Key scene generation method and system for autonomous ship intelligent navigation test

Abstract

The invention discloses a key scene generation method for an autonomous ship intelligent navigation test. The key scene generation method comprises the following steps: identifying a natural navigation scene and a key navigation scene from ship navigation historical data, and constructing a database; the distribution of the natural navigation parameters of the ships is obtained through statistics from the natural navigation scene database, and the ship navigation parameters, the ship motion parameters and the collision risk degree between the ships in the key scene are obtained from the key navigation scene database; constructing an initial scene in the test scene based on the natural navigation parameter distribution of the ship; inputting parameters obtained in the key navigation scene database into a WGAN-GP model, and generalizing to form a key scene navigation parameter generation model; and according to the navigation parameters of the background ship and the external access tested ship in the kth step length, predicting and forming the navigation parameters of the background ship in the (k + 1) th step length, and iterating to form a continuous test scene. According to the invention, a scene with a higher probability of occurrence of an encounter situation can be established, and scene support is provided for test and evaluation of an intelligent navigation technology.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent ship autonomous navigation test scenario construction, and in particular to a key scenario generation method and system for autonomous ship intelligent navigation test. Background Art

[0002] In recent years, with the rapid development of technologies such as the Internet of Things, big data, and artificial intelligence, the application scope of autonomous ship unmanned systems has gradually expanded. Progressive ship intelligence based on digitalization, with intelligence as a breakthrough and autonomy as a goal has become a new trend and new hotspot in the development of the shipbuilding industry. Realizing the research and development of intelligent ship navigation function testing and evaluation technologies, testing autonomous ships, and verifying their performance in safety and reliability are important factors in ensuring that intelligent ships move from theory to reality.

[0003] The simulation test method can effectively reduce the test cost, avoid the occurrence of dangerous situations, and has higher test efficiency. The autonomous navigation simulation test is designed to verify the decision-making ability of the autonomous navigation system in a virtual environment. Considering the test cost and the flexibility of test scenario construction, the high-coverage scenario generation test method has become an important technical means to realize autonomous navigation testing.

[0004] Therefore, for the autonomous navigation test environment, how to extract situations, reconstruct scenarios, and derive real high-risk scenarios is a technical problem that needs to be solved urgently.

[0005] In view of this, it is necessary to propose a new technical solution to solve the above technical problems. Summary of the invention

[0006] The present invention is made to solve the above-mentioned problem, and aims to provide a key scenario generation method for autonomous ship intelligent navigation test.

[0007] The present invention provides a key scenario generation method for autonomous ship intelligent navigation test, which has the following characteristics and comprises the following steps: step 1, identifying natural navigation scenarios and key navigation scenarios from ship navigation history data, and respectively constructing a natural navigation scenario database and a key navigation scenario database; step 2, obtaining the distribution of ship natural navigation parameters from the natural navigation scenario database, and obtaining the ship navigation parameters, ship motion parameters and the collision risk model between ships in the key scenario from the key navigation scenario database; step 3, constructing an initial scenario in the test scenario based on the distribution of ship natural navigation parameters; step 4, inputting the ship navigation parameters, ship motion parameters and the collision risk model between ships in the key scenario into the WGAN-GP model, and generalizing to form a key scenario navigation parameter generation model; step 5, based on the key scenario navigation parameter generation model, according to the navigation parameters of the background ship within the kth step length and the externally connected tested ship obtained in the initial scenario, predicting the navigation parameters of the background ship within the k+1th step length, and iteratively forming continuous test scenarios.

[0008] In the key scenario generation method for autonomous ship intelligent navigation test provided by the present invention, it can also have the following characteristics: wherein, step 1 includes the following steps: step 1-1, collecting the navigation data of marine ships within a certain time and space, and pre-processing the acquired AIS historical data; step 1-2, based on the collision avoidance rules, identifying the key navigation scenes and natural navigation scenes according to the heading difference range and speed, distance and other parameters under different encounter situations. Wherein, in step 1-1, the pre-processing process is: decoding the AIS coded data, and performing data cleaning, timestamp conversion, linear interpolation to supplement trajectory points, Douglas-Peuker trajectory compression, and screening of valid navigation data on the decoded AIS data. Wherein, the valid navigation data at least includes the ship MMSI code, ship name, heading, speed, position and time information.

