A ship collision avoidance decision trigger timing modeling method imitating human social cognitive processes

Abstract

The application provides a ship collision avoidance decision trigger opportunity modeling method which imitates human social cognitive process, comprising: obtaining ship AIS data, extracting the longitude, latitude, heading, speed and length information of the ship in the AIS data, and constructing an eccentric ellipse ship field; calculating the remaining distance DDV of a target ship TS before invading the own ship OS field, the remaining time TDV of the target ship TS before invading the own ship OS field, and the distance D between the ships; constructing a drift rate function, forming the unit time information accumulation and the current time information accumulation total amount; constructing a decision boundary, and forming a decision trigger result based on the information accumulation and the intervention scenario. The application is based on the variable drift diffusion model in the human cognitive modeling field, learns from the unique evidence accumulation mechanism in the human cognitive process, converts the instantaneous risk parameter into the evidence accumulation amount under the unit time, replaces the threshold determination trigger method into the mode whether the evidence amount reaches the trigger boundary, and imitates the human social cognitive decision mode by introducing the safe decision length.

Description

Technical Field

[0001] This invention relates to the field of maritime traffic management technology, and more particularly to a method for modeling the timing of ship collision avoidance decisions that mimics the cognitive processes of human society. Background Technology

[0002] Collision avoidance decision-making by autonomous vessels is a crucial task related to their navigational safety. The nature of the collision avoidance problem dictates that a vessel's navigational decision-making behavior is not solely governed by its own rational logic, but must also consider the cognitive characteristics of other vessels during encounters, ensuring that collision avoidance decisions are socially acceptable while guaranteeing navigational safety. The decision trigger model, acting as a switch for autonomous vessel collision avoidance, directly determines whether to take collision avoidance action and the timing of formulating collision avoidance strategies, making it a key issue in the collision avoidance process.

[0003] Currently, most collision avoidance decision-triggering models rely on real-time calculations of inter-ship collision risks, comparing these risks with pre-set risk thresholds to determine the timing of a decision. This model is susceptible to fluctuations in instantaneous data, leading to unstable collision avoidance decision triggering. Furthermore, decision triggering based on risk thresholds is a form of ship automation rather than autonomy, which differs significantly from human cognitive decision-making processes. Consequently, the resulting collision avoidance behavior is difficult for other vessels to accept. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for modeling the triggering timing of ship collision avoidance decisions, mimicking the human social cognitive process. This invention designs a trigger, VDDM-CADT (the specific Variable Drift Diffusion Model for ship Collision Avoidance Decision Trigger), which is based on the Variable Drift Diffusion Model (VDDM) commonly used in human cognitive modeling. It draws upon the evidence accumulation mechanism unique to human cognition, converting instantaneous risk parameters into the amount of evidence accumulated per unit time. The threshold-based triggering method is replaced with a mode based on whether the amount of evidence reaches the triggering boundary. By introducing the concept of safety decision duration, the trigger can mimic the human social cognitive decision-making pattern.

[0005] The technical means employed in this invention are as follows:

[0006] A method for modeling the triggering timing of ship collision avoidance decisions, mimicking human social cognitive processes, includes:

[0007] S1. Obtain ship AIS data, extract the ship's latitude, longitude, heading, speed and length information from the AIS data, and construct an eccentric elliptical ship domain.

[0008] S2. Based on the constructed eccentric elliptical ship domain, calculate the remaining distance DDV before another ship's TS invades the ship's OS domain, the remaining time TDV before another ship's TS invades the ship's OS domain, and the inter-ship distance D.

[0009] S3. Construct the drift rate function f[i] to form the cumulative amount of information per unit time and the total cumulative amount of information at the current moment;

[0010] S4. Construct decision boundaries and generate decision-triggered results based on the accumulated amount of information and the intervention scenario.

[0011] Furthermore, in step S1, the eccentric elliptical ship domain is constructed, specifically including:

[0012] The major axis of the ellipse is parallel to the ship's heading (OS). The semi-major axis of the ellipse is 'a', and the semi-minor axis is 'b'. The displacement of the ship from the center of the ellipse along the semi-major axis to the stern is Δa, and the displacement of the ship from the center of the ellipse along the semi-minor axis to the left is Δb. The values ​​of a, b, Δa, and Δb are set as follows:

[0013] a = 10l;

[0014] b = 5l;

[0015] Δa = 2.5l;

[0016] Δb = 1.25l;

[0017] Where l represents the length of the ship.

