display status after receiving 403B (other ship 210B) 404A Warning / emphasis display status (Other vessel 210A) 404B Warning / emphasis display status (Other vessel 210B)

Claims

1. An autonomous navigation system for controlling the navigation of an unmanned vessel performing autonomous navigation, An acquisition unit that acquires information regarding objects to be avoided that exist around the unmanned vessel and information regarding the size of the navigable water area around the unmanned vessel, The navigation control unit controls the propulsion mechanism and steering mechanism of the unmanned vessel to perform an avoidance maneuver when the proximity state between the unmanned vessel and the object to be avoided, determined based on the information acquired by the acquisition unit, satisfies predetermined proximity conditions, The navigation control unit sets an avoidance target course to resolve the proximity condition at the start of the avoidance operation, and if the width of the navigable water area is greater than or equal to a predetermined reference value, it controls the unmanned vessel to perform a turning operation from its current course to the avoidance target course while maintaining a speed greater than or equal to a lower limit speed set based on the speed of the unmanned vessel before the start of the avoidance operation. Autonomous navigation system.

2. The unmanned vessel is controlled to include a period of time after the turning motion or change in speed, during which the unmanned vessel's course is restricted from being changed toward the object to be avoided until the relative position of the object to be avoided with respect to the unmanned vessel satisfies the predetermined conditions for completion of passage. The autonomous navigation system according to claim 1.

3. If the estimated time or relative distance for the unmanned vessel to reach the object to be avoided is shorter than the threshold in the approach condition, the navigation control unit will perform a speed change in the avoidance operation to reduce the speed of the unmanned vessel to below the lower limit speed. The autonomous navigation system according to claim 1.

4. The conditions for completing the passage include at least one of the following: the relative distance of the object to be avoided as seen from the unmanned vessel begins to increase, or the relative azimuth angle of the object to be avoided with respect to the direction of travel of the unmanned vessel exceeds a predetermined angular threshold. The autonomous navigation system according to claim 2.

5. The navigation control unit controls the unmanned vessel based on the target trajectory calculated within its own device or the target trajectory obtained from an external device via communication. During normal navigation when the approach to the object to be avoided does not satisfy the approach conditions, the navigation of the unmanned vessel along the target trajectory, including gradual changes in course, is permitted. At the start of the avoidance operation, the unmanned vessel is controlled based on the target trajectory, which transitions to the avoidance target course through a series of consecutive turning maneuvers, by restricting navigation along the target trajectory, including the stepwise course changes. The autonomous navigation system according to claim 1.

6. If, during a period in which the navigation control unit restricts the unmanned vessel from changing its course toward the object to be avoided, a new approach condition between the unmanned vessel and a third vessel or obstacle different from the object to be avoided meets predetermined conditions, the control unit releases or relaxes the restriction on changing the course toward the object to be avoided, and causes the unmanned vessel to perform a secondary evasive action to resolve the new approach condition. The autonomous navigation system according to claim 2.

7. The autonomous navigation system further comprises an intent information transmission unit that transmits intent information to an external source based on either the actual trajectory, trajectory history, or predicted trajectory of the unmanned vessel. The navigation control unit causes the intention information, updated based on the avoidance target course, to be transmitted from the intention information transmission unit immediately before or during the execution of the series of consecutive turning maneuvers. The autonomous navigation system according to claim 1.

8. The acquisition unit acquires the range of deviation in the unmanned vessel's course caused by disturbances including at least one of waves, wind, or currents, and the navigation control unit, when performing the avoidance operation under the influence of the disturbance, sets the amount of course change or speed change in the turning operation to be larger the greater the deviation, and performs the control accordingly. The autonomous navigation system according to claim 1.

9. An autonomous navigation system for controlling the navigation of a group of unmanned vessels, including multiple unmanned vessels performing autonomous navigation, A navigation control unit that controls the propulsion mechanism and steering mechanism of each unmanned vessel belonging to the aforementioned group of unmanned vessels, The system includes an action setting unit that sets actions for multiple unmanned vessels belonging to the aforementioned group of unmanned vessels, including steering amounts and / or speed change amounts in avoidance maneuvers. The operation setting unit sets an operation for the multiple unmanned vessels to perform an avoidance operation in substantially the same direction when at least one of the unmanned vessels in the group starts an avoidance operation based on its proximity to an object to be avoided. In the aforementioned avoidance maneuver in substantially the same direction, the operation of the plurality of unmanned boats is set such that the amount of steering or the amount of speed change in the avoidance maneuver differs between the inner boat located on the inside of the turn and the outer boat located on the outside of the turn. The navigation control unit causes at least one of the plurality of unmanned vessels based on the set operation to perform a change of course or change of speed from the current course as a turning operation or change of speed that is greater than the amount of steering or change of speed observed for at least one of the unmanned vessels during a predetermined period before the start of the operation, and to perform the turning operation while maintaining a speed equal to or greater than the lower limit speed set based on the speed of at least one of the unmanned vessels before the start of the avoidance operation. Autonomous navigation system.

10. The operation setting unit sets the speed of the outer boat higher than the speed of the inner boat in order to maintain the relative positional relationship between the plurality of unmanned boats within a predetermined allowable range during the execution of the avoidance operation. The autonomous navigation system according to claim 9.

11. The operation setting unit sets the steering amount of the inner boat to be greater than the steering amount of the outer boat in order to maintain the relative positional relationship between the plurality of unmanned boats within a predetermined allowable range during the execution of the avoidance operation. The autonomous navigation system according to claim 9.