[0009] The key scenario generation method for autonomous ship intelligent navigation test provided by the present invention may also have the following features: wherein, in step 1-2, the key navigation scenario is a scenario in which there are three encounter situations, namely, head-on encounter, overtaking and cross encounter, during the navigation process of the ship, and the other normal navigation scenarios are natural navigation scenarios.

[0010] In the key scenario generation method for autonomous ship intelligent navigation test provided by the present invention, it can also have the following characteristics: wherein step 2 includes the following steps: step 2-1, based on the natural navigation scenario database, calculating the distribution of navigation parameters of the ship under natural navigation; step 2-2, based on the key navigation scenario database, calculating the ship motion parameters under different situations; step 2-3, calculating the collision risk under different situations, specifically: determining the collision risk assessment index between ships, establishing the index membership function matrix, using the hierarchical analysis method to determine the weight of each index, and constructing the collision risk function:

[0011]

[0012] The collision risk assessment indicators are DCPA, TCPA and the distance between ships, that is, I = {DCPA, TCPA, D}. The corresponding weight vector constructed by the indicators in set I is W = [w DCPA ,w TCPA ,w D ], where:

[0013] w DCPA +w TCPA +w D =1 (2)

[0014] The indicator membership function is:

[0015] U=[μ DCPA ,μ TCPA ,μ D ] (3)

[0016] In the formula,

[0017]

[0018] The key scenario generation method for autonomous ship intelligent navigation test provided by the present invention may also have the following features: wherein the navigation parameter distribution includes speed distribution, heading distribution and position distribution. The ship motion parameters include DCPA and TCPA.

[0019] The key scenario generation method for autonomous ship intelligent navigation test provided by the present invention may also have the following characteristics: wherein, in step 3, the construction of the initial scenario is based on the distribution of natural navigation parameters of the ship, and the Monte Carlo sampling method is used to obtain the initial state of the key scenario of the autonomous navigation test, including the number and position of background ships, the initial heading, the initial speed, and the position of the access ship to be tested.

[0020] In the key scenario generation method for autonomous ship intelligent navigation test provided by the present invention, it can also have the following characteristics: wherein, the specific implementation process of step 4 is as follows: step 4-1, in order to consider the mutual influence between ships and the spatiotemporal changes of navigation parameters, the navigation parameter difference between the background ship and the ship under test, the position of the ship and the collision risk are used to form a data set as the input of the key scenario navigation parameter generation model; step 4-2, the key scenario data set is trained using the WGAN-GP model, and generalized to form a key scenario navigation parameter generation model.

[0021] The key scenario generation method for autonomous ship intelligent navigation test provided by the present invention may also have the following features: wherein, in step 4-2, the calculation process of the key scenario navigation parameter generation model is:

[0022]

[0023] Where D(x) is the output value of the discriminator, G(z) is the output value of the generator; x~P r Indicates that the distribution of input sample x is P r , z~P g The distribution of the input noise z is P g , E[·] is the expected function.

[0024] The key scenario generation method for autonomous ship intelligent navigation test provided by the present invention may also have the following features: wherein, in step 5, the background ship navigation parameter sequence is:

[0025]

[0026] In the formula, is the maneuvering behavior parameter difference of the key ship at the k+1th step, v represents the speed, θ represents the heading, represents the maneuvering behavior of the measured ship in the k+1th step, represents the maneuvering behavior of the key ship in the k+1th step.

[0027] The present invention also provides a key scenario generation system for autonomous ship intelligent navigation test, which has the following characteristics, including: a database construction unit: identifying natural navigation scenarios and key navigation scenarios from historical ship navigation data, and constructing a natural navigation scenario database and a key navigation scenario database respectively; a ship parameter acquisition unit: statistically acquiring the distribution of natural navigation parameters of ships from the natural navigation scenario database, and acquiring the ship navigation parameters, ship motion parameters and the collision risk model between ships in the key scenario from the key navigation scenario database; an initial scenario construction unit: constructing the initial scenario in the test scenario based on the distribution of natural navigation parameters of ships; a key scenario navigation parameter model generation unit: inputting the ship navigation parameters, ship motion parameters and the collision risk model between ships in the key scenario into the WGAN-GP model, and generalizing them to form a key scenario navigation parameter generation model; a scenario generation unit: based on the key scenario navigation parameter generation model, according to the navigation parameters of the background ships and the externally connected tested ships within the kth step length obtained from the initial scenario, predicting the navigation parameters of the background ships within the k+1th step length, and iteratively forming continuous test scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flow chart of a key scenario generation method for autonomous ship intelligent navigation testing in an embodiment of the present invention;

