Unmanned ship path planning and collision avoidance decision-making method and system based on multi-factor fusion

Abstract

The invention discloses an unmanned ship path planning and collision avoidance decision-making method and system based on multi-factor fusion. Comprising the steps that S1, an unmanned ship multi-sensor fusion sensing network is constructed, environment data are collected in real time, mechanical parameters at the collision moment are captured on an unmanned ship, and the unmanned ship conducts sensing recognition on collision severity; s2, defining a path tracking excitation function, and adaptively adjusting dynamic parameters required in a path control process; s3, setting an excitation mechanism for collision avoidance of the static obstacle and the dynamic obstacle, and evaluating the threat degree of the static obstacle and the dynamic obstacle; and S4, integrating the path tracking response, the static obstacle collision avoidance excitation mechanism and the dynamic obstacle collision avoidance excitation mechanism, so that the unmanned ship balances the relationship among different task targets, and performs path planning and collision avoidance decision under multiple tasks. According to the technical scheme provided by the invention, the problems of path planning and obstacle avoidance decision making of the unmanned ship are effectively solved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous navigation control and path planning of unmanned surface ships, and in particular to an unmanned ship path planning and collision avoidance decision method and system based on multi-factor fusion. Background Art

[0002] With the rapid development of science and technology, unmanned ships are increasingly widely used in the marine field. In the military field, unmanned ships can perform reconnaissance, surveillance, anti-submarine, anti-mine and other tasks, and can replace manned ships in dangerous waters to perform tasks, reduce the risk of casualties, and improve combat effectiveness and strategic deterrence; in the civilian field, unmanned ships have shown great potential in marine resource exploration, environmental monitoring, maritime rescue, and shipping assistance; in maritime rescue missions, unmanned ships can quickly reach the scene of the accident, deliver rescue supplies or provide information support to rescuers, and improve rescue efficiency.

[0003] Reinforcement learning is an important technical means to achieve autonomous navigation of unmanned ships. Unmanned ships can continuously learn and optimize their own behavior strategies based on environmental feedback to achieve specific goals. However, in practical applications, the reinforcement learning process faces many problems. For example, it is not adaptable enough to complex marine environments. The marine environment is highly complex and dynamic, including changeable sea conditions (such as waves, currents, wind direction and speed, etc.), different geographical areas (such as ports, narrow waterways, open seas, etc.) and various types of obstacles (static islands, reefs, buoys and dynamic other ships, etc.). Traditional reinforcement learning methods often use fixed parameters and simple calculation modes, which are difficult to adapt to such complex and changing environments; it is difficult to coordinate multiple task goals. In actual operation, unmanned ships usually need to take into account multiple task goals at the same time, such as accurate path tracking, efficient navigation speed, safe obstacle avoidance, and satisfaction of task-specific requirements (such as rapid response in rescue missions, data accuracy in environmental monitoring tasks, etc.). It is difficult for traditional methods to find a reasonable balance between these interrelated and potentially conflicting goals. For example, when pursuing rapid path tracking, potential obstacle risks may be ignored, or when over-emphasizing collision avoidance, navigation efficiency may be greatly reduced, and the task cannot be completed on time; lack of dynamic adjustment capabilities, the environment and task requirements faced by unmanned ships during navigation are constantly changing, and traditional methods cannot respond to these changes in a timely manner. For example, when encountering sudden severe weather or temporary high-priority tasks, traditional methods cannot automatically adjust the incentive calculation strategy, making it impossible for the unmanned ship to respond flexibly, which may lead to mission failure or threaten navigation safety. Summary of the invention

[0004] In view of the defects of related technologies, the present invention provides an unmanned ship path planning and collision avoidance decision-making method and system based on multi-factor fusion, which improves the autonomous navigation performance of the unmanned ship, enables it to efficiently and safely perform diversified tasks in a complex and changeable marine environment, and improves and optimizes the traditional reinforcement learning process to overcome the limitations of the existing technology.

