A Formal Expression Method for Ship Collision Avoidance Rule Scenario Knowledge

Abstract

The present invention discloses a method for formal expression of scenario knowledge of ship collision avoidance rules. The method includes: defining different water traffic scenarios; conducting element analysis and semantic expression on the water traffic scenarios involved in ship collision avoidance rules, and abstractly expressing the water traffic scenario as a conceptual model of a quadruple: scenario = {E, A, R, B}; that is, the water traffic scenario model includes entities E involved in the water traffic scenario, attributes A of objects, relationships R between objects, and behavioral elements B of objects; combining the structural characteristics of the water traffic scenario model in collision avoidance rules, and using the finite state machine method for extension; based on the extended finite state machine method, conducting ontology modeling on ship collision avoidance rule scenario knowledge and constructing a formal description in the form of a scenario knowledge graph. The present invention converts the scenario knowledge involved in collision avoidance rules into content recognizable and processable by a computer, forming knowledge with structured and formal expression, enabling a machine to understand the scenario knowledge through learning.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence knowledge representation, and in particular to a method for formalizing the expression of ship collision avoidance rule scenario knowledge. Background Art

[0002] With the continuous breakthroughs in software and hardware technologies such as sensors and artificial intelligence, and the rapid development of ship intelligence and unmanned operation, the main body of ship autonomous driving has gradually changed from humans to machines. Ship drivers can quickly integrate various scenario information during ship navigation, capture key information from the complex navigation scenario information, establish the correlation between key element information, and consider the constraints of water traffic rules on ship behavior in different water traffic scenarios, so as to form knowledge guiding ship behavior. However, it is difficult for machines to achieve this because machines lack the knowledge basis for understanding and judging complex navigation scenarios and cannot complete the transformation from data information to scenario knowledge and navigation situation knowledge. In the information world, it is necessary to convert various scenario knowledge descriptions understood by drivers into content that can be recognized and processed by computers, and form knowledge with structured and formalized expressions, so that machines can learn this knowledge to achieve a similar understanding of scenario knowledge as humans.

[0003] Domestic and foreign scholars have done a lot of work on traffic scenario modeling and expression and achieved relatively remarkable research results. However, there are still some scientific problems to be solved, specifically including the following aspects: 1) Focusing on analyzing typical scenarios and lacking the analysis of the constituent elements of scenario knowledge. Most studies only focus on the ship collision avoidance algorithm research of ships in typical scenarios at the data level, so they fail to reveal the constituent elements of scenario knowledge and the relationship between elements, and it is difficult to completely express scenario knowledge. 2) Focusing on analyzing single objects and single features and lacking the structured modeling and expression of scenario knowledge. The current research mainly focuses on the research of autonomous collision avoidance methods for the spatial characteristics of ships in open waters, often only involving single ships, single waters, and single features. It ignores that water scenarios involve numerous water traffic objects, and each object has different characteristics, such as time, space, attributes, relationships, etc. Therefore, the research needs to clarify the connotation of scenario knowledge and systematically study the construction of scenario knowledge models. It is necessary to further improve the logical structure of scenario knowledge reasoning to achieve the formalized expression of scenario knowledge. 3) Focusing on listing scenario instances and lacking the research on the abstraction and inferability of scenarios. The current research only lists scenarios in a descriptive manner and lacks the abstraction and inferability of scenarios. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a formal expression method for scenario knowledge of ship collision avoidance rules in view of the defects in the prior art, convert the scenario knowledge involved in the collision avoidance rules and various scenario knowledge understood by the driver into content that can be recognized and processed by a computer, and form knowledge with structured and formal expression, so that the machine can learn this knowledge to achieve a similar understanding degree of scenario knowledge as humans.

[0005] The technical solution adopted by the present invention to solve its technical problems is as follows:

[0006] The present invention provides a formal expression method for scenario knowledge of ship collision avoidance rules, and this method includes the following steps:

[0007] Step 1: Obtain data of different water traffic scenarios and define the water traffic scenarios.