[0018] Further, step S2 specifically includes:

[0019] S21. The scaling factor f is based on the maximum extent to which another ship's TS intrudes into the ship's OS domain. min Calculate the remaining distance DDV before another vessel's TS intrudes into the OS domain of this vessel. The calculation formula is as follows:

[0020] DDV = max(1-f min ,0)

[0021] Where DDV is determined by the domain scaling factor f, and f is a quadratic function f(t) of time t, then, given that f has a solution, we obtain the scaling factor f when another ship's TS invades the domain of this ship to the maximum extent. min Specifically, f min If f > 1, it proves that other ships' TS will never invade the OS domain of this ship; if f min =1, which means that the other ship's TS just cuts through the boundary of this ship's OS; if f min <1 indicates that other ships' TS will inevitably intrude into the territory of this ship's OS;

[0022] S22. Calculate the remaining time TDV before another vessel's TS intrudes into the vessel's OS domain. The calculation formula is as follows:

[0023] f(t) = 1

[0024] TDV is obtained by solving the formula f(t) = 1.

[0025] Further, step S3 specifically includes:

[0026] S31, the drift rate includes the theoretical instantaneous safe decision time (TISDT) and the theoretical standard safe decision time (TSSDT), where K i This represents the duration of the safety decision at any given time i. The maximum safety decision duration for the current TDV sub-interval is represented by the following formula:

[0027] K i =h ψ (DDV i D i )

[0028]

[0029] Among them, h ψ It is the parsing function of SDT, DDV i D represents the DDV value at time i. i This represents the distance between the two ships at time i;

[0030] S32. Calculate the evidence accumulation direction g[i] at time i, using the following formula:

[0031]

[0032] Among them, DDV i crit express The corresponding DDV value;

[0033] S33. Based on steps S31 and S32, the drift rate function f[i] is defined as follows:

[0034]

[0035] Where Δt is the time step, λ is the tuning factor, ε is a constant, and ∑t i Let |∑t| be the cumulative decision time at the current time i. i -K i | Reflects the rate at which evidence accumulates;

[0036] S34. Construct the information accumulation model of VDDM-CADT, using the following formula:

[0037] ΔX[i] = X[i] - X[i-1]

[0038] =ωf[i]+W[i]

[0039] Where X[i] represents the total amount of information accumulated at time i, ΔX[i] represents the amount of information accumulated per unit time with a time step of Δt, where Δt is in seconds, and W[i] represents the diffusion noise; it is assumed that the increment of the diffusion part follows a standard Wiener process, indicating that the change of the diffusion increment in any finite time follows a normal distribution, expressed as W[i]~N(0,Δtσ) 2 ω represents the scale parameter, reflecting the proportion of the remaining accumulated information at the distance to the trigger threshold to the current trigger threshold. The closer the accumulated information value is to the trigger threshold at the current moment, the smaller the scale factor will be assigned to the calculated ΔX[i]. The formula for calculating the scale parameter ω is as follows:

[0040]

[0041] Among them, B i Let α represent the information accumulation boundary at time i, where α is the power function parameter and α∈(0,1);

[0042] S35. When T0 is set as the initial time for information accumulation, calculate T using Δt as the unit time length. d The formula for calculating the accumulated information at each moment is as follows:

[0043]

[0044] Furthermore, the theoretical instantaneous safety decision duration TISDT and the theoretical standard safety decision duration TSSDT in step S31 are both related to the safety decision duration SDT. SDT represents the upper limit of the collision avoidance decision triggering time for the ship's navigator under the premise of ensuring navigational safety, that is, the maximum time to determine whether to make a collision avoidance decision. The setting of the safety decision duration SDT is related to the remaining distance DDV, inter-ship distance D, and TDV before another vessel TS intrudes into the ship's OS domain, as follows:

[0045] The remaining distance DDV before another vessel's TS intrudes into the vessel's OS and the inter-ship distance D represent the urgency of the current situation. That is, the clearer the urgency of the current situation, the shorter the SDT will be, and vice versa. The greater the remaining time TDV before another vessel's TS intrudes into the vessel's OS, the more time the human operator has to make a safety decision, and the longer the SDT will be.

[0046] Furthermore, the specific calculation process for the Security Decision Time (SDT) includes:

[0047] The remaining time TDV before another vessel's TS intrudes into the OS domain of this vessel, the remaining distance DDV before another vessel's TS intrudes into the OS domain of this vessel, and the distance D are limited to specific ranges and intervals. The unit of the remaining time TDV before another vessel's TS intrudes into the OS domain of this vessel is hours, and the unit of D is nautical miles.

[0048] Divide the remaining time TDV before another ship's TS invades the ship's OS domain into six sub-intervals [TDV] j ,TDV j+1 For each group of other vessels' TS before intruding into the vessel's OS domain, based on the survey and analysis results of the ship's drivers, the remaining distance DDV before intruding into the vessel's OS domain and the proportion of remaining time TDV before intruding into the vessel's OS domain under different ship-to-ship distances D are calculated respectively.