[0029] Figure 2 This is a block diagram of a key scenario generation method for autonomous ship intelligent navigation testing in an embodiment of the present invention;

[0030] Figure 3 It is a schematic diagram of the network framework of the key scenario navigation parameter generation model provided in the embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to make the technical means, creative features, objectives and effects of the present invention easy to understand, the following embodiments are combined with the accompanying drawings to specifically describe a key scene generation method for autonomous ship intelligent navigation testing of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] In this embodiment, a key scenario generation method for autonomous ship intelligent navigation testing is provided.

[0033] Figure 1 It is a flow chart of a key scenario generation method for autonomous ship intelligent navigation testing in an embodiment of the present invention.

[0034] Figure 2It is a block diagram of the key scenario generation method for autonomous ship intelligent navigation test in an embodiment of the present invention.

[0035] like Figure 1-2 As shown, the key scenario generation method for autonomous ship intelligent navigation test in this embodiment includes the following steps:

[0036] Step 1: Identify natural navigation scenarios and key navigation scenarios from historical ship navigation data, and construct a natural navigation scenario database and a key navigation scenario database respectively. Specifically, the following steps are included:

[0037] Step 1-1, collect the navigation data of marine vessels within a certain time and space, and pre-process the acquired AIS historical data.

[0038] Step 1-2, based on the collision avoidance rules, identify key navigation scenarios and natural navigation scenarios according to the heading difference range, speed, distance and other parameters under different encounter situations.

[0039] In this embodiment, the preprocessing process is:

[0040] Decode the AIS coded data, and perform data cleaning, timestamp conversion, linear interpolation to supplement track points, Douglas-Peuker track compression, and filter effective navigation data on the decoded AIS data. Effective navigation data includes ship MMSI code, ship name, heading, speed, location, time and other information.

[0041] In this embodiment, the identification of key navigation scenes and natural navigation scenes is based on the International Regulations for Preventing Collisions at Sea, 1972.

[0042] Step 2: Statistically obtain the distribution of natural navigation parameters of ships from the natural navigation scenario database, and obtain the ship navigation parameters, ship motion parameters and collision risk model between ships in key scenarios from the key navigation scenario database.

[0043] The above navigation parameter distribution includes speed distribution, heading distribution and position distribution, and the ship motion parameters include DCPA and TCPA.

[0044] The specific implementation of step 2 is:

[0045] Step 2-1, based on the natural navigation scenario database, calculate the navigation parameter distribution of the ship under natural navigation;

[0046] Step 2-2, based on the key navigation scenario database, calculate the ship motion parameters under different situations;

[0047] Step 2-3, calculate the collision risk under different situations, specifically:

[0048] Determine the ship collision risk assessment index, establish the index membership function matrix, use the hierarchical analysis method to determine the weight of each index, and construct the collision risk function:

[0049]

[0050] In this embodiment, the collision risk assessment index is specifically DCPA, TCPA and the distance between ships, that is, I = {DCPA, TCPA, D}, and the corresponding weight vector constructed by the index in set I is W = [w DCPA ,w TCPA ,w D ], where:

[0051] w DCPA +w TCPA +w D =1 (2)

[0052] The indicator membership function is:

[0053] U=[μ DCPA ,μ TCPA ,μ D ] (3)

[0054] Where:

[0055]

[0056] Step 3: construct the initial scene in the test scenario based on the distribution of natural navigation parameters of ships. In this embodiment, the construction of the initial scene is based on the distribution of natural navigation parameters of ships, and the Monte Carlo sampling method is used to obtain the initial state of the key scene of the autonomous navigation test, including the number, position, initial heading, initial speed of background ships, and the position of the access ship under test.

[0057] Step 4: Input the ship navigation parameters, ship motion parameters and collision risk model between ships in the key scenarios into the WGAN-GP model to generalize the key scenario navigation parameter generation model. The specific implementation process is as follows:

[0058] Step 4-1: In order to consider the mutual influence between ships and the spatiotemporal changes of navigation parameters, the navigation parameter difference between the background ship and the measured ship, the position of the ship and the collision risk constitute a data set as the input of the key scene navigation parameter generation model;

[0059] In step 4-2, the WGAN-GP model is used to train the key scenario dataset and generalize it to form a key scenario navigation parameter generation model.