[0005] The technical solution is as follows:

[0006] A method for unmanned ship path planning and collision avoidance decision-making based on multi-factor fusion, comprising the following steps:

[0007] Step S1: construct a multi-sensor fusion perception network for unmanned ships, integrate GPS, lidar, sonar and multi-spectral visual sensors, collect environmental data in real time, and deploy high-sensitivity piezoelectric impact sensors and high-precision optical collision angle measuring instruments in key collision areas of the unmanned ship shell to capture the mechanical parameters at the moment of collision, so that the unmanned ship can perceive and identify the severity of the collision;

[0008] Step S2: setting an incentive mechanism for the unmanned ship path tracking response, simulating the navigation process of the unmanned ship under various path deviation conditions, defining the path tracking incentive function, and adaptively adjusting the dynamic parameters required in the path control process;

[0009] Step S3: setting incentive mechanisms for static obstacle avoidance and dynamic obstacle avoidance, using regression analysis and machine learning to determine incentive functions for static obstacle avoidance and dynamic obstacle avoidance, and evaluating the threat levels of static obstacles and dynamic obstacles by continuously optimizing the sensor data processing algorithm and incentive function calculation model;

[0010] Step S4: Integrate the path tracking response, static obstacle and dynamic obstacle avoidance incentive mechanisms, and fuse their respective incentive functions into an organic whole, so that the unmanned ship can balance the relationship between different mission objectives and make path planning and collision avoidance decisions under multiple tasks.

[0011] An unmanned ship path planning and collision avoidance decision system based on multi-factor fusion, the system is used to implement the unmanned ship path planning and collision avoidance decision method based on multi-factor fusion, comprising:

[0012] The unmanned ship multi-sensor fusion perception network module is used to integrate GPS, laser radar, sonar and multi-spectral visual sensors to collect environmental data in real time, and deploy high-sensitivity piezoelectric impact sensors and high-precision optical collision angle measuring instruments in the key collision area of ​​the unmanned ship's hull to capture the mechanical parameters at the moment of collision, so that the unmanned ship can perceive and identify the severity of the collision;

[0013] The unmanned ship path tracking response excitation module is used to simulate the navigation process of the unmanned ship under various path deviation conditions, define the path tracking excitation function, and adaptively adjust the dynamic parameters required in the path control process;

[0014] The unmanned ship static obstacle and dynamic obstacle collision avoidance incentive module is used to determine the static obstacle collision avoidance incentive function by using regression analysis and machine learning, and to evaluate the threat level of static and dynamic obstacles by continuously optimizing the sensor data processing algorithm and incentive function calculation model;

[0015] The unmanned ship incentive integration module is used to integrate path tracking response, static obstacle and dynamic obstacle avoidance incentive mechanisms, and fuse the respective incentive functions into an organic whole, so that the unmanned ship can balance the relationship between different mission objectives and make path planning and collision avoidance decisions under multiple tasks.

[0016] The present invention has the following beneficial effects:

[0017] The present invention provides an unmanned ship path planning and collision avoidance decision-making method and system based on multi-factor fusion. Taking into account the game situation of executing path planning and obstacle avoidance in the process of unmanned ship driving, it is creatively proposed to integrate path tracking response, static obstacle and dynamic obstacle collision avoidance incentive mechanism into an organic whole. In this process, the path tracking incentive is deeply studied, and multivariate functions and dynamic adaptation parameters are integrated; static obstacle collision avoidance incentive is optimized, and distance speed perception and multidimensional environmental factors are integrated; dynamic obstacle collision avoidance incentive is advanced in all directions, COLREGs rules are deeply integrated and dynamic parameters and situation perception factors are intelligently adjusted; total incentive calculation is systematically optimized, multiple factors are integrated and task-oriented adaptive parameter tuning is realized, and the innovative synergy of each step significantly improves the decision rationality, navigation safety and efficiency, collision avoidance success rate, task adaptability and overall task effectiveness of unmanned ships in multi-task execution in complex environments, and comprehensively enhances its comprehensive performance and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of an unmanned ship path planning and collision avoidance decision method based on multi-factor fusion provided by an embodiment of the present invention;

[0019] Figure 2 is a flow chart of the unmanned ship sensing and identifying the severity of a collision provided by an embodiment of the present invention;

[0020] Figure 3 is a flow chart of setting an incentive mechanism for unmanned ship path tracking response provided by an embodiment of the present invention;

[0021] Figure 4It is a flow chart of the incentive mechanism for the unmanned ship to avoid static obstacles and dynamic obstacles provided by an embodiment of the present invention;

[0022] Figure 5 It is a structural diagram of an unmanned ship path planning and collision avoidance decision system based on multi-factor fusion provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0024] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0025] See also Figure 1 The present invention provides an unmanned ship path planning and collision avoidance decision method based on multi-factor fusion, comprising the following steps:

[0026] Step S1: construct a multi-sensor fusion perception network for unmanned ships, integrate GPS, lidar, sonar and multi-spectral visual sensors, collect environmental data in real time, and deploy high-sensitivity piezoelectric impact sensors and high-precision optical collision angle measuring instruments in key collision areas of the unmanned ship shell to capture the mechanical parameters at the moment of collision, so that the unmanned ship can perceive and identify the severity of the collision;

[0027] Step S2: setting an incentive mechanism for the unmanned ship path tracking response, simulating the navigation process of the unmanned ship under various path deviation conditions, defining the path tracking incentive function, and adaptively adjusting the dynamic parameters required in the path control process;

[0028] Step S3: setting incentive mechanisms for static obstacle avoidance and dynamic obstacle avoidance, using regression analysis and machine learning to determine incentive functions for static obstacle avoidance and dynamic obstacle avoidance, and evaluating the threat levels of static obstacles and dynamic obstacles by continuously optimizing the sensor data processing algorithm and incentive function calculation model;

[0029] Step S4: Integrate the path tracking response, static obstacle and dynamic obstacle avoidance incentive mechanisms, and fuse their respective incentive functions into an organic whole, so that the unmanned ship can balance the relationship between different mission objectives and make path planning and collision avoidance decisions under multiple tasks.

[0030] See also Figure 2 ,Step S1: Build a multi-sensor fusion perception network for unmanned ships, integrate GPS, lidar, sonar and multi-spectral visual sensors, collect environmental data in real time, and deploy high-sensitivity piezoelectric impact sensors and high-precision optical collision angle measuring instruments in the key collision area of ​​the unmanned ship shell to capture the mechanical parameters at the moment of collision, so that the unmanned ship can perceive and identify the severity of the collision, specifically:

[0031] Step S11: deploy a high-precision global positioning system, laser radar, sonar and visual camera multi-type sensors on the unmanned ship to build a full-range environmental perception system; use GPS to obtain the precise geographical location information of the ship, laser radar and sonar to scan the surrounding environment in real time, detect the location, distance and shape of obstacles, and the visual camera provides intuitive image data for auxiliary judgment; compare and analyze these sensor data with pre-stored different scene feature data, and use pattern recognition algorithms to accurately identify the current environmental scene of the unmanned ship;

[0032] Step S12: Install a collision detection device consisting of a high-sensitivity combined collision sensor based on a pressure sensor and an accelerometer at the key parts of the unmanned ship to accurately measure the magnitude, direction and duration of the impact force at the moment of collision, and then calculate the collision angle θ collision and the collision relative velocity v collision Based on a large amount of actual collision test data, a weighted function ω(θ collision ), and define the collision excitation function Enable the unmanned ship to perceive and identify the severity of the collision.

[0033] In an embodiment provided by the present invention, for a head-on collision, a larger weighting value (such as 10) is assigned due to its extremely high risk of damage to the ship structure and equipment; for a side collision, a relatively smaller weighting value (such as 5) is assigned due to its relatively small degree of damage; the calculated collision angle and relative speed are substituted into the collision excitation function to obtain an accurate collision excitation value, providing a strong negative feedback signal for the intelligent agent, prompting it to avoid similar high-risk behaviors in subsequent decisions.

[0034] See also Figure 3 ,Step S2: Set an incentive mechanism for the unmanned ship path tracking response, simulate the navigation process of the unmanned ship under various path deviation conditions, define the path tracking incentive function, and adaptively adjust the dynamic parameters required in the path control process, specifically:

[0035] Step S21: Obtain the lateral error ε of the unmanned ship in real time through the inertial navigation system, GPS and other sensors on the unmanned ship (t) , speed u (t) and heading information;

[0036] Step S22: Combine the exponential activation function exp(-γ ε |ε (t) |) and Gaussian activation function The excitation characteristics within different path error ranges are constructed to simulate the navigation process of the unmanned ship under various path deviation conditions. The influence of the two excitation functions on the path adjustment behavior of the unmanned ship is recorded and analyzed, and the weight α is set for optimization. The α value combination that can make the fusion performance of the two excitation functions optimal within different error intervals is found. At the same time, the adjustment function κ(ε (t) ), dynamically adjust the intensity of the speed and heading excitation terms according to the size of the lateral error, and define the path tracking excitation function as:

[0037]

[0038] where γ ε is the weight coefficient of the lateral error and the path fit, which is used to adjust the control strength; U max is the maximum drivable speed of the unmanned ship, and the dynamic parameters required in the path control process are adaptively adjusted.