[0049] The analytical expression for the Safety Decision Duration (SDT) is obtained through data fitting. Based on the actual values ​​of the independent variables (specific values ​​of DDV and D), the SDT is calculated as follows:

[0050]

[0051] Among them, TDV j+1 This represents the TDV value at time j+1. j Represents the TDV value at time j, DDV j+1 This represents the DDV value at time j+1. j D represents the DDV value at time j. j+1 D represents the distance between the two ships at time j+1. j Let j represent the distance between the two ships at time j.

[0052] Furthermore, assuming that the numerical expression of the Security Decision Duration (SDT) is a special Gaussian smooth function, namely the bulge function, is defined as:

[0053]

[0054] Where A represents the amplitude of the bulge function, i.e. the highest value at the center point, r represents the center radius of the bulge function, (x0, y0) represents the center point of the bulge function, and d represents the minimum margin of the bulge function.

[0055] Based on the numerical expression of the Security Decision Duration (SDT), the SDT calculation formula SDT(DDV) is obtained. i D i ),as follows:

[0056] when Then:

[0057]

[0058] when Then:

[0059] SDT(DDV i D i ) = SDT min

[0060] Where DDV0 and D0 represent the coordinates of the center point of the bulge function; SDT max This represents the proportion of the safe decision time (SDT) to the remaining time (TDV) before another vessel's TS intrudes into the vessel's OS, under the most complex combination of distances between ships (DDV and D) before the other vessel's TS intrudes into the vessel's OS. min This represents the proportion of the minimum decision-making time for collision avoidance to the remaining time TDV before another vessel's TS intrudes into the vessel's OS domain; r SDT Represents the radius of the bulge function; DDV e and D e ′ are the center and radius of the bulge function, respectively;

[0061] Since the range of the inter-ship distance D is not consistent with the remaining distance DDV before other ships' TS intrudes into the OS domain of this ship, it is necessary to map the value of the inter-ship distance D from the interval [2,6] to the interval [0,1], specifically as follows:

[0062] D′=(D-2) / 4.

[0063] Further, step S4 specifically includes:

[0064] S41. Construct the decision boundary, which is a function of TDV. The design references the Sigmoid activation function, which has the characteristic of compressing a large numerical range into the interval (0, 1), and is defined as follows:

[0065]

[0066] Where b0 is the parameter for adjusting the boundary size, k is the sensitivity of the boundary to TDV, μ is the critical value of the boundary at the baseline, and k0 is the minimum value of the boundary constraint.

[0067] S42. Calculate the decision trigger result based on the accumulated information. The calculation formula is as follows:

[0068]

[0069] in, X is the outcome triggered by the intervention in the decision. i B(TDV) represents the total amount of information accumulated at the current moment.i ) represents the size of the information boundary at the current moment, μ is the scaling factor, and C1 and C2 are the decision to make a collision avoidance decision and the decision not to make a collision avoidance decision, respectively;

[0070] S43, Based on intervention scenarios ξ i The decision trigger result is calculated using the following formula:

[0071]

[0072] Among them, intervention scenario ξ i Including Condi_1 and Condi_2, Condi_1 is defined as the total accumulated information duration at the current time being greater than or equal to the current TISDT, Condi_1: Σt i ≥K i Condi_2 is defined as the total change in TISDT between the α sampling times prior to time i and the current time being less than or equal to the actual accumulated time, where Condi_2:K i-α -K i ≤αΔt;

[0073] S44, When intervening in scenario ξ i If both Condi_1 and Condi_2 are satisfied, execute C1; otherwise, execute C2.

[0074] Compared with the prior art, the present invention has the following advantages:

[0075] 1. The present invention provides a modeling method for ship collision avoidance decision triggering timing that mimics the cognitive process of human society. The constructed decision triggering model does not rely on the instantaneous value of collision risk parameters between ships, but is modeled according to the trend of collision risk changes within the safety decision time interval. The decision triggering timing calculated in this way is a stable output obtained after a certain observation period, thus avoiding the randomness and volatility problems in previous triggering calculations.

[0076] 2. The present invention provides a modeling method for ship collision avoidance decision triggering timing that mimics the cognitive process of human society. For the first time, it models the collision avoidance decision triggering mechanism of autonomous ships based on the Variable Drift Diffusion Model (VDDM). The decision triggering model that integrates the VDDM framework can approximate the cognitive decision-making process of human drivers.

[0077] 3. This invention provides a modeling method for ship collision avoidance decision triggering timing that mimics the cognitive process of human society. It explores a feasible path for integrating cognitive models into autonomous navigation systems, considers the necessity for autonomous ships to make human-like decisions in mixed scenarios, and provides certain theoretical and technical support for the socialized operation of autonomous ships.