[0008] Step 2: Conduct element analysis and semantic expression on the water traffic scenarios involved in the ship collision avoidance rules. The ship collision avoidance rules include the International Regulations for Preventing Collisions at Sea, the Inland River Collision Avoidance Rules, and local rules; abstractly express the water traffic scenario as a conceptual model of a quadruple: scenario = {E, A, R, B}; that is, the water traffic scenario model includes the entities E involved in the water traffic scenario, the attributes A of the objects, the relationships r between the objects, and the behavioral elements B of the objects.

[0009] Step 3: Combine the structural characteristics of the water traffic scenario model in the collision avoidance rules and use the finite state machine method for extension.

[0010] Step 4: Based on the extended finite state machine method, conduct ontology modeling on the scenario knowledge of the ship collision avoidance rules and construct a formal description in the form of a scenario knowledge graph.

[0011] Further, the method of Step 1 of the present invention specifically includes:

[0012] Considering that the water traffic scenario has time continuity and space topology, define the water traffic scenario as a water area complex with a specific structure and function formed by the mutual connection and interaction of various water traffic elements within a certain water area and different spatio-temporal ranges; the water traffic elements include the natural environment, navigation infrastructure, ships, and shipboard personnel; the water traffic scenario includes 6 elements: time, space, traffic objects, things, phenomena, and events.

[0013] Further, the method of Step 2 of the present invention specifically includes:

[0014] Abstractly express the water traffic scenario as a conceptual model of a quadruple: scenario = {E, A, R, B}; where scenario represents the water traffic scenario of the ship collision avoidance rules.

[0015] E represents the entities involved in the water traffic scenario, i.e., water traffic objects, including: people, ships, and environmental objects;

[0016] A represents the attributes of the objects, i.e., the corresponding properties that the objects possess, including: ship size, speed, heading, environmental visibility, and traffic density;

[0017] R represents the relationships between objects, i.e., the interrelated relationships between people and ships, people and the environment, ships and ships, ships and the environment, and the environment and the environment;

[0018] B represents the behavioral elements of the objects, i.e., the dynamic changes over time of the attribute elements of the objects or the relationship elements between objects.

[0019] Furthermore, the method for expansion using the finite state machine method in step 3 of the present invention specifically includes:

[0020] (1) Adding a description mechanism for time and space features; the expanded finite state machine scenario = {E, A, R, B, T, S, W} is a set of time conditions, T is a set of space conditions, S is the input conditions other than time and space, and W is defined as the set of trigger conditions;

[0021] (2) Expanding the finite state machine into a hierarchical form (scenario1, scenario2,..., scenario n ), n ≥ 3;

[0022] (3) State transitions are allowed to exist in parallel, expanding the key feature set U of each scenario state, recording the start and end times of the state; and expanding the scenario state transitions into normal transitions and concurrent transitions, and scenario state transitions must meet all the requirements of scenario feature changes; when the time condition does not belong to the time feature of the previous state, a normal transition is made, otherwise a concurrent transition is made.

[0023] Furthermore, the method for expansion using the finite state machine method in step 3 of the present invention specifically includes:

[0024] The expanded finite state machine M is a 9-tuple (S, so, I, O, σ, γ, Y, F, T), where:

[0025] (1) S represents a finite non-empty set of scenario states, i.e., Scenario = {Scenario1, Scenario2,..., Scenario n};

[0026] (2) so ∈ S represents the initial state of the attributes, relationships, and behaviors of the objects in the scenario of the finite state machine;

[0027] (3) I and O respectively represent finite non - empty sets of input symbols and output symbols;

[0028] (4) σ represents the state transition function;

[0029] Use the maximum state - related ratio as the state transition path, as shown in the following formula:

[0030]

[0031] where w is the state - related ratio, k1 and k2 are correction coefficients, S input is the number of input variables of the state, S output is the number of output variables related to the input state, and b is the number of input and output state variables;

[0032] Design the state transition function, as shown in the following formula:

[0033]

[0034] where w i represents the related ratio of the i - th state variable, and z represents the number of all state transition paths;

[0035] (5) γ represents the output function;

[0036] Scenario output =γ(σ·Scenario input ) (3)

[0037] Scenario output and Scenario input represent the output and input element variables of the scenario, and such variables include eigenvectors and eigenvalues such as attributes, relationships, and behaviors changing over time and space;