[0039] In one embodiment provided by the present invention, the combined effect of the exponential excitation function and the Gaussian excitation has an important effect on the continuously changing excitation value based on the distance (such as the yaw error); it can encourage the unmanned ship to gradually converge to the desired state, and according to this continuously changing characteristic, combined with the speed and heading factors introduced by itself, the excitation function changes more smoothly and reasonably under different path deviations and ship motion states. For example, in the process of the unmanned ship approaching the precise path from a large path deviation, according to the gradual characteristics of the combination of the exponential excitation function and the Gaussian excitation, the weight of the excitation term based on speed and heading is better adjusted, so that the ship can quickly approach the path and maintain a stable motion state during the adjustment process.

[0040] In one embodiment of the present invention, when the lateral error sensor detects |ε (t) When |≥5, it indicates that the path deviation is large. At this time, the preset parameter adjustment algorithm is started, and the increased γ ε For example, when the initial value 0.5 is increased to 0.8, the correction of lateral errors is strengthened, prompting the unmanned ship to quickly adjust its heading and speed. (t) When |≤1, γ is automatically reduced to avoid path oscillation caused by over-adjustment. ε to 0.2, so that the intelligent agent can remain stable when approaching the path; according to the sea area information provided by the electronic chart system or the feedback of the terrain detection sonar, the type of sea area where the unmanned ship is located is judged. In open waters, in order to give the unmanned ship a larger exploration space and the ability to cope with environmental changes, σ is set to 10 in the Gaussian excitation; in narrow waterways or areas with high-precision path requirements, σ ​​is reduced to 2 to improve the accuracy requirements of path tracking; at the same time, the speed sensor monitors the ship speed in real time. When u (t) <0.5U max When u is , the ship speed is considered to be low. Through the preset speed incentive adjustment rules, the weight coefficient of the speed incentive item is appropriately increased (such as increasing by 0.2) to encourage the unmanned ship to increase its speed. (t) >0.9U max When the speed is too high, the weight of the speed incentive item is reduced (such as reducing it by 0.3) to prevent speeding from affecting the path tracking accuracy and ensure that the unmanned ship can maintain good path tracking performance under different speed conditions.

[0041] See also Figure 4 ,Step S3: Set the incentive mechanism for static obstacle and dynamic obstacle avoidance, use regression analysis and machine learning to determine the incentive functions for static obstacle avoidance and dynamic obstacle avoidance, and evaluate the threat level of static obstacles and dynamic obstacles by continuously optimizing the sensor data processing algorithm and the incentive function calculation model, specifically:

[0042] Step S31: Equip the unmanned ship with a high-precision laser rangefinder, Doppler radar rangefinder and speed sensor to accurately measure the distance change rate between the unmanned ship and the static obstacle in real time. Combined with the analysis of a large amount of actual navigation data, the weighting coefficient β is determined using regression analysis and machine learning algorithms. At the same time, the angle between the current heading of the unmanned ship and the direction of the obstacle is obtained through the ship heading sensor, steering angle sensor, water flow sensor and electronic compass equipment. Ship steering capability parameter ω turn , water velocity v current and direction θ current Information, these parameters together with the obstacle distance x and the angle θ between the unmanned ship and the obstacle, construct the static obstacle avoidance incentive function:

[0043]

[0044] where γ θ,stat It is a parameter related to the relative angle between the unmanned ship and the static obstacle. Through this parameter, the weight of the excitation calculation can be adjusted according to the relative angle. When the relative angle is in front, the value of this item will become smaller, thereby reducing the entire excitation value and enhancing the importance of collision avoidance. δ is a coefficient used to weight factors related to the position of static obstacles. γ x is the distance weighting index, β is the spacing change rate weighting coefficient, χ is the ship speed weighting coefficient, is the steering capability weighting coefficient.

[0045] In an embodiment provided by the present invention, when the direction of the water flow is consistent with the direction of the ship heading towards the obstacle (the angle with the ship's heading is small) and the water flow speed is large, the incentive item is appropriately increased through a preset water flow impact compensation algorithm to compensate for the impact of the water flow on the speed of the ship approaching the obstacle; when the ship's steering ability is limited (small), the incentive weight related to the steering ability is increased, prompting the unmanned ship to plan a more reasonable collision avoidance path in advance.