[0078] Based on the above reasons, this invention can be widely applied in fields such as maritime transportation. Attached Figure Description

[0079] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0080] Figure 1 This is a flowchart of the method of the present invention.

[0081] Figure 2 This is a schematic diagram of an eccentric elliptical ship constructed according to the present invention.

[0082] Figure 3 The SDT function graphs under different TDV intervals provided by this invention.

[0083] In the figure: (a) TDV ∈ [0.1, 0.2); (b) TDV ∈ [0.2, 0.3); (c) TDV ∈ [0.3, 0.4); (d) TDV ∈ [0.4, 0.5); (e) TDV ∈ [0.5, 0.6); (f) TDV ∈ [0.6, 0.7);

[0084] Figure 4 A schematic diagram of a ship encounter situation provided by the present invention.

[0085] In the image: (a) Scene I; (b) Scene II; (c) Scene III;

[0086] Figure 5 A schematic diagram of the initial parameters of a ship provided by the present invention.

[0087] Figure 6 This is a schematic diagram illustrating the changes in domain parameters under scenario I provided by the present invention.

[0088] In the figure: (a) DDV change record of TS-1; (b) TDV change record of TS-1;

[0089] Figure 7 This is a schematic diagram of domain parameter changes under scenario II provided by the present invention.

[0090] In the figure: (a) DDV change record of TS-1; (b) TDV change record of TS-1; (c) DDV change record of TS-2; (d) TDV change record of TS-2;

[0091] Figure 8 This is a schematic diagram of domain parameter changes under scenario III provided by the present invention.

[0092] In the figure: (a) DDV change record of TS-1; (b) TDV change record of TS-1; (c) DDV change record of TS-2; (d) TDV change record of TS-2; (e) DDV change record of TS-3; (f) TDV change record of TS-3;

[0093] Figure 9 This is a diagram showing the accumulation of model information and the recording of boundary changes for scenario I provided by the present invention.

[0094] Figure 10 This is a diagram showing the accumulation of model information and the recording of boundary changes for scenario II provided by the present invention.

[0095] The diagram shows: the information accumulation and triggering process of TS-1; the information accumulation and triggering process of TS-2.

[0096] Figure 11 This is a diagram showing the accumulation of model information and the recording of boundary changes for Scenario III provided by the present invention.

[0097] The diagram shows: (a) the information accumulation and triggering process of TS-1; (b) the information accumulation and triggering process of TS-2; and (c) the information accumulation and triggering process of TS-3. Detailed Implementation

[0098] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0099] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0100] like Figure 1As shown, this invention provides a method for modeling the triggering timing of ship collision avoidance decisions, which mimics the cognitive process of human society, including:

[0101] S1. Obtain ship AIS data, extract the ship's latitude, longitude, heading, speed and length information from the AIS data, and construct an eccentric elliptical ship domain.

[0102] S2. Based on the constructed eccentric elliptical ship domain, calculate the remaining distance DDV before another ship's TS invades the ship's OS domain, the remaining time TDV before another ship's TS invades the ship's OS domain, and the inter-ship distance D.

[0103] S3. Construct the drift rate function f[i] to form the cumulative amount of information per unit time and the total cumulative amount of information at the current moment;

[0104] S4. Construct decision boundaries and generate decision-triggered results based on the accumulated amount of information and the intervention scenario.

[0105] In a specific implementation, as a preferred embodiment of the present invention, step S1, constructing the eccentric elliptical ship domain, specifically includes:

[0106] The major axis of the ellipse is parallel to the ship's (OS) heading. The semi-major axis of the ellipse is 'a', and the semi-minor axis is 'b'. The displacement of the ship from the center of the ellipse along the semi-major axis to the stern is Δa, and the displacement of the ship from the center of the ellipse along the semi-minor axis to the left is Δb. Values ​​for a, b, Δa, and Δb are set as follows: Figure 2 As shown, the details are as follows:

[0107] a = 10l;

[0108] b = 5l;

[0109] Δa = 2.5l;

[0110] Δb = 1.25l;

[0111] Where l represents the length of the ship.

[0112] In a specific implementation, as a preferred embodiment of the present invention, step S2 specifically includes:

[0113] S21. The scaling factor f is based on the maximum extent to which another ship's TS intrudes into the ship's OS domain. min Calculate the remaining distance DDV before another vessel's TS intrudes into the OS domain of this vessel. The calculation formula is as follows:

[0114] DDV = max(1-f min ,0)

[0115] Where DDV is determined by the domain scaling factor f, and f is a quadratic function f(t) of time t, then, given that f has a solution, we obtain the scaling factor f when another ship's TS invades the domain of this ship to the maximum extent. min Specifically, f min If f > 1, it proves that other ships' TS will never invade the OS domain of this ship; if f min =1, which means that the other ship's TS just cuts through the boundary of this ship's OS; if f min <1 indicates that other ships' TS will inevitably intrude into the territory of this ship's OS;

[0116] S22. Calculate the remaining time TDV before another vessel's TS intrudes into the vessel's OS domain. The calculation formula is as follows:

[0117] f(t) = 1

[0118] TDV is obtained by solving the formula f(t) = 1.