[0038] (6) Y represents the context, that is, the set of variables of consecutive scenarios;

[0039] (7) represents the set of termination states of the extended finite - state machine;

[0040] (8) T represents a finite non - empty set of state migrations, and any migration t in T is represented as a six - tuple (si, sj, i, o, p, a), where:

[0041] 1) si, sj ∈ S respectively represent the start state and end state of the migration t;

[0042] 2) i ∈ I and o ∈ O respectively represent the input symbol and output symbol of the migration t;

[0043] 3) p represents the guard condition that needs to be satisfied for the occurrence of transition t, which is a logical expression;

[0044] 4) a represents a set of assignment statements that modify the values of context variables when transition t occurs;

[0045] The process of scenario transition is described as follows: When the current state of the finite state machine M is S, if it receives the input symbol a, it will transfer to the state σ(S, a) and generate the output symbol γ(S, a), where the actual semantics of the input and output symbols are events, messages, or operations.

[0046] Furthermore, the method in step 4 of the present invention specifically includes:

[0047] Based on the ship collision avoidance rule knowledge extended by an extended finite state machine, each complete input scenario can be regarded as a directed path graph from the state transition function to the output scenario output variables. Link each path graph to form a ship collision avoidance scenario knowledge graph. The specific method is as follows:

[0048] According to the obtained scenario directed path graph Scenario input →Scenario output , further calculate the similarity of the directed path graph based on semantic information, which is used to splice the directed path graphs to construct a ship collision avoidance scenario knowledge graph;

[0049] The similarity of semantic information between a directed path graph V Scenarioinput (i) and another directed path graph V Scenariooutput (j) is:

[0050]

[0051] where Ti and Tj are the semantic information vectors of the two directed graphs. Perform label propagation on the graph. If the size of the set of directed path graphs is N(L), then define V Scenarioinput (i), V Scenariooutput (j)'s propagation probability P is:

[0052]

[0053] Propagate according to the propagation probability from high to low until the data of the constructed set of directed path graphs is completely propagated, forming a complete ship collision avoidance scenario knowledge graph.

[0054] Furthermore, the method of the present invention also includes storing scenario knowledge using the neo4j graph database and visually displaying the scenario knowledge graph using the d3.js software.

[0055] The beneficial effects produced by the present invention are as follows: The method for formal expression of ship collision avoidance rule scenario knowledge of the present invention considers the constraints of ship collision avoidance rules on ship behavior in different water traffic scenarios, abstractly analyzes water traffic scenario information according to ship collision avoidance rules, establishes the association relationship between key element information to form a water traffic scenario model, conducts ontology modeling on the water traffic knowledge in the rules and constructs a water traffic scenario knowledge graph, converts the scenario knowledge involved in collision avoidance rules and various scenario knowledge understood by drivers into content recognizable and processable by a computer, forms knowledge expressed in a structured and formal way, so that a machine can learn this knowledge to achieve a similar understanding level of scenario knowledge as humans. Description of the Drawings

[0056] The present invention will be further described below in conjunction with the drawings and embodiments. In the drawings:

[0057] Figure 1 is the flowchart of the method for formal expression of ship collision avoidance rule scenario knowledge of the present invention;

[0058] Figure 2 is the composition diagram of ship collision avoidance rule scenario elements of the present invention;

[0059] Figure 3 is the knowledge composition diagram of ship collision avoidance rule scenario entity elements of the present invention;

[0060] Figure 4 is the knowledge composition diagram of ship collision avoidance rule scenario attribute elements of the present invention;

[0061] Figure 5 is the knowledge composition diagram of ship collision avoidance rule scenario relationship elements of the present invention;

[0062] Figure 6 is the knowledge composition diagram of ship collision avoidance rule scenario behavior elements of the present invention;

[0063] Figure 7 is the example diagram of the ship collision avoidance rule scenario knowledge graph of the present invention. Detailed Embodiments

[0064] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0065] The present invention proposes a method for formal expression of ship collision avoidance rule scenario knowledge, Figures 1 to 7 is an embodiment of a method for formal expression of ship collision avoidance rule scenario knowledge proposed by the present invention.

