Intelligent decision-making method based on maritime collision avoidance rules

Abstract

The application discloses an intelligent decision-making method based on maritime collision avoidance rules and belongs to the technical field of intelligent ship collision avoidance decision-making, and comprises the following steps: S10, setting a global route of a ship autonomous cruising task; S20, acquiring ship positioning and environment perception prediction data; S30, acquiring a local reference trajectory of the ship through a local trajectory calculation module; S40, constructing a ship navigation encounter situation model; S50, constructing a multivariate composite collision avoidance risk assessment model; S60, constructing a behavior decision tree model based on maritime collision avoidance rules; S70, a trajectory planning module performs trajectory planning based on the behavior decision tree model result; and S80, establishing a monitoring continuous decision-making operation mechanism; the application, under the premise of guaranteeing ship navigation safety and efficiency and under the premise of complying with the International Maritime Collision Regulations, explores the interactive collision avoidance behavior between autonomous ships and manned ships in a complex and changeable sea environment, so as to realize fast and reasonable collision avoidance decision-making, thereby improving the navigation safety and efficiency.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent ship collision avoidance decision-making, and particularly relates to an intelligent decision-making method based on maritime collision avoidance rules. BACKGROUND

[0002] With the rapid development of artificial intelligence and unmanned driving technology, unmanned ships have become an important research field of maritime intelligent transportation, and ship collision avoidance decision-making is a crucial research content. Effective collision avoidance decision-making can ensure the safety of ship navigation, reduce life and property losses and environmental pollution.

[0003] However, in the marine environment, the time-varying nature of wind, flow, surge, and wave is strong, which greatly affects the safety of ship navigation. In addition, the near-sea water area has strong structured navigation characteristics, and there are many types of channel navigation and a large amount of navigation information, such as light floats, light beacons, channel structures, navigation signal lights, and small fishing boats that do not follow rules. Therefore, an intelligent decision-making method based on maritime collision avoidance rules is needed to solve the above problems. SUMMARY

[0004] The purpose of the present application is to provide an intelligent decision-making method based on maritime collision avoidance rules to solve the problems raised in the background.

[0005] To achieve the above purpose, the present application provides the following technical solution: an intelligent decision-making method based on maritime collision avoidance rules, comprising the following steps:

[0006] S10, setting a global route for a ship autonomous cruise task;

[0007] S20, acquiring ship positioning and environment perception prediction data;

[0008] S30, acquiring a local reference trajectory of the ship through a local trajectory calculation module;

[0009] S40, constructing a ship navigation encounter situation model;

[0010] S50, constructing a multi-element compound collision avoidance risk assessment model;

[0011] S60, constructing a behavior decision tree model based on maritime collision avoidance rules;

[0012] S70, trajectory planning module based on behavior decision tree model results for trajectory planning;

[0013] S80, establishing a monitoring continuous decision-making operation mechanism.

[0014] As a preferred scheme, in the step S10, the step S10 of setting a global route for a ship autonomous cruise task comprises the following steps:

[0015] S101, sequentially mark the cruise task route of the intelligent ship on the electronic map, and record the positions on the marked cruise task route to obtain a point queue of the cruise task route;

[0016] S102, convert the point queue of the cruise task route into a Cartesian coordinate system through a longitude and latitude & Cartesian coordinate conversion module;

[0017] S103, compensate the adjacent two points in the global route point queue in the Cartesian coordinate system by using a sample point interpolation algorithm to obtain an expected point queue, and take the expected point queue as a global route.

[0018] As a preferred scheme, in "step S20", the step S20 of acquiring ship positioning and environment perception prediction data comprises the following steps:

[0019] S201, the ship acquires the current GPS longitude and latitude in real time through a global positioning system carried by itself, and converts the longitude and latitude into a Cartesian coordinate system through a longitude and latitude & Cartesian coordinate system conversion module;

[0020] S202, the ship acquires the current driving speed in real time through a speed calculation module configured by itself;

[0021] S203, the ship acquires the current driving heading angle in real time through a heading calculation module configured by itself;

[0022] S204, the ship obtains obstacle boundary data through a perception prediction module.

[0023] As a preferred scheme, in "step S30", the step S30 of acquiring a local reference trajectory of the ship through a local trajectory calculation module comprises the following steps:

[0024] S301, calling a trajectory planning module to generate a current local reference path based on the global route.