[0119] In a specific implementation, as a preferred embodiment of the present invention, step S3 specifically includes:

[0120] S31, the drift rate includes the theoretical instantaneous safe decision time (TISDT) and the theoretical standard safe decision time (TSSDT), where K i This represents the duration of the safety decision at any given time i. The maximum safety decision duration for the current TDV sub-interval is represented by the following formula:

[0121] K i =h ψ (DDV i D i )

[0122]

[0123] Among them, h ψ It is the parsing function of SDT, DDV i D represents the DDV value at time i. i This represents the distance between the two ships at time i.

[0124] S32. Calculate the evidence accumulation direction g[i] at time i, using the following formula:

[0125]

[0126] Among them, DDV i crit express The corresponding DDV value;

[0127] S33. Based on steps S31 and S32, the drift rate function f[i] is defined as follows:

[0128]

[0129] Where Δt is the time step in seconds, λ is the tuning factor, ε is a constant, and ∑t i Let |∑t| be the cumulative decision time at the current time i. i -K i | Reflects the rate at which evidence accumulates;

[0130] S34. Construct the information accumulation model of VDDM-CADT, using the following formula:

[0131] ΔX[i] = X[i] - X[i-1]

[0132] =ωf[i]+W[i]

[0133] Where X[i] represents the total amount of information accumulated at time i, ΔX[i] represents the amount of information accumulated per unit time with a time step of Δt, where Δt is in seconds, and W[i] represents the diffusion noise; it is assumed that the increment of the diffusion part follows a standard Wiener process, indicating that the change of the diffusion increment in any finite time follows a normal distribution, expressed as W[i]~N(0,Δtσ) 2 ω represents the scale parameter, reflecting the proportion of the remaining accumulated information at the distance to the trigger threshold to the current trigger threshold. The closer the accumulated information value is to the trigger threshold at the current moment, the smaller the scale factor will be assigned to the calculated ΔX[i]. The formula for calculating the scale parameter ω is as follows:

[0134]

[0135] Among them, B i Let α represent the information accumulation boundary at time i, where α is the power function parameter and α∈(0,1);

[0136] S35. When T0 is set as the initial time for information accumulation, calculate T using Δt as the unit time length. d The formula for calculating the accumulated information at each moment is as follows:

[0137]

[0138] In specific implementation, as a preferred embodiment of the present invention, the theoretical instantaneous safety decision duration TISDT and the theoretical standard safety decision duration TSSDT in step S31 are both related to the safety decision duration SDT. SDT represents the upper limit of the collision avoidance decision triggering time for the ship's navigator under the premise of ensuring navigational safety, i.e., the maximum time for determining whether to make a collision avoidance decision. The setting of the safety decision duration SDT is related to the remaining distance DDV, inter-ship distance D, and TDV before another vessel's TS intrudes into the ship's OS domain, as detailed below:

[0139] The remaining distance DDV before another vessel's TS intrudes into the vessel's OS and the inter-ship distance D represent the urgency of the current situation. That is, the clearer the urgency of the current situation, the shorter the SDT will be, and vice versa. The greater the remaining time TDV before another vessel's TS intrudes into the vessel's OS, the more time the human operator has to make a safety decision, and the longer the SDT will be.

[0140] In specific implementation, as a preferred embodiment of the present invention, the specific calculation process of the Security Decision Duration (SDT) includes:

[0141] The remaining time TDV before another vessel's TS intrudes into the OS domain of this vessel, the remaining distance DDV before another vessel's TS intrudes into the OS domain of this vessel, and the distance D are limited to specific ranges and intervals. The unit of the remaining time TDV before another vessel's TS intrudes into the OS domain of this vessel is hours, and the unit of D is nautical miles.

[0142] Divide the remaining time TDV before another ship's TS invades the ship's OS domain into six sub-intervals [TDV] j ,TDV j+1 For each group of other vessels' TS before intruding into the vessel's OS domain, based on the survey and analysis results of the ship's drivers, the remaining distance DDV before intruding into the vessel's OS domain and the proportion of remaining time TDV before intruding into the vessel's OS domain under different ship-to-ship distances D are calculated respectively.

[0143] The analytical expression for the Security Decision Duration (SDT) is obtained through data fitting, and the SDT is calculated based on the actual values ​​of the independent variables (specific values ​​of DDV and D), such as... Figure 3 As shown, the details are as follows:

[0144]

[0145] Among them, TDV j+1 This represents the TDV value at time j+1. j Represents the TDV value at time j, DDV j+1 This represents the DDV value at time j+1. jD represents the DDV value at time j. j+1 D represents the distance between the two ships at time j+1. j Let j represent the distance between the two ships at time j.