[0046] Step S32: Based on AIS data, radar image recognition and machine learning target classification technology, the situation of dynamic obstacle targets in the unmanned ship navigation area is mastered, and according to the COLREGs rules and the direction and speed of the unmanned ship, the weighted parameter γ of the dynamic obstacle close to the unmanned ship in the starboard area, port area and stern area is calculated. right , γ left and γ back Perform real-time optimization;

[0047] In an embodiment provided by the present invention, when a dynamic obstacle approaches the unmanned ship from the starboard side, the influence weight of the obstacle in the area on the activation function is adjusted according to factors such as distance; when the starboard side is close to the target ship, for example, less than 80 meters, γright The increase from 0.5 to 0.95 significantly improves the decision weight of collision avoidance in the starboard area, prompting the USV to take evasive action on the target ship approaching the starboard side first, which complies with the requirements of the COLREGs rules for starboard avoidance. When in a crossing encounter situation, for example, when the speed of the dynamic obstacle appearing on the port side is greater than 0.5 times the maximum speed and the distance to the unmanned ship is between 60 meters and 180 meters, the γ left , ensuring appropriate response to dynamic obstacles on the port side in crossing encounter scenarios, achieving a balance between safe collision avoidance and efficient navigation; when dynamic obstacles approach quickly from the stern, increasing γ back The value of is used to enhance the weight of obstacles in the stern direction in the incentive function, so that the unmanned ship can detect the danger of the ship coming from behind in time and take effective collision avoidance measures to prevent accidents such as rear-end collision.

[0048] Step S33: Define the dynamic collision avoidance excitation factor ξ(v target ,x,θ relative ,ω USV ) Improve the success rate of avoiding collision with dynamic obstacles, and define where v target is the speed of the dynamic obstacle, x is the relative distance between the unmanned ship and the dynamic obstacle, θ relative is the relative azimuth between the unmanned ship and the dynamic obstacle, ω USV It is the turning ability of the unmanned ship;

[0049] Step S34: Construct dynamic obstacle avoidance incentive function:

[0050]

[0051] Among them, γ θ,dyn is a parameter related to the relative angle of the dynamic obstacle. It can adjust the weight of this item in the excitation calculation according to the relative angle of the target ship. y is the velocity component of the dynamic obstacle in the direction of the unmanned ship, and ξ v (θ,v y ) is a velocity component weighting parameter based on the angle and velocity of the dynamic obstacle, which is specifically a piecewise function:

[0052]

[0053] η1, η2 and η3 are constant weighting factors, reflecting COLREGs compliance level C rules and The weight of .

[0054] Step S4: Integrate the path tracking response, static obstacle and dynamic obstacle avoidance incentive mechanisms, and fuse their respective incentive functions into an organic whole, so that the unmanned ship can balance the relationship between different mission objectives and make decisions on path planning and collision avoidance under multiple tasks, specifically:

[0055] Integrate the path tracking response, static obstacle and dynamic obstacle avoidance incentive mechanism, and merge their respective incentive functions into an organic whole. The overall incentive function is:

[0056]

[0057] where λ (t) is the weight associated with path tracking, which changes dynamically according to different path tracking factors. Its role is to determine the relative importance of path tracking incentives according to the current situation when calculating the total incentive, while 1-λ (t) It reflects the game between the path and the obstacle avoidance incentive; It includes additional incentives related to environmental stability, mission priority, etc. In terms of environmental stability, if the sea conditions are good and the water flow is stable, environmental factors that are conducive to navigation appear, The part related to environmental stability will be given a positive incentive; in terms of task priority, if the current task is an emergency rescue task and the behavior of the unmanned ship helps to improve the rescue efficiency, then a positive incentive will be given according to the task priority.

[0058] In an embodiment provided by the present invention, different scene feature data include typical features of ports, narrow waterways, open seas, etc. After the scene type is determined, the weighting coefficient is initialized to a corresponding value according to a preset scene and weighting coefficient mapping table. In a complex port environment or a narrow waterway scene, due to dense obstacles and limited space, λ (t) It is set to a lower value (such as 0.2) to highlight the importance of collision avoidance incentives. In open and obstacle-free sea scenarios, it will be set to a higher value (such as 0.8) to focus on path tracking incentives and provide basic guidance for subsequent decision-making.