[0025] As a preferred scheme, in "step S40", the step S40 of constructing a ship navigation encounter situation model comprises the following steps:

[0026] S401, the International Regulations for Preventing Collisions at Sea provides three encounter scenarios: meeting, crossing and overtaking, but in the actual ship navigation process, the ship has different right-of-way for different encounter situations, so it is necessary to divide the encounter scenarios in detail;

[0027] S402, based on the International Regulations for Preventing Collisions at Sea, a ship navigation encounter area model is established, six encounter areas are divided according to the position relationship of the ship, which are ship bow, right side transverse front, right side transverse rear, ship stern, left side transverse rear, and left side transverse front; taking the ship body position as the center, the ship head direction as the radian 0, and dividing clockwise; the radian area boundary point calculation tool is used to calculate the boundary point coordinates of each encounter area respectively;

[0028] S403, based on the ship navigation encounter area model, in each encounter area, the ship navigation encounter scene situation model is set according to the International Regulations for Preventing Collisions at Sea.

[0029] As a preferred scheme, in "step S50", the step S50 of constructing a multi-element composite collision avoidance risk assessment model comprises the following steps:

[0030] S501, constructing a local reference path 2DBox detection frame set based on local reference path data;

[0031] S502, constructing a DCPA space collision avoidance risk detection model;

[0032] S503, constructing a TCPA time collision avoidance risk detection model.

[0033] As a preferred scheme, in "step S60", the step S60 of constructing a behavior decision tree model based on the International Regulations for Preventing Collisions at Sea comprises the following steps:

[0034] S601, constructing a behavior decision tree feature state;

[0035] S602, constructing a behavior decision tree classification result;

[0036] S603, constructing a behavior decision tree classification model, and setting corresponding decision results under different feature states to constitute training data, and training the behavior decision tree classification model;

[0037] S604, setting a ship current data feature state set;

[0038] S605, based on the ship current data feature state set, predicting the current decision tree result of the ship through the behavior decision tree classification model.

[0039] As a preferred scheme, in "step S70", the step S70 of trajectory planning module based on behavior decision tree model result for trajectory planning comprises the following steps:

[0040] S701, if the behavior decision tree model decision result is , set the current maximum speed boundary of the ship to 0, and plan a stop trajectory through the trajectory planning module with the maximum deceleration parameter of the ship;

[0041] S702, if the behavior decision tree model decision result is , based on the step, the spatial nearest collision avoidance danger distance and the minimum collision avoidance time obtained in the multi-element composite collision avoidance danger assessment model are calculated to obtain the maximum speed boundary of the ship for avoidance;

[0042] S703, if the behavior decision tree model decision result is , the maximum speed parameter of the ship itself is restored, and a right-around overtaking trajectory is planned through the trajectory planning module;

[0043] S704, if the behavior decision tree model decision result is , the maximum speed parameter of the ship itself is restored, and the original channel passing trajectory is planned through the trajectory planning module;

[0044] S705, the trajectory planning module result is sent to the ship control module for execution.

[0045] As a preferred scheme, in the step S80, the step S80 establishes a monitoring continuous decision running mechanism, including the following steps:

[0046] S801, a monitoring continuous decision running program is established, which is continuously cycled from step S20 to step S70, and the decision planning result is adjusted in real time until the autonomous cruise global route navigation is completed.

[0047] Compared with the prior art, the beneficial effects of the present application are:

[0048] The present application, under the premise of ensuring the safety and efficiency of ship navigation, and under the premise of complying with the International Regulations for Preventing Collisions at Sea, explores the interactive collision avoidance behavior between autonomous ships and manned ships in complex and variable sea environments, in order to achieve fast and reasonable collision avoidance decision, thereby improving the safety and efficiency of navigation. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 is a schematic diagram of the overall process of the present application;

[0050] Figure 2 is a flowchart of step S10 of the present application;

[0051] Figure 3 is a flowchart of step S20 of the present application;

[0052] Figure 4 is a flowchart of step S30 of the present application;

[0053] Figure 5 is a flowchart of step S50 of the present application;

[0054] Figure 6Flowchart of step S60 of the present application;

[0055] Figure 7 Flowchart of step S70 of the present application;

[0056] Figure 8 Flowchart of step S80 of the present application. DETAILED DESCRIPTION

[0057] The present application will be further described below in conjunction with examples.