[0060] Figure 3 It is a schematic diagram of the network framework of the key scenario navigation parameter generation model provided in the embodiment of the present invention.

[0061] like Figure 3 As shown, the calculation process of the key scenario navigation parameter generation model in this embodiment is:

[0062]

[0063] Where D(x) is the output value of the discriminator, G(z) is the output value of the generator; x~P r Indicates that the distribution of input sample x is P r , z~P g The distribution of the input noise z is P g , E[·] is the expected function.

[0064] Step 5: Generate a model based on the navigation parameters of key scenarios. According to the navigation parameters of the background ships and the externally connected ships under test within the kth step obtained in the initial scenario, predict the navigation parameters of the background ships within the k+1th step, and iterate to form continuous test scenarios.

[0065] In this embodiment, the background ship navigation parameter sequence is:

[0066]

[0067] In the formula, is the maneuvering behavior parameter difference of the key ship at the k+1th step, v represents the speed, θ represents the heading, represents the maneuvering behavior of the measured ship in the k+1th step, represents the maneuvering behavior of the key ship in the k+1th step.

[0068] This embodiment also provides a key scenario generation system for autonomous ship intelligent navigation testing, including:

[0069] Database construction unit: constructing a natural navigation scenario database and a key navigation scenario database using the method in step 1 of this embodiment;

[0070] Ship parameter acquisition unit: using the method in step 2 of this embodiment to acquire ship parameters;

[0071] Initial scene construction unit: constructing the initial scene using the method in step 3 of this embodiment;

[0072] A key scenario navigation parameter model generation unit: using the method in step 4 of this embodiment to form a key scenario navigation parameter generation model;

[0073] Scenario generation unit: Generate a test scenario using the method in step 5 of this embodiment.

[0074] Functions and Effects of the Embodiments

[0075] The key scenario generation method for the intelligent navigation test of autonomous ships involved in this embodiment can generate adaptive high-risk test scenarios in real time according to the navigation status of the tested ship, effectively improving the test and evaluation efficiency of the autonomous navigation capability of intelligent ships. Identify the characteristics of high-risk key scenarios for ship navigation, build a test scenario generation model by changing the maneuvering behavior of the background ship, predict the possible situations of the tested ship and the key ship in subsequent movements, use the improved generative adversarial network to generate key test scenarios for autonomous navigation for simulation, improve the generation probability under different encounter situations during the test process, and realize the construction of high-risk key test scenarios for ship intelligent navigation. While ensuring the authenticity of the scenario, it can increase the probability of high-risk scenarios, provide technical support for solving the real-time generation of intelligent ship autonomous navigation test scenarios, and is of great significance for realizing real-time testing of autonomous navigation capabilities.

[0076] The above-mentioned embodiments are preferred examples of the present invention and are not intended to limit the protection scope of the present invention.

Claims

1. A key scenario generation method for autonomous ship intelligent navigation test, characterized in that: The following steps are involved: Step 1: Identify natural navigation scenarios and key navigation scenarios from historical ship navigation data, and construct a natural navigation scenario database and a key navigation scenario database respectively; Step 2, obtaining the distribution of natural navigation parameters of ships from the natural navigation scenario database, and obtaining the navigation parameters, motion parameters and collision risk model of ships in key scenarios from the key navigation scenario database; Step 3, constructing an initial scene in the test scene based on the natural navigation parameter distribution of the ship; Step 4, inputting the ship navigation parameters, the ship motion parameters and the ship collision risk model in the key scene into the WGAN-GP model, and generalizing them to form a key scene navigation parameter generation model; Step 5, based on the key scenario navigation parameter generation model, according to the navigation parameters of the background ship in the kth step and the external access ship under test obtained in the initial scenario, the navigation parameters of the background ship in the k+1th step are predicted, and continuous test scenarios are formed iteratively.

2. The key scenario generation method for autonomous ship intelligent navigation test according to claim 1 is characterized by: in, Step 1 includes the following steps: Step 1-1, collecting the navigation data of marine vessels within a certain time and space, and pre-processing the acquired AIS historical data; Step 1-2: Based on the collision avoidance rules, identify key navigation scenarios and natural navigation scenarios according to the heading difference range, speed, distance and other parameters under different encounter situations. Among them, in step 1-1, the preprocessing process is: Decode the AIS coded data, and perform data cleaning, time stamp conversion, linear interpolation to supplement track points, Douglas-Peuker track compression, and screening of valid navigation data on the decoded AIS data. The valid navigation data includes at least the ship's MMSI code, ship name, heading, speed, location and time information.