[0146] In specific implementation, as a preferred embodiment of the present invention, this embodiment considers that the human cognitive decision-making time reflects the degree of entanglement in judging things. Therefore, the closer to the intermediate state of the independent variable, the larger the value of SDT should be. In addition, for the value of the independent variable exceeding a certain limit, SDT should also take into account the shortest human decision-making time, that is, the shortest time to immediately make a collision avoidance decision and trigger a judgment. The present invention assumes that the numerical expression of the safety decision-making time SDT is a special Gaussian smooth function, that is, a bulge function, defined as:

[0147]

[0148] Where A represents the amplitude of the bulge function, i.e. the highest value at the center point, r represents the center radius of the bulge function, (x0, y0) represents the center point of the bulge function, and d represents the minimum margin of the bulge function.

[0149] Based on the numerical expression of the Security Decision Duration (SDT), the SDT calculation formula SDT(DDV) is obtained. i D i ),as follows:

[0150] when Then:

[0151]

[0152] when Then:

[0153] SDT(DDV i D i ) = SDT min

[0154] Where DDV0 and D0 represent the coordinates of the center point of the bulge function; SDT max This represents the proportion of the safe decision time (SDT) to the remaining time (TDV) before another vessel's TS intrudes into the vessel's OS, under the most complex combination of distances between ships (DDV and D) before the other vessel's TS intrudes into the vessel's OS. min This represents the proportion of the minimum decision-making time for collision avoidance to the remaining time TDV before another vessel's TS intrudes into the vessel's OS domain; r SDT Represents the radius of the bulge function; DDV e and D e ′ are the center and radius of the bulge function, respectively;

[0155] Since the range of the inter-ship distance D is not consistent with the remaining distance DDV before other ships' TS intrudes into the OS domain of this ship, it is necessary to map the value of the inter-ship distance D from the interval [2,6] to the interval [0,1], specifically as follows:

[0156] D′=(D-2) / 4.

[0157] In a specific implementation, as a preferred embodiment of the present invention, step S4 specifically includes:

[0158] S41. Construct the decision boundary, which is a function of TDV. The design references the Sigmoid activation function, which has the characteristic of compressing a large numerical range into the interval (0, 1), and is defined as follows:

[0159]

[0160] Where b0 is the parameter for adjusting the boundary size, k is the sensitivity of the boundary to TDV, μ is the critical value of the boundary at the baseline, and k0 is the minimum value of the boundary constraint.

[0161] S42. Calculate the decision trigger result based on the accumulated information. The calculation formula is as follows:

[0162]

[0163] in, X is the outcome triggered by the intervention in the decision. i B(TDV) represents the total amount of information accumulated at the current moment. i ) represents the size of the information boundary at the current moment, μ is the scaling factor, and C1 and C2 are the decision to make a collision avoidance decision and the decision not to make a collision avoidance decision, respectively;

[0164] S43, Based on intervention scenarios ξ i The decision trigger result is calculated using the following formula:

[0165]

[0166] Among them, intervention scenario ξ i Including Condi_1 and Condi_2, Condi_1 is defined as the total accumulated information duration at the current time being greater than or equal to the current TISDT, Condi_1: Σt i ≥K i Condi_2 is defined as the total change in TISDT between the α sampling times prior to time i and the current time being less than or equal to the actual accumulated time, where Condi_2:K i-α -K i ≤αΔt;

[0167] S44, When intervening in scenario ξ i If both Condi_1 and Condi_2 are satisfied, execute C1; otherwise, execute C2.

[0168] Example

[0169] Based on this verification platform, this invention designs three different ship encounter scenarios to verify the rationality of the decision timing output by the VDDM-CADT model, specifically as follows: Figure 4 As shown, where:

[0170] Scenario I is a typical two-ship encounter situation. In this invention, TS-1 is set to continuously turn to port to pass the starboard angle of OS during the encounter.

[0171] Scenario II is a three-ship encounter scenario. OS encounters TS-1 and TS-2 in a cross-encounter situation, but none of the three ships take any action in this scenario.

[0172] Scenario III presents a more complex four-ship encounter situation. In this scenario, OS encounters TS-1, TS-2, and TS-3 in a right-hand cross encounter, a head-on encounter, and a left-hand cross encounter, respectively. The other three ships besides OS take different actions. TS-1 first turns left and then continues to turn right, and finally turns left to near its original course after passing the clear. TS-2 first turns left and then turns right, and finally turns left to return to its original course. TS-3 turns left to resume its original course after its initial right turn.