[0066] Refer to Figure 1 ,Figure 1 This is an embodiment of a method for formalizing the knowledge of ship collision avoidance rule scenarios in the present invention, including:

[0067] S101 summarizes and proposes the definition and connotation of the water traffic scenario;

[0068] S02 conducts element analysis and semantic expression on the water traffic scenarios involved in ship collision avoidance rules. It should be noted that the ship collision avoidance rules here include the International Regulations for Preventing Collisions at Sea, the Inland Navigation Rules, and local rules. The water traffic scenario is abstractly expressed as a conceptual model of a quadruple using the finite state machine method: scenario = {E, A, R, B}; that is, the water traffic scenario includes the entities E involved in the water traffic scenario, the attributes A of the objects, the relationships R between the objects, and the behavioral elements B of the objects;

[0069] S103 combines the structural characteristics of the water traffic scenario model in the collision avoidance rules and is extended using the finite state machine method;

[0070] S104, based on the extended finite state machine method, conducts ontology modeling on the ship collision avoidance rule scenario knowledge and constructs a formal description in the form of a scenario knowledge graph;

[0071] As a preferred technical solution, the element analysis, semantic expression of the water traffic scenarios involved in ship collision avoidance rules, and the formal expression of ship collision avoidance rule scenarios include:

[0072] 1. The water traffic scenario is abstractly expressed as a conceptual model of a quadruple using the finite state machine method: scenario = {E, A, R, B};

[0073] Among them, scenario represents the water traffic scenario of ship collision avoidance rules,

[0074] E represents the entities involved in the water traffic scenario, which can be divided into physical entities and logical entities. Physical entities are the main participants in ship behaviors and events, including ships, participating personnel (such as lookouts, helmsmen, and decision-makers), and the spatial environment when ship behaviors and events occur, etc.; logical entities mainly include the states of ships (underway, at anchor, aground, and moored to the shore), sensing data (AIS, GPS, and radar), etc. And it is expressed using a triple: E = {ship, human, environment};

[0075] A represents the attributes of an object. Attribute elements describe entities and their behaviors and events. Attribute elements can be divided into parameter attributes and functional attributes. Among them, parameter attributes are used to describe the attribute information of entities. Parameter attributes are further divided into inherent attributes (such as the type and size of a ship) and dynamic attributes (such as the position, heading, and speed of a ship), etc. Parameter attributes are used for further analysis and calculation; functional attributes are the functional information of entities, such as describing the maneuverability of a ship (restricted maneuverability ship, ship restricted by draft, etc.)

[0076] R represents the relationships between objects, that is, the interrelationships between people and ships, people and the environment, ships and ships, ships and the environment, and the environment and the environment; mainly including spatial relationships, temporal relationships, and semantic relationships.

[0077] B represents the behavioral elements of an object, that is, the dynamic changes of the attribute elements of an object or the relationship elements between objects over time. In collision avoidance rules, it can specifically refer to human behaviors (such as lookout, steering, etc.), ship behaviors (turning, changing speed, approaching and leaving, etc.), etc. The combination of multiple behavioral elements of entity elements over a time period constitutes event elements, such as berthing and leaving events, collision events. Behavioral elements can guide entity elements to perform complex actions under spatio-temporal elements

[0078] 2. Combining the structural characteristics of the water traffic scenario model in collision avoidance rules, use the finite state machine method for extension:

[0079] It is mainly extended from the following three aspects:

[0080] (1) Add a description mechanism for time and space characteristics; the extended finite state machine scenario = {E, A, R, B, T, S, W}, where T is the set of time conditions, S is the set of input conditions other than time and space, and W is defined as the set of trigger conditions;

[0081] (2) Expand the finite state machine into a hierarchical form (scenario1, scenario2,..., scenario n ), n ≥ 3;

[0082] (3) State transitions are allowed to exist in parallel. Expand the key feature set U of each scenario state, and the start and end times of the state need to be recorded; and expand the scenario state transition into a normal transition and a concurrent transition. All requirements for scenario feature changes must be met before a scenario state transition occurs; when the time condition does not belong to the time characteristics of the previous state, a normal transition is performed, otherwise a concurrent transition is performed.