3. The key scenario generation method for autonomous ship intelligent navigation test according to claim 2 is characterized in that: in, In step 1-2, the key navigation scenarios are scenarios in which there are three encounter situations during the navigation of ships, namely, encounter, overtaking and crossing encounter, and the other normal navigation scenarios are natural navigation scenarios.

4. The key scenario generation method for autonomous ship intelligent navigation test according to claim 1, Features: Wherein, step 2 comprises the following steps: Step 2-1, calculating the distribution of navigation parameters of the ship under natural navigation based on the natural navigation scenario database; Step 2-2, calculating the ship motion parameters under different situations based on the key navigation scenario database; Step 2-3, calculate the collision risk under different situations, specifically: Determine the ship collision risk assessment index, establish the index membership function matrix, use the hierarchical analysis method to determine the weight of each index, and construct the collision risk function: The collision risk assessment index is specifically DCPA, TCPA and the distance between ships, that is, I = {DCPA, TCPA, D}, and the corresponding weight vector constructed by the index in set I is W = [w DCPA ,w TCPA ,w D ], where: In DCPA +in TCPA +in D =1 (2) The indicator membership function is: U=[μ DCPA ,m TCPA ,m D ] (3) In the formula, 5. The key scenario generation method for autonomous ship intelligent navigation test according to claim 4 is characterized in that: in, The navigation parameter distribution includes speed distribution, heading distribution and position distribution, and the ship motion parameters include DCPA and TCPA.

6. The key scenario generation method for autonomous ship intelligent navigation test according to claim 1 is characterized by: in, In step 3, the construction of the initial scene is based on the distribution of the natural navigation parameters of the ship, and the Monte Carlo sampling method is used to obtain the initial state of the key scene of the autonomous navigation test, including the number, position, initial heading, initial speed of background ships, and the position of the access ship under test.

7. The key scenario generation method for autonomous ship intelligent navigation test according to claim 1 is characterized by: in, The specific implementation process of step 4 is as follows: Step 4-1: In order to consider the mutual influence between ships and the spatiotemporal changes of navigation parameters, the navigation parameter difference between the background ship and the measured ship, the position of the ship and the collision risk constitute a data set as the input of the key scene navigation parameter generation model; In step 4-2, the WGAN-GP model is used to train the key scenario dataset and generalize it to form a key scenario navigation parameter generation model.

8. The key scenario generation method for autonomous ship intelligent navigation test according to claim 7 is characterized by: in, In step 4-2, the calculation process of the key scenario navigation parameter generation model is: Where D(x) is the output value of the discriminator, G(z) is the output value of the generator; x~P r Indicates that the distribution of input sample x is P r , z~P g The distribution of the input noise z is P g , E[·] is the expected function.

9. The key scenario generation method for autonomous ship intelligent navigation test according to claim 1 is characterized in that: in, In step 5, the background ship navigation parameter sequence is: In the formula, is the maneuvering behavior parameter difference of the key ship at the k+1th step, v represents the speed, θ represents the heading, represents the maneuvering behavior of the measured ship in the k+1th step, represents the maneuvering behavior of the key ship in the k+1th step.

10. A key scenario generation system for autonomous ship intelligent navigation testing, characterized in that: include: Database construction department: Identify natural navigation scenarios and key navigation scenarios from historical ship navigation data, and construct natural navigation scenario database and key navigation scenario database respectively; Ship parameter acquisition unit: statistically acquires the distribution of ship natural navigation parameters from the natural navigation scenario database, and acquires ship navigation parameters, ship motion parameters and collision risk model between ships in key scenarios from the key navigation scenario database; An initial scene construction unit: constructing an initial scene in a test scene based on the natural navigation parameter distribution of the ship; Key scenario navigation parameter model generation unit: input the ship navigation parameters, the ship motion parameters and the collision risk model between ships in the key scenario into the WGAN-GP model, and generalize them to form a key scenario navigation parameter generation model; Scenario generation unit: Based on the key scenario navigation parameter generation model, the navigation parameters of the background ship within the kth step and the externally connected ship under test are obtained from the initial scenario, and the navigation parameters of the background ship within the k+1th step are predicted to iterate and form continuous test scenarios.

Images