[0173] These three ship encounter scenarios involve two, three, and four ships respectively. Considering the experimental objective of triggering collision avoidance decisions, the scenario-based experimental design logic of this invention is as follows:

[0174] The OS maintains a constant direction and speed, while the TS (Ship Traverse) is controlled by the pilot (who can either take action or remain inactive). Initial ship parameters for these three scenarios are as follows: Figure 5 As shown. Figure 5 The data not only shows the initial motion information of each ship, but also records the duration of the three encounter scenarios and the ship turning process.

[0175] Figure 6-8 The graph shows the changes in DDV and TDV parameters for various scenarios.

[0176] For scenario I, the parameter change records are as follows: Figure 6 As shown, due to TS-1 taking two actions during the encounter, DDV and TDV fluctuated, with the maximum DDV between the two ships approaching 1.

[0177] For scenario II, the parameter change records are as follows: Figure 7As shown, since TS-1, TS-2 and OS did not take any action, the fluctuation range of parameter changes was not large. The interesting phenomenon is that the TDV value of TS-1 changed abruptly at the last moment. This is because the instability of the ship motion parameters caused a second DDV, that is, during the encounter between the two ships, TS-1 had two DDV ship positions relative to OS.

[0178] For scenario III, the parameter change records are as follows: Figure 8 As shown, although OS did not take any action, TS-1, TS-2 and TS-3 all took additional actions, so the parameter change curves were significantly affected.

[0179] In summary, the following important conclusions can be drawn from the analysis of the domain-based parameter changes between ships in the three scenarios: When ships do not take any action (turning / changing speed), due to the influence of the marine environment, the ship's course and speed will still experience small-scale disturbances, and therefore the domain-based parameters between ships will also fluctuate slightly; when ships take action, the domain intrusion parameters between ships will change significantly. The smoothness of this change is related to the frequency of action. If the interval is short, the parameter change is rapid; if the interval is long, the parameter change is slow.

[0180] Figure 9-11 A graph recording the accumulation of information and boundary changes for each scenario.

[0181] For TS-1 in scenario I, its information accumulation change curve is as follows: Figure 9 As shown. In the initial stage of information accumulation, the accumulation of evidence between ships rapidly shifted towards "not triggering a decision," reaching the trigger boundary for the first time. After this, the total amount of evidence accumulated between ships returned to zero and collection resumed. Since the OS did not take any action, its information accumulation process continued to shift negatively. During this process, TS-1 performed a left turn to avoid a collision, causing a sharp increase in the DDV between the two ships. This caused the evidence accumulation value between the ships to turn positive and shift towards "triggering a decision." Unfortunately, also due to TS-1's left turn, the DDV between the two ships first increased and then decreased. Therefore, after the DDV decreased and stabilized, the amount of decision-making evidence accumulated between the ships also began to change slowly, never reaching the trigger boundary until the very end of the scenario.

[0182] For Scenario II, OS, TS-1, and TS-2 did not take any action during the entire ship encounter. The information accumulation change curves for this scenario are as follows: Figure 10As shown in (a) and (b) of the figure, since all ships maintain their heading and speed during the encounter, the information accumulation between the ships is always biased towards the negative side. For TS-1 and TS-2, the frequency of evidence accumulation triggering the boundary differs due to the difference in risk parameters relative to OS. Because TS-2 has a higher DDV value than TS-1, the evidence accumulation speed of TS-2 is slightly faster, i.e., TS-1 triggers three times, while TS-2 only triggers twice. After the collision risk parameters tend to stabilize, the information accumulation process of the two ships relative to OS also tends to stabilize.

[0183] Scenario III presents a complex ship encounter situation. The information accumulation and change curves for this scenario are as follows: Figure 11 As shown in (a), (b), and (c), the OS encountered three TS vessels during a prolonged encounter (over 30 minutes). Each of the three TS vessels' pilots, based on their individual perceptions, employed a series of different avoidance maneuvers, including both compliance with and non-compliance with COLREGS. TS-1 had the longest initial trigger time for a natural decision. This was because TS-1 initially turned slowly to the left and then accelerated to the left, continuously increasing the DDV between the two vessels and accumulating evidence. Afterward, TS-1 began a series of right turns, rapidly decreasing the inter-ship domain parameter and causing the information accumulation to rapidly expand negatively, repeatedly reaching the negative decision boundary. Besides TS-1, TS-2 initially encountered the OS in a typical cross-encounter situation. Due to the extremely high inter-ship domain parameter values, TS-2's information accumulation curve repeatedly triggered the upper decision boundary from the initial moment. Following this, TS-2 first took a left turn to avoid a collision, which slightly decreased the domain parameter between the ships. Then, needing to avoid TS-1 on its left, TS-2 took a right turn. It is noteworthy that during this maneuver, the parameter value between TS-2 and OS experienced a brief, rapid increase, followed by a rapid decrease to 0 and a sustained value at that level. However, due to the information accumulation mechanism employed in this invention's model, no jump in the accumulated evidence occurred during the information accumulation process. For TS-3, initially positioned on OS's port side, it was a give-way vessel according to COLREGS regulations. Therefore, TS-3 proactively took reasonable avoidance action during the encounter, i.e., turning right past OS's stern. Because the initial domain parameter value between OS and TS-3 was small, and TS-3 took effective avoidance action, its accumulated information value initially increased negatively, and after the parameter values ​​stabilized, the information accumulation process entered a stable fluctuation phase.