[0083] 3. Based on the extended finite state machine method, conduct ontology modeling on ship collision avoidance rule scenarios and construct a formal description in the form of a scenario knowledge graph

[0084] Use ontology construction software to model the above-mentioned knowledge of the water traffic scenario of ship collision avoidance rules, further construct a scenario knowledge graph, store the scenario knowledge using the neo4j graph database, and visually display the scenario knowledge graph using d3.js software.

[0085] Referring to Figures 2 to 6 , the composition diagram of the scenario elements of a ship collision avoidance rule of the present invention analyzes the scenario knowledge of ship collision avoidance rules from four aspects, including entity elements, attribute elements, relationship elements, and behavior elements, and analyzes these types of elements layer by layer;

[0086] With time and location as the basic framework. Time and location describe the spatio-temporal characteristics of the scenario and are the prerequisite conditions for the existence of other dimensional characteristics. Among them, time is a measure describing the order of occurrence of the scenario state, and location records the space or position where the state occurs.

[0087] With attributes and behaviors as the core features. Attributes and behaviors describe the self-characteristics of the scenario and are an important manifestation of the development degree of the scenario. Among them, relatively static attribute features depict the form of the scenario; dynamic behavior features depict the evolution trend of the scenario.

[0088] Referring to Figure 7 , Figure 7 is an example diagram of the scenario knowledge graph of the ship collision avoidance rule of the present invention, stores the scenario knowledge using the neo4j graph database, and visually displays the scenario knowledge graph using d3.js software.

[0089] It should be understood that for those of ordinary skill in the art, improvements or transformations can be made according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of the present invention.

Claims

1. A method for formalizing the expression of ship collision avoidance rule scenarios, characterized in that, The method includes the following steps: Step 1: Obtain data of different water traffic scenarios and define the water traffic scenarios. Step 2: Conduct element analysis and semantic expression on the water traffic scenarios involved in ship collision avoidance rules. The ship collision avoidance rules include the International Regulations for Preventing Collisions at Sea, the Inland Navigation Rules, and local rules. Abstractly express the water traffic scenario as a conceptual model of a quadruple: scenario = {E, A, R, B}; that is, the water traffic scenario model includes the entities E involved in the water traffic scenario, the attributes A of the objects, the relationships R between the objects, and the behavioral elements B of the objects. Step 3: Combine the structural characteristics of the water traffic scenario model in the collision avoidance rules and use the finite state machine method for extension. Step 4: Based on the extended finite state machine method, conduct ontology modeling on the ship collision avoidance rule scenario knowledge and construct a formal description in the form of a scenario knowledge graph. The method in Step 4 specifically includes: Based on the ship collision avoidance rule knowledge extended by the extended finite state machine, the path from each complete input scenario to the output scenario main variable through the state transition function can be regarded as a directed path graph. Link each path graph to form a ship collision avoidance scenario knowledge graph. The specific method is as follows: According to the obtained scenario directed path graph Scenario input →Scenario output , further calculate the similarity of the directed path graph based on semantic information, which is used to splice the directed path graphs to construct a ship collision avoidance scenario knowledge graph; A directed path graph V Scenarioinput (i) and the semantic information similarity of another directed path graph V Scenariooutput (j) is: Among them, Ti and Tj are semantic information vectors of two directed graphs. Label propagation is performed on the graph. If the size of the set of directed path graphs is N(L), then the propagation probability P of V Scenarioinput (i) and V Scenariooutput (j) is defined as: Propagate according to the decreasing propagation probability until the data of the constructed set of directed path graphs is completely propagated to form a complete ship collision avoidance scenario knowledge graph.

2. The method for formalizing the expression of ship collision avoidance rule scenarios according to claim 1, characterized in that, The method in Step 1 specifically includes: Considering that the water traffic scenario has time continuity and spatial topology, define the water traffic scenario as a water complex with a specific structure and function formed by the mutual connection and interaction of various water traffic elements within a certain water area and different spatio-temporal ranges. The water traffic elements include the natural environment, navigation infrastructure, ships, and onboard personnel. The water traffic scenario includes six elements: time, space, traffic objects, things, phenomena, and events.