[0058] The following examples are used to illustrate the present application, but cannot be used to limit the protection scope of the present application. The conditions in the examples can be further adjusted according to specific conditions, and simple improvements of the method of the present application under the concept of the present application all belong to the scope of protection of the present application.

[0059] Please refer to Figures 1-8 The present application provides an intelligent decision-making method based on maritime collision avoidance rules, comprising the following steps:

[0060] S10, setting a global route of a ship autonomous cruise task;

[0061] S20, acquiring ship positioning and environment perception prediction data;

[0062] S30, acquiring a ship local reference trajectory through a local trajectory calculation module;

[0063] S40, constructing a ship navigation encounter situation model;

[0064] S50, constructing a multi-element composite collision avoidance risk assessment model;

[0065] S60, constructing a behavior decision tree model based on maritime collision avoidance rules;

[0066] S70, a trajectory planning module performs trajectory planning based on the results of the behavior decision tree model;

[0067] S80, establishing a monitoring continuous decision-making operation mechanism.

[0068] In "step S10", step S10 of setting a global route of a ship autonomous cruise task comprises the following steps:

[0069] S101, marking a cruise task route of an intelligent ship on an electronic map in sequence, and recording the positions on the marked cruise task route to obtain a point queue of the cruise task route;

[0070] S102, converting the above cruise task route point queue into a Cartesian coordinate system through a longitude and latitude & Cartesian coordinate conversion module;

[0071] S103, the sample point interpolation algorithm is used for waypoint compensation on the adjacent two points in the global route point queue in the Cartesian coordinate system, to obtain the expected point queue, and the expected point queue is taken as the global route, that is .

[0072] In "step S20", the ship positioning and environment perception prediction data acquisition step S20 includes the following steps:

[0073] S201, the ship obtains the current GPS longitude and latitude in real time through the global positioning system carried by itself, denoted as , and converts it into a Cartesian coordinate system through a longitude and latitude & Cartesian coordinate system conversion module, denoted as ;

[0074] S202, the ship obtains the current driving speed in real time through the speed calculation module configured by itself, denoted as ;

[0075] S203, the ship obtains the current driving heading angle in real time through the heading calculation module configured by itself, denoted as ;

[0076] S204, the ship obtains obstacle boundary data through the perception prediction module, including two types of data:

[0077] One type is static obstacle boundary line data, which is composed of line geometry, specifically ;

[0078] One type is obstacle data, including static obstacles and dynamic obstacles, which are composed of polygon geometry: ;

[0079] Among them, the dynamic obstacle data is specifically all obstacle boundaries at time ; obstacle speed , obstacle direction .

[0080] In "step S30", the ship local reference trajectory is obtained by the local trajectory calculation module in step S30, including the following steps:

[0081] S301, call the trajectory planning module to generate the current local reference path based on the global route, which is obtained by time dimension interpolation, denoted as , wherein represents the first point coordinate of the local reference path at time represents the sample point coordinate of the local reference path at time represents the tail point coordinate of the local reference path.

[0082] In "step S40", step S40 constructs a ship navigation encounter situation model, including the following steps:

[0083] S401, the International Regulations for Preventing Collisions at Sea gives three encounter scenarios of meeting, crossing and overtaking, but in the actual process of ship navigation, ships have different route rights for different encounter situations, so it is necessary to divide the encounter scenarios in detail;

[0084] S402, based on the International Regulations for Preventing Collisions at Sea, a ship navigation encounter area model is established, and six encounter areas are divided according to the position relationship of the ship, which are bow, right side transverse front, right side transverse rear, stern, left side transverse rear and left side transverse front; taking the ship body position as the center, the parameter is the radius, the ship head direction is the radian 0, and the clockwise division, is the bow range, is the right side transverse front range, is the right side transverse rear range, is the stern range, is the left side transverse rear range, is the left side transverse front range; the boundary point coordinates of each encounter area are calculated by the arc area boundary point calculation tool respectively:

[0085] The bow encounter area boundary point coordinate set is:

[0086] ;

[0087] The right side transverse front encounter area boundary point coordinate set is:

[0088] ;

[0089] The right side transverse rear encounter area boundary point coordinate set is:

[0090] ;

[0091] The stern encounter area boundary point coordinate set is:

[0092] ;

[0093] The left side transverse rear encounter area boundary point coordinate set is:

[0094] ;

[0095] The left side transverse front encounter area boundary point coordinate set is:

[0096] ;

[0097] S403, based on the ship navigation encounter area model, in each encounter area, combined with the International Regulations for Preventing Collisions at Sea, set the ship navigation encounter scene situation model, divided into: HO encounter scene, CR cross encounter scene, OT overtaking scene and FO following scene;

[0098] set the ship heading angle as , the ship speed as , the target obstacle ship heading angle as and the speed as ;

[0099] calculate the difference in degrees of the ship and the target obstacle ship heading angle ; encounter scene The specific calculation formula of the model is as follows:

[0100] ,

[0101] As shown in the above formula, when the angle difference between the subject ship and the obstacle ship satisfies , it is an HO encounter scene; when the angle difference between the subject ship and the obstacle ship satisfies , it is a CR cross scene; when the angle difference between the subject ship and the obstacle ship satisfies ,

[0102] and the speed satisfies , it is an OT overtaking scene; when the angle difference between the subject ship and the obstacle ship satisfies , and the speed satisfies , it is an FO following scene.

[0103] In "step S50", step S50 constructs a multi-element composite collision avoidance risk assessment model, including the following steps:

[0104] S501, based on the local reference path data, a local reference path 2DBox detection frame set is constructed;

[0105] The step of constructing a local reference path 2DBox detection frame based on local reference path data includes the following steps:

[0106] S5011, set the ship safety detection shape (lateral and longitudinal parameters), which is generally the length and width of the ship itself plus a certain safety buffer distance, denoted as ;

[0107] S5012, calculate the heading angle of each t-time sample point in the local reference path ;

[0108] The step of calculating the heading angle of each t-time sample point in the local reference path includes the following steps:

[0109] Select a reference point: The time point is the location where the tangential angle will be calculated;

[0110] Determine adjacent points: The point in time is the next point adjacent to the reference point, that is, the next point on the local reference path;

[0111] Calculate the direction vector: Using adjacent coordinates, calculate the direction vector (vector), which represents the direction from the reference point to the next point. The method for calculating the direction vector is as follows: ,in These are the coordinates of the reference point. These are the coordinates of adjacent points; calculate the orientation angle: use the arctangent function to calculate the tangential angle:

[0112] ;

[0113] For local reference paths Sample points at time t The above steps yield a queue of local reference path orientation angles: ;

[0114] S5013, Based on local reference path coordinates and local reference path orientation angle traverse each local path Time sample point, based on the coordinates of that point Orientation angle and ship safety shape Calculate the coordinates of the 2DBox bounding box corresponding to the sample point. This includes the following steps:

[0115] Calculate the coordinates of each vertex of the bounding box:

[0116] For the coordinates of the first corner point (top left corner):

[0117] ,

[0118] ,

[0119] For the coordinates of the second corner point (top right corner):

[0120] ,

[0121] ,

[0122] For the coordinates of the third corner point (bottom right corner):

[0123] ,

[0124] ,

[0125] For the fourth corner point (lower left corner) coordinates:

[0126] ,

[0127] ,

[0128] Therefore, the calculated sample point corresponding to the bounding box coordinates:

[0129] .

[0130] S5014, traverse the local reference path every moment sample point, execute the above steps, generate the 2DBox boundary box set corresponding to the local reference path:

[0131] ;

[0132] S502, construct a DCPA space collision avoidance risk detection model;

[0133] The step of constructing a DCPA space collision avoidance risk detection model includes the following steps:

[0134] S5021, traverse the 2DBox detection frame generated by the local reference path every t moment , and all boundary lines & obstacle geometric data to detect whether there is a collision:

[0135] Calculate whether the sample point bounding box intersects the boundary line through the line and polygon intersection calculation module;

[0136] Calculate whether the sample point bounding box intersects each obstacle bounding box through the two polygon intersection calculation module, if it is a dynamic obstacle, only intersect with moment bounding box to judge intersection;

[0137] If there is an intersection, it is considered that the sample point position is a spatial nearest collision avoidance danger point, denoted as , and then end the sample point traversal.