[0059] In an embodiment provided by the present invention, when an unmanned ship is in an emergency rescue mission, the unmanned ship can adjust the strategy according to the mission characteristics and parameters, give priority to rapid arrival at the target and safe collision avoidance, and effectively improve the rescue efficiency; by accurately increasing the speed incentive weight and strengthening the incentive intensity of obstacles in key areas, the unmanned ship can significantly shorten the rescue time and increase the probability of successful rescue under the premise of ensuring safety, thereby winning precious time for saving lives and reducing accident losses, and greatly improving the execution efficiency of emergency rescue missions. For example, in a maritime rescue operation, the unmanned ship can quickly cross the open sea according to the adjusted incentive function, and when approaching the accident site, it can quickly reach the rescue site by relying on its high vigilance and effective collision avoidance of obstacles in key areas, thereby increasing the chance of survival of trapped people.

[0060] In an embodiment provided by the present invention, when the unmanned ship is in the material transportation task, it ensures the stable operation of the ship and the safety of the cargo during the entire transportation process, reduces the risk of cargo damage, and ensures the integrity of the material transportation. By reasonably reducing the speed incentive weight, increasing the incentive for smooth navigation, and increasing the incentive intensity that affects the safe operation of the cargo, the risk of ship shaking and cargo displacement caused by improper operation of the unmanned ship is effectively suppressed, the possibility of damage to the cargo during transportation is reduced, and the materials can be delivered to the destination intact and safely, providing a solid guarantee for the smooth completion of the material transportation task. For example, in the process of long-distance material transportation, the unmanned ship can maintain a stable speed and a small heading adjustment to avoid collision, damage or displacement of the cargo due to violent ship movement, and ensure that the quality and quantity of the materials are not affected.

[0061] In an embodiment provided by the present invention, when the unmanned ship is carrying out the ocean monitoring task, the unmanned ship can efficiently carry out monitoring work in a wide ocean area, reasonably balance the monitoring range with navigation safety, data quality and other factors; by expanding the scope of exploration area incentives, increasing the incentive weight of sensor data quality and incentives for poor navigation status, the unmanned ship is encouraged to actively explore more areas and maintain a good navigation state to obtain high-quality monitoring data, thereby significantly improving the effect and quality of ocean monitoring, providing more abundant and accurate data support for the fields of marine scientific research, environmental protection, resource exploration, etc., and effectively promoting the development and decision-making of related fields. For example, in the large-area ocean monitoring task, the unmanned ship can expand the monitoring range according to the optimized incentive function, while ensuring the accuracy of sensor data, and obtain more information on the marine environment, ecology and other aspects, providing a scientific basis for the rational development of marine resources and the formulation of environmental protection policies.

[0062] See also Figure 5 The present invention provides an unmanned ship path planning and collision avoidance decision system 100 based on multi-factor fusion, the system comprising:

[0063] The unmanned ship multi-sensor fusion perception network module 101 is used to integrate GPS, laser radar, sonar and multi-spectral visual sensors to collect environmental data in real time, and deploy high-sensitivity piezoelectric impact sensors and high-precision optical collision angle measuring instruments in the key collision area of ​​the unmanned ship shell to capture the mechanical parameters at the moment of collision, so that the unmanned ship can perceive and identify the severity of the collision;

[0064] The unmanned ship path tracking response excitation module 102 is used to simulate the navigation process of the unmanned ship under various path deviation conditions, define the path tracking excitation function, and adaptively adjust the dynamic parameters required in the path control process;

[0065] The unmanned ship static obstacle and dynamic obstacle collision avoidance incentive module 103 is used to determine the static obstacle collision avoidance incentive function by using regression analysis and machine learning, and to evaluate the threat level of static obstacles and dynamic obstacles by continuously optimizing the sensor data processing algorithm and the incentive function calculation model;

[0066] The unmanned ship incentive integration module 104 is used to integrate the path tracking response, static obstacle and dynamic obstacle avoidance incentive mechanisms, and merge the respective incentive functions into an organic whole, so that the unmanned ship can balance the relationship between different mission objectives and make path planning and collision avoidance decisions under multiple tasks.