3. The method for formalizing the expression of ship collision avoidance rule scenarios according to claim 1, characterized in that, The method in Step 2 specifically includes: Abstractly express the water traffic scenario as a conceptual model of a quadruple: scenario = {E, A, R, B}; where scenario represents the water traffic scenario of the ship collision avoidance rules. E represents the entities involved in the water traffic scenario, that is, water traffic objects, including: people, ships, and environmental objects. A represents the attributes of the objects, that is, the corresponding properties of the objects, including: ship size, speed, course, environmental visibility, and traffic density. R represents the relationships between the objects, that is, the mutual correlation relationships between people and ships, people and the environment, ships and ships, ships and the environment, and the environment and the environment. B represents the behavioral elements of the objects, that is, the dynamic changes of the attribute elements of the objects or the relationship elements between the objects over time.

4. The method for formalizing the expression of ship collision avoidance rule scenarios according to claim 1, characterized in that, The method of using the finite state machine method for extension in Step 3 specifically includes: (1) Add a description mechanism for time and space characteristics; the extended finite state machine scenario = {E, A, R, B, T, S, W} is a set of time conditions, T is a set of space conditions, S is the input conditions other than time and space, and W is defined as a set of trigger conditions. (2) Expand the finite state machine into a hierarchical form (scenario1, Scenario2, …, scenario n ), n≥3; (3) State transitions are allowed to exist in parallel, expanding the key feature set U of each scenario state, recording the start and end times of the state; and expanding the scenario state transition into a regular transition and a concurrent transition. The scenario state transition must meet all the requirements of scenario feature changes before it occurs; when the time condition does not belong to the time feature of the previous state, a regular transition is made, otherwise a concurrent transition is made.

5. The method for formalizing the expression of ship collision avoidance rule scenarios according to claim 4, characterized in that, The method of expansion using the finite state machine method in step 3 specifically includes: The expanded finite state machine M is a 9-tuple (S, so, I, O, σ, γ, Y, F, T), where: (1)S represents a finite non-empty set of scenario states, i.e., Scenario = {Scenario1, Scenario2, …, Scenario n}; (2) so ∈ S represents the initial state of the attributes, relationships, and behaviors of the objects in the scenario of the finite state machine; (3) I and O represent the finite non-empty input symbol set and output symbol set respectively; (4) σ represents the state transition function; Using the maximum state-related ratio as the state transition path, as shown in the following formula: Among them, w is the state-related ratio, k1 and k2 are correction coefficients, S input is the number of input variables of the state, S output is the number of output variables related to the input state, and b is the number of input and output state variables; Design the state transition function, as shown in the following formula: where w i represents the relevant ratio of the i-th state variable, and z represents the total number of state transition paths; (5) γ represents the output function; Scenario output = γ(σ·Scenario input ) (3) Scenario output and Scenario input represent the output and input factor variables of the scenario, and such variables include eigenvectors and eigenvalues such as attributes, relationships, and behaviors that change over time and space; (6) Y represents the context, that is, the variable set of consecutive scenarios; (7) represents the set of termination states of the extended finite state machine; (8) T represents the finite non-empty state transition set, and any transition t in T is represented as a six-tuple (si, sj, i, o, p, a), where: 1) si, sj ∈ S represent the start state and end state of the transition t respectively; 2) i ∈ I and o ∈ O represent the input symbol and output symbol of the transition t respectively; 3) p represents the guard condition required for the transition t to occur, which is a logical expression; 4) a represents a set of assignment statements for modifying the values of context variables when the transition t occurs; The scenario transition process is described as: when the current state of the finite state machine M is S, if it receives the input symbol a, it will transfer to the state σ(S, a) and generate the output symbol γ(S, a), where the actual semantics of the input and output symbols are events, messages, or operations.

6. The formal expression method for ship collision avoidance rule scenario knowledge according to claim 1, characterized in that, This method also includes storing scenario knowledge using the neo4j graph database and visually displaying the scenario knowledge graph using the d3.js software.

Images