12. The aforementioned operation setting unit controls the boats to perform the avoidance maneuver while maintaining the relative distance, relative speed, or relative angle between multiple boats within a predetermined range. The autonomous navigation system according to claim 9.

13. The operation setting unit sets an operation to cancel the avoidance operation for the multiple unmanned vessels in substantially the same direction when the relative positional relationship between the closest unmanned vessel among the multiple unmanned vessels that is closest to the object to be avoided and the object to be avoided satisfies a predetermined passing completion condition indicating that the object to be avoided has been passed. The autonomous navigation system according to claim 9.

14. The conditions for completing the passage include at least one of the following: the relative distance of the object to be avoided as seen from the nearest unmanned vessel begins to increase, or the relative azimuth angle of the object to be avoided, with respect to the direction of travel of the nearest unmanned vessel, exceeds a predetermined angular threshold. The autonomous navigation system according to claim 13.

15. A computer-based method for controlling the navigation of an autonomously navigating unmanned vessel, The steps include obtaining information about objects to be avoided that exist around the unmanned vessel and information about the size of the navigable water area around the unmanned vessel, The step includes controlling the propulsion mechanism and steering mechanism of the unmanned vessel to perform an avoidance maneuver when the proximity state between the unmanned vessel and the object to be avoided, as determined based on the acquired information, satisfies predetermined proximity conditions, In the step of executing the avoidance operation, at the start of the avoidance operation, a target avoidance course to resolve the approaching state is set, and if the width of the navigable water area is greater than or equal to a predetermined reference value, the unmanned vessel is controlled to perform a turning operation from its current course to the target avoidance course, while maintaining a speed greater than or equal to a lower limit speed set based on the speed of the unmanned vessel before the start of the avoidance operation. Navigation control method.

16. A method for controlling the navigation of a group of unmanned vessels, including multiple unmanned vessels that perform autonomous navigation, which is executed by a computer, For multiple unmanned vessels belonging to the aforementioned group of unmanned vessels, the operation setting step involves setting operations for multiple vessels, including steering amount and / or speed change amount in avoidance maneuvers. This includes a navigation control step that controls the propulsion mechanism and steering mechanism of each unmanned vessel belonging to the aforementioned group of unmanned vessels, In the operation setting step, an operation is set so that at least one of the unmanned vessels in the group of unmanned vessels performs an avoidance operation in substantially the same direction based on its proximity to the object to be avoided. In the aforementioned avoidance maneuvers in substantially the same direction, the operations of the plurality of unmanned boats are set such that the amount of steering or the amount of speed change in the avoidance maneuvers differs between the inner boat located on the inside of the turn and the outer boat located on the outside of the turn. In the navigation control step, control is performed on at least one of the plurality of unmanned vessels based on the set operation to change course or speed from the current course by a turning operation or speed change that is greater than the amount of steering or speed change observed for at least one of the unmanned vessels during a predetermined period before the start operation, and the turning operation is performed while maintaining a speed equal to or greater than the lower limit speed set based on the speed of at least one of the unmanned vessels before the start of the avoidance operation. Navigation control method.

17. An autonomous navigation program for causing a computer to function as a component of an autonomous navigation system that controls the navigation of an unmanned vessel performing autonomous navigation, The computer is configured as an acquisition unit to acquire information about objects to be avoided that exist around the unmanned vessel and information about the size of the navigable water area around the unmanned vessel. Based on the information acquired by the acquisition unit, if the proximity state between the unmanned vessel and the object to be avoided, as determined, satisfies predetermined proximity conditions, the navigation control unit functions to control the propulsion mechanism and steering mechanism of the unmanned vessel to perform an avoidance maneuver. The computer, which functions as the navigation control unit, is instructed to set an avoidance target course to resolve the approaching state at the start of the avoidance operation, and if the width of the navigable water area is greater than or equal to a predetermined reference value, the unmanned vessel is instructed to maintain a speed greater than or equal to a lower limit speed set based on the speed of the unmanned vessel before the start of the avoidance operation, and to perform a turning operation from the unmanned vessel's current course to the avoidance target course. Autonomous navigation program.

18. An autonomous navigation program for causing a computer to function as a component of an autonomous navigation system that controls the navigation of a group of unmanned vessels, including multiple unmanned vessels performing autonomous navigation, The aforementioned computer, A navigation control unit that controls the propulsion mechanism and steering mechanism of each unmanned vessel belonging to the aforementioned group of unmanned vessels. For multiple unmanned vessels belonging to the aforementioned group of unmanned vessels, the unit functions as an action setting unit that sets actions for multiple vessels, including steering amount and / or speed change amount in evasive maneuvers. The computer, which functions as the operation setting unit, is instructed to set an operation for the multiple unmanned vessels to perform an avoidance operation in substantially the same direction when at least one of the unmanned vessels in the group of unmanned vessels starts an avoidance operation based on its proximity to an object to be avoided. In the aforementioned avoidance maneuvers in substantially the same direction, the operations of the plurality of unmanned boats are set such that the amount of steering or the amount of speed change in the avoidance maneuvers differs between the inner boat located on the inside of the turn and the outer boat located on the outside of the turn. The computer, which functions as the navigation control unit, is instructed to control at least one of the plurality of unmanned vessels based on the set operation to change course or speed from its current course by a turning operation or speed change that is greater than the amount of steering or speed change observed for at least one of the unmanned vessels during a predetermined period before the start of the operation, and to perform the turning operation while maintaining a speed equal to or greater than the lower limit speed set based on the speed of at least one of the unmanned vessels before the start of the avoidance operation. Autonomous navigation program.

Images