Claims

1. A method for unmanned ship path planning and collision avoidance decision-making based on multi-factor fusion, characterized in that: The following steps are involved: Step S1: construct a multi-sensor fusion perception network for unmanned ships, integrate GPS, lidar, sonar and multi-spectral visual sensors, collect environmental data in real time, and deploy high-sensitivity piezoelectric impact sensors and high-precision optical collision angle measuring instruments in key collision areas of the unmanned ship shell to capture the mechanical parameters at the moment of collision, so that the unmanned ship can perceive and identify the severity of the collision; Step S2: setting an incentive mechanism for the unmanned ship path tracking response, simulating the navigation process of the unmanned ship under various path deviation conditions, defining the path tracking incentive function, and adaptively adjusting the dynamic parameters required in the path control process; Step S3: setting incentive mechanisms for static obstacle avoidance and dynamic obstacle avoidance, using regression analysis and machine learning to determine incentive functions for static obstacle avoidance and dynamic obstacle avoidance, and evaluating the threat levels of static obstacles and dynamic obstacles by continuously optimizing the sensor data processing algorithm and incentive function calculation model; Step S4: Integrate the path tracking response, static obstacle and dynamic obstacle avoidance incentive mechanisms, and fuse their respective incentive functions into an organic whole, so that the unmanned ship can balance the relationship between different mission objectives and make path planning and collision avoidance decisions under multiple tasks.

2. The unmanned ship path planning and collision avoidance decision method based on multi-factor fusion as claimed in claim 1, characterized in that: The step S1 specifically includes: Step S11: deploy a high-precision global positioning system, laser radar, sonar and visual camera multi-type sensors on the unmanned ship to build a full-range environmental perception system; use GPS to obtain the precise geographical location information of the ship, laser radar and sonar to scan the surrounding environment in real time, detect the location, distance and shape of obstacles, and the visual camera provides intuitive image data for auxiliary judgment; compare and analyze these sensor data with pre-stored different scene feature data, and use pattern recognition algorithms to accurately identify the current environmental scene of the unmanned ship; Step S12: Install a collision detection device consisting of a high-sensitivity combined collision sensor based on a pressure sensor and an accelerometer at the key parts of the unmanned ship to accurately measure the magnitude, direction and duration of the impact force at the moment of collision, and then calculate the collision angle θ collision and the collision relative velocity v collision Based on a large amount of actual collision test data, a weighted function ω(θ collision ), and define the collision excitation function Enable the unmanned ship to perceive and identify the severity of the collision.

3. The unmanned ship path planning and collision avoidance decision-making method based on multi-factor fusion as claimed in claim 1, characterized in that: The step S2 specifically includes: Step S21: Obtain the lateral error ε of the unmanned ship in real time through the inertial navigation system, GPS and other sensors on the unmanned ship (t) , speed u (t) and heading information; Step S22: Combine the exponential activation function exp(-γ ε |ε (t) |) and Gaussian activation function The excitation characteristics within different path error ranges are constructed to simulate the navigation process of the unmanned ship under various path deviation conditions. The influence of the two excitation functions on the path adjustment behavior of the unmanned ship is recorded and analyzed, and the weight α is set for optimization. The α value combination that can make the fusion performance of the two excitation functions optimal within different error intervals is found. At the same time, the adjustment function κ(ε (t) ), dynamically adjust the intensity of the speed and heading excitation terms according to the size of the lateral error, and define the path tracking excitation function as: where γ ε is the weight coefficient of the lateral error and the path fit, which is used to adjust the control strength; U max is the maximum drivable speed of the unmanned ship, and the dynamic parameters required in the path control process are adaptively adjusted.