[0138] S5022, calculate the spatial minimum collision avoidance danger distance based on the local reference path and the spatial nearest collision avoidance danger point :

[0139] ,

[0140] Among them is the nearest collision avoidance point The time, For The coordinates corresponding to the point, For The next point in time The coordinates corresponding to the point, That is, the minimum space collision danger distance;

[0141] If the local reference path is completed All sample points are traversed, and no spatial nearest collision danger point is generated , Set the minimum space collision danger distance to positive infinity;

[0142] S5023, combined with the International Regulations for Preventing Collisions at Sea, the smaller the space collision danger distance, the greater the space collision danger degree of the ship, and the space collision danger degree membership function of the ship is constructed For:

[0143] ;

[0144] In the formula: The minimum safe meeting distance parameter is usually 20 times the ship length, The cumulative s length parameter of the local reference path is usually the perception range distance + 12 times the ship length.

[0145] S503, construct a TCPA time collision danger degree detection model;

[0146] The step of constructing the TCPA time collision danger degree detection model includes the following steps:

[0147] S5031, construct a single dynamic obstacle and the nearest collision danger time detector of the ship:

[0148] Traverse the 2DBox detection frame generated by the local reference path every t time , With the boundary box Of the dynamic obstacle every t time Whether to collide: calculate whether the sample point boundary box Intersect with the boundary box of the dynamic obstacle at t time, if they intersect, consider that the t time of this sample point is the nearest collision danger time point of the ship and the dynamic obstacle, mark it as , And end the sample point traversal;

[0149] If all sample points of the local reference path are traversed and no nearest collision danger time point is generated, set the nearest collision danger time point of the ship and the dynamic obstacle to positive infinity.

[0150] S5032, construct a single dynamic obstacle and the ship's encounter scene detector:

[0151] If the dynamic obstacle's closest collision danger time is positive infinity after the closest collision danger time detection, it is considered that the dynamic obstacle has no encounter relationship with the ship;

[0152] Determine the dynamic obstacle's encounter region: Based on the ship's encounter situation model, the ship's encounter region boundary set is obtained: the bow encounter region , the starboard quarter encounter region , the port quarter encounter region , the stern encounter region , the starboard bow encounter region , and the port bow encounter region ;

[0153] Whether the dynamic obstacle's position coordinates are in a certain encounter region is calculated by the point-in-polygon calculation module.

[0154] If the dynamic obstacle is inside a certain encounter region, set the encounter region as the dynamic obstacle's encounter region, which is used for subsequent decision-making.

[0155] If the dynamic obstacle is not inside any encounter region, it is considered that the dynamic obstacle has no encounter relationship with the ship.

[0156] Determine the dynamic obstacle's encounter scenario: Based on the ship's encounter situation model, the encounter scenario situation model is obtained, and the dynamic obstacle's encounter scenario is calculated. The encounter scenario result is one of the following scenarios: meeting, crossing, overtaking, and following.

[0157] S5033, iterate all dynamic obstacle data to generate each dynamic obstacle's closest collision danger time, encounter region, and encounter scenario; calculate the minimum collision danger time of all dynamic obstacles:

[0158] ;

[0159] In the formula , represents the calculated closest collision danger time of each dynamic obstacle.

[0160] S5034, combined with the International Regulations for Preventing Collisions at Sea, the smaller the minimum collision danger time, the greater the ship's time collision danger degree, and the ship's spatial collision danger degree membership function is constructed :

[0161] ;

[0162] In the formula , T is the minimum safe encounter time parameter, is the local reference path end state t time parameter.

[0163] S504, constructing a membership function of the space collision risk model and a membership function of the time collision risk model , constructing a multi-element composite collision risk assessment model, and the specific formula is as follows:

[0164] ,

[0165] In the formula, is a space collision risk weight, is a space collision risk result, is a time collision risk weight, is a time collision risk result, is a final collision risk detection result;

[0166] The higher the collision risk is, the more dangerous the current ship navigation state is;

[0167] In the "step S60", the step S60 of constructing a behavior decision tree model based on maritime collision rules includes the following steps:

[0168] S601, constructing a behavior decision tree feature state; the decision tree feature state is set as:

[0169] S6011, whether in an emergency collision avoidance scenario, represented as: ;

[0170] S6012, whether in a speed reduction avoidance scenario, represented as: ;

[0171] S6013, whether in a overtaking and bypassing scenario, represented as .