4. The unmanned ship path planning and collision avoidance decision-making method based on multi-factor fusion as claimed in claim 1, characterized in that: The step S3 specifically includes: Step S31: Equip the unmanned ship with a high-precision laser rangefinder, Doppler radar rangefinder and speed sensor to accurately measure the distance change rate between the unmanned ship and the static obstacle in real time. Combined with the analysis of a large amount of actual navigation data, the weighting coefficient β is determined using regression analysis and machine learning algorithms. At the same time, the angle between the current heading of the unmanned ship and the direction of the obstacle is obtained through the ship heading sensor, steering angle sensor, water flow sensor and electronic compass equipment. Ship steering capability parameter ω turn , water velocity v current and direction θ current Information, these parameters together with the obstacle distance x and the angle θ between the unmanned ship and the obstacle, construct the static obstacle avoidance incentive function: where γ θ,stat It is a parameter related to the relative angle between the unmanned ship and the static obstacle. Through this parameter, the weight of the excitation calculation can be adjusted according to the relative angle. When the relative angle is in front, the value of this item will become smaller, thereby reducing the entire excitation value and enhancing the importance of collision avoidance. δ is a coefficient used to weight factors related to the position of static obstacles. γ x is the distance weighting index, β is the spacing change rate weighting coefficient, χ is the ship speed weighting coefficient, is the steering capability weighting coefficient; Step S32: Based on AIS data, radar image recognition and machine learning target classification technology, the situation of dynamic obstacle targets in the unmanned ship navigation area is mastered, and according to the COLREGs rules and the direction and speed of the unmanned ship, the weighted parameter γ of the dynamic obstacle close to the unmanned ship in the starboard area, port area and stern area is calculated. right , γ left and γ back Perform real-time optimization; Step S33: Define the dynamic collision avoidance excitation factor ξ(v target ,x,θ relative ,ω USV ) Improve the success rate of avoiding collision with dynamic obstacles, and define where v target is the speed of the dynamic obstacle, x is the relative distance between the unmanned ship and the dynamic obstacle, θ relative is the relative azimuth between the unmanned ship and the dynamic obstacle, ω USV It is the turning ability of the unmanned ship; Step S34: Construct dynamic obstacle avoidance incentive function: Among them, γ θ,dyn is a parameter related to the relative angle of the dynamic obstacle. It can adjust the weight of this item in the excitation calculation according to the relative angle of the target ship. y is the velocity component of the dynamic obstacle in the direction of the unmanned ship, and ξ v (θ,v y ) is a velocity component weighting parameter based on the angle and velocity of the dynamic obstacle, which is specifically a piecewise function: η1, η2 and η3 are constant weighting factors, reflecting COLREGs compliance level C rules and The weight of .

5. The unmanned ship path planning and collision avoidance decision method based on multi-factor fusion as claimed in claim 1, characterized in that: The step S4 specifically includes: Integrate the path tracking response, static obstacle and dynamic obstacle avoidance incentive mechanism, and merge their respective incentive functions into an organic whole. The overall incentive function is: where λ (t) is the weight associated with path tracking, which changes dynamically according to different path tracking factors. Its role is to determine the relative importance of path tracking incentives according to the current situation when calculating the total incentive, while 1-λ (t) It reflects the game between the path and the obstacle avoidance incentive; It includes additional incentives related to environmental stability, mission priority, etc. In terms of environmental stability, if the sea conditions are good and the water flow is stable, environmental factors that are conducive to navigation appear, The part related to environmental stability will be given a positive incentive; in terms of task priority, if the current task is an emergency rescue task and the behavior of the unmanned ship helps to improve the rescue efficiency, then a positive incentive will be given according to the task priority.

6. An unmanned ship path planning and collision avoidance decision system 100 based on multi-factor fusion, used to implement the unmanned ship path planning and collision avoidance decision method based on multi-factor fusion as described in any one of claims 1 to 5, characterized in that: The system includes: The unmanned ship multi-sensor fusion perception network module 101 is used to integrate GPS, laser radar, sonar and multi-spectral visual sensors to collect environmental data in real time, and deploy high-sensitivity piezoelectric impact sensors and high-precision optical collision angle measuring instruments in the key collision area of ​​the unmanned ship shell to capture the mechanical parameters at the moment of collision, so that the unmanned ship can perceive and identify the severity of the collision; The unmanned ship path tracking response excitation module 102 is used to simulate the navigation process of the unmanned ship under various path deviation conditions, define the path tracking excitation function, and adaptively adjust the dynamic parameters required in the path control process; The unmanned ship static obstacle and dynamic obstacle collision avoidance incentive module 103 is used to determine the static obstacle collision avoidance incentive function by using regression analysis and machine learning, and to evaluate the threat level of static obstacles and dynamic obstacles by continuously optimizing the sensor data processing algorithm and the incentive function calculation model; The unmanned ship incentive integration module 104 is used to integrate the path tracking response, static obstacle and dynamic obstacle avoidance incentive mechanisms, and merge the respective incentive functions into an organic whole, so that the unmanned ship can balance the relationship between different mission objectives and make path planning and collision avoidance decisions under multiple tasks.

Images