[0172] S602, constructing a behavior decision tree classification result; the decision tree classification result is set as:

[0173] S6021, a static decision result, indicating that the ship should reduce to 0 at the maximum deceleration and remain static, denoted as ;

[0174] S6022, an avoidance decision result, indicating that the ship needs to maintain the heading and reduce speed to avoid, denoted as ;

[0175] S6023, an overtaking decision result, indicating that the ship needs to overtake and bypass at this time, according to the International Regulations for Preventing Collisions at Sea, when two motor ships meet on opposite or nearly opposite courses so as to involve risk of collision, each ship shall alter her course to starboard, so as to pass on the other's port side; therefore, it should be right overtaking and bypassing, denoted as ;

[0176] S6024, the passing decision result indicates that the ship keeps the original course and speed and normally sails in the original channel, denoted as .

[0177] S603, a behavior decision tree classification model is constructed, and corresponding decision results in different feature states are set to constitute training data, and the behavior decision tree classification model is trained:

[0178] S6031, the feature state set of the training data is set to whether the emergency collision avoidance scene is true, whether the speed reduction avoidance scene is true, and whether the overtaking and detour scene is true , the corresponding decision result is static .

[0179] S6032, the feature state set of the training data is set to whether the emergency collision avoidance scene is true, whether the speed reduction avoidance scene is true, and whether the overtaking and detour scene is true , the corresponding decision result is static .

[0180] S6033, the feature state set of the training data is set to whether the emergency collision avoidance scene is true, whether the speed reduction avoidance scene is true, and whether the overtaking and detour scene is true , the corresponding decision result is static .

[0181] S6034, the feature state set of the training data is set to whether the emergency collision avoidance scene is true, whether the speed reduction avoidance scene is true, and whether the overtaking and detour scene is true , the corresponding decision result is static .

[0182] S6035, the feature state set of the training data is set to whether the emergency collision avoidance scene is true, whether the speed reduction avoidance scene is true, and whether the overtaking and detour scene is true , the corresponding decision result is avoidance .

[0183] S6036, the feature state set of the training data is set to whether the emergency collision avoidance scene is true, whether the speed reduction avoidance scene is true, and whether the overtaking and detour scene is true , the corresponding decision result is avoidance .

[0184] ​​​​​​​​​​​​S6037, set the training data feature state set as whether an emergency collision avoidance scenario false, whether to slow down and avoid the scene false and whether to overtake and detour the scene false, the decision result is to pass through

[0185] S6038, set the training data feature state set as whether an emergency collision avoidance scenario false, whether to slow down and avoid the scene false and whether to overtake and detour the scene true, the decision result is to overtake

[0186] S604, set the current data feature state set of the ship:

[0187] S6041, whether an emergency collision avoidance scenario feature state decision: collision risk degree obtained in the step of constructing a multi-element composite collision risk assessment model

[0188] when the collision risk degree is greater than 0, set whether an emergency collision avoidance scenario feature state is true, otherwise false;

[0189] S6042, whether to slow down and avoid the scene feature state decision: based on the detection result of the ship navigation encounter scene situation model obtained in the step of constructing a multi-element composite collision risk assessment model, combined with the International Regulations for Preventing Collisions at Sea, the ship on the right side of the ship should give way to the ship;

[0190] when the right side of the ship encounters the area in front of the right side of the ship and the right side of the ship encounters the area behind the right side of the ship and other obstacle ships exist encounter scene relationship, set whether to slow down and avoid the scene feature state is true;

[0191] In addition, when the bow encounter area of the ship and other obstacle ships exist following, crossing encounter scene relationship, set whether to slow down and avoid the scene feature state is true;

[0192] In addition, when the stern encounter area of the ship exists overtake encounter scene relationship, set whether to slow down and avoid the scene feature state is true;

[0193] ​​​In the actual navigation practice, in order to improve the safety, when the left side of the ship is transversely encountered in the front of the area and the high-speed obstacle ship (here, the high-speed ship is determined by setting a speed threshold in practice, and if the obstacle ship is greater than the threshold, it is considered to belong to the high-speed obstacle ship) exists in the encounter scene relationship, set whether to slow down to avoid the scene The feature state is true;

[0194] If none of the above decision conditions are met, set whether to slow down to avoid the scene The feature state is false;

[0195] S6043, whether to overtake and bypass the scene Feature state decision: based on the collision avoidance danger degree obtained in the step of constructing a multi-element composite collision avoidance danger degree evaluation model And the ship navigation encounter scene situation model detection result;

[0196] When the collision avoidance danger degree The ship has no dynamic obstacle encounter scene relationship in each encounter area, that is, only a static obstacle collision avoidance scene, or the ship's bow encounter area exists in the encounter, overtaking encounter scene relationship with other obstacle ships, then set whether to overtake and bypass the scene The feature state is true, otherwise it is false;

[0197] S605, based on the current data feature state set of the ship: ,

[0198] The current decision tree result of the ship is predicted by the behavior decision tree classification model, which should be: One of the formulas;

[0199] In "step S70", the trajectory planning module based on the behavior decision tree model result trajectory planning includes the following steps:

[0200] S701, if the behavior decision tree model decision result is , set the current maximum speed boundary of the ship to 0, and plan a stop trajectory with the maximum deceleration parameter of the ship through the trajectory planning module;

[0201] S702, if the behavior decision tree model decision result is , based on the spatial nearest collision avoidance danger distance And the minimum collision avoidance time , calculate the maximum avoidance speed boundary of the ship, including the following steps:

[0202] S7021, based on the spatial nearest collision avoidance danger distance Calculate the avoidance maximum speed boundary :

[0203] wherein, is the maximum deceleration parameter of the ship;

[0204] S7022, based on the minimum collision danger time calculate the maximum speed boundary of avoidance :

[0205] ,

[0206] wherein, is the maximum deceleration parameter of the ship;

[0207] S7023, based on the maximum speed boundary of spatial collision danger and the maximum speed boundary of time collision danger, set the maximum speed boundary of avoidance decision result ; plan the avoidance trajectory by the trajectory planning module with the maximum deceleration parameter of the ship;

[0208] S703, if the behavior decision tree model decision result is , restore the maximum speed parameter of the ship itself, and plan the right-around overtaking trajectory by the trajectory planning module;

[0209] S704, if the behavior decision tree model decision result is , restore the maximum speed parameter of the ship itself, and plan the original channel passing trajectory by the trajectory planning module;

[0210] S705, send the above trajectory planning module result to the ship control module for execution.

[0211] In "step S80", step S80 establishes a monitoring continuous decision running mechanism, including the following steps:

[0212] S801, establish a monitoring continuous decision running program, continuously cycle from step S20 to step S70, adjust the decision planning result in real time until the autonomous cruise global route navigation is completed.

[0213] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.​

Claims

1. An intelligent decision making method based on the rules of collision avoidance at sea, characterized in that: The method comprises the following steps: S10, setting a global route of a ship autonomous cruise task; S20, acquiring ship positioning and environment perception prediction data; S30, acquiring a ship local reference trajectory through a local trajectory calculation module; S40, constructing a ship navigation encounter situation model; S50, constructing a multi-element composite collision avoidance risk assessment model; S60, constructing a behavior decision tree model based on maritime collision avoidance rules; S70, the trajectory planning module performs trajectory planning based on the behavior decision tree model result; The step S70 trajectory planning module based on the behavior decision tree model result trajectory planning includes the following steps: S701, if the behavior decision tree model decision result is , set the current maximum speed boundary of the ship to 0, and plan a stop trajectory through the trajectory planning module with the maximum deceleration parameter of the ship; S702、if the behavior decision tree model decision result is , based on the step of constructing a multi-element composite collision avoidance danger assessment model, the spatial nearest collision avoidance danger distance and the minimum collision avoidance time , calculate the maximum speed boundary of the ship to avoid, including the following steps: S7021、Based on the space nearest collision avoidance dangerous distance Calculate the maximum speed boundary of avoidance : wherein is the maximum deceleration parameter of the ship; S7022、Based on the minimum collision avoidance danger time Calculate the maximum speed boundary of avoidance : wherein, is the final velocity set to 0, is the maximum deceleration parameter of the ship; S7023、based on the space collision avoidance danger maximum speed boundary and the time collision avoidance danger maximum speed boundary, set the maximum speed boundary of the avoidance decision result ; plan the avoidance trajectory with the ship maximum deceleration parameter through the trajectory planning module; S703, if the behavior decision tree model decision result is , the ship itself maximum speed parameter is restored, and a right-around overtaking trajectory is planned through the trajectory planning module. S704, if the behavior decision tree model decision result is , the ship itself maximum speed parameter is restored, and the original channel passing trajectory is planned through the trajectory planning module. S705, the trajectory planning module result is sent to the ship control module for execution; S80, establishing a monitoring continuous decision operation mechanism. 2.The intelligent decision-making method based on the maritime collision avoidance rule according to claim 1, characterized in that: In "step S10", the step S10 sets a global route of a ship autonomous cruise task, which includes the following steps: S101, marking the cruise task route of the intelligent ship on the electronic map in turn, and recording the positions on the marked cruise task route to obtain a point queue of the cruise task route; S102, converting the above-mentioned cruise task route point queue into a Cartesian coordinate system through a longitude and latitude & Cartesian coordinate conversion module; S103, using a sample point interpolation algorithm to compensate the adjacent two points in the global route point queue in the Cartesian coordinate system to obtain an expected point queue, and taking the expected point queue as the global route. 3.The intelligent decision-making method based on the maritime collision avoidance rule of claim 1, wherein: In "step S20", the step S20 acquires ship positioning and environment perception prediction data, which includes the following steps: S201, the ship acquires the current GPS longitude and latitude in real time through the global positioning system carried by itself, and converts it into a Cartesian coordinate system through a longitude and latitude & Cartesian coordinate system conversion module; S202, the ship acquires the current driving speed in real time through the speed calculation module configured by itself; S203, the ship acquires the current driving heading angle in real time through the heading calculation module configured by itself; S204, the ship obtains obstacle boundary data through the perception prediction module. 4.The intelligent decision-making method based on the maritime collision avoidance rule of claim 1, wherein: In "step S30", the step S30 acquires a ship local reference trajectory through a local trajectory calculation module, which includes the following steps: S301, calling the trajectory planning module to generate a current local reference path based on the global route.

5. The intelligent decision making method based on COLREGs as claimed in claim 1, wherein: In "step S40", the step S40 constructs a ship navigation encounter situation model, which includes the following steps: S401, the International Regulations for Preventing Collisions at Sea gives three encounter scenarios: meeting, crossing and overtaking, but in the actual ship navigation process, the ship has different navigation rights for different encounter situations, so the encounter scenarios need to be divided in detail; S402, based on the International Regulations for Preventing Collisions at Sea, a ship navigation encounter area model is established, which divides six encounter areas according to the position relationship of the ship, namely, the bow, the right side, the stern, the left side, and the front; taking the ship body position as the center and the ship heading as the radian 0, the areas are divided clockwise; the boundary point coordinates of each encounter area are calculated through the radian area boundary point calculation tool; S403, based on the ship navigation encounter area model, in each encounter area, set the ship navigation encounter scene situation model combined with the International Regulations for Preventing Collisions at Sea. 6.The intelligent decision making method based on maritime collision avoidance rules of claim 1, wherein: In "step S50", the step S50 of constructing a multi-element composite collision avoidance risk assessment model includes the following steps: S501, based on the local reference path data, construct a local reference path 2DBox detection frame set; S502, construct a DCPA space collision avoidance risk detection model; S503, construct a TCPA time collision avoidance risk detection model.

7. The intelligent decision making method based on COLREGs as claimed in claim 1, wherein: In "step S60", the step S60 of constructing a behavior decision tree model based on the International Regulations for Preventing Collisions at Sea includes the following steps: S601, construct a behavior decision tree feature state; S602, construct a behavior decision tree classification result; S603, construct a behavior decision tree classification model, and set the corresponding decision results under different feature states to constitute training data, and train the behavior decision tree classification model; S604, set the current data feature state set of the ship; S605, based on the current data feature state set of the ship, predict the current decision tree result of the ship through the behavior decision tree classification model. 8.The intelligent decision making method based on maritime collision avoidance rules of claim 1, wherein: In "step S80", the step S80 of establishing a monitoring continuous decision running mechanism includes the following steps: S801, establish a monitoring continuous decision running program, continuously run from step S20 to step S70, adjust the decision planning result in real time, and complete the autonomous cruise global route navigation.

Images