[0184] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for modeling the triggering timing of ship collision avoidance decisions, mimicking the cognitive process of human society, characterized in that, include: S1. Obtain ship AIS data, extract the ship's latitude, longitude, heading, speed and length information from the AIS data, and construct an eccentric elliptical ship domain. S2. Based on the constructed eccentric elliptical ship domain, calculate the remaining distance DDV before another ship's TS invades the ship's OS domain, the remaining time TDV before another ship's TS invades the ship's OS domain, and the inter-ship distance D. S3. Construct the drift rate function This forms the cumulative amount of information per unit time and the total cumulative amount of information at the current moment, including: S31, the drift rate includes the Theoretical Instantaneous Safe Decision Time (TISDT) and the Theoretical Standard Safe Decision Time (TSSDT), where, Represents any current time. i The time required for security decision-making The maximum safety decision duration for the current TDV sub-interval is represented by the following formula: in, It is the parsing function of SDT. Indicates time i The DDV value below, Indicates time i The distance between the two ships; S32, Calculation Time The direction of evidence accumulation below The calculation formula is as follows: in, express The corresponding DDV value; S33. Based on steps S31 and S32, the drift rate function The definition is as follows: in, For time step, For tuning factor, For fixed constants, For the current moment The cumulative decision time under the following conditions It reflects the rate at which evidence accumulates; S34. Construct the information accumulation model of VDDM-CADT, using the following formula: in, Indicates the current The total amount of information that has been accumulated at this moment. This means that in the context of This represents the cumulative amount of information per unit time within a given time step. The unit is seconds. Representing diffuse noise; assuming the increment of the diffuse component follows a standard Wiener process, indicating that the change in the diffusion increment over any finite time interval follows a normal distribution, expressed as... ; The scale parameter reflects the proportion of the remaining accumulated information at the distance to the trigger threshold to the current trigger threshold. The closer the accumulated information value is to the trigger threshold at the current moment, the better the calculated value. A smaller scale factor will be assigned; scale parameter The calculation formula is as follows: in, Indicates time The information accumulation boundary below, The parameter is an exponential function, and ; S35, when set At the initial moment of information accumulation, with Calculate per unit time length The formula for calculating the accumulated information at each moment is as follows: S4. Construct decision boundaries and generate decision-triggered results based on the accumulated amount of information and the intervention scenario.

2. The method for modeling the triggering timing of ship collision avoidance decisions, which mimics the cognitive process of human society, as described in claim 1, is characterized in that... In step S1, the eccentric elliptical ship domain is constructed, specifically including: The major axis of the elliptical domain is parallel to the heading of the ship's OS, and the semi-major axis of the elliptical domain is... a The minor semi-axis of the ellipse is b The displacement of the ship from the center of the ellipse along the semi-major axis to the stern is: The displacement of the ship to the left along the minor semi-axis from the center of the ellipse is: and set a , b , and The values ​​are as follows: in, l This refers to the captain of the ship.

3. The method for modeling the triggering timing of ship collision avoidance decisions, which mimics the cognitive process of human society, as described in claim 1, is characterized in that... Step S2 specifically includes: S21. Scaling factor based on the maximum extent to which another ship's TS intrudes into the ship's OS domain. Calculate the remaining distance before another ship's TS invades the OS domain of this ship. The calculation formula is as follows: Wherein, DDV is composed of the domain scaling factor. Decision, and It's about time. quadratic function Then in Given a solution, the scaling factor is obtained when the other ship's TS (Transportation System) invades the maximum extent of the ship's OS (Operational System) domain. Specifically, This proves that other ships' TS will never intrude into the OS domain of this ship; if This indicates that another ship's TS has just crossed the boundary of this ship's OS; if This indicates that other ships' TS will inevitably intrude into the territory of this ship's OS; S22. Calculate the remaining time TDV before another vessel's TS intrudes into the vessel's OS domain. The calculation formula is as follows: in, According to the formula The solution is obtained.

4. The method for modeling the triggering timing of ship collision avoidance decisions, which mimics the cognitive process of human society, as described in claim 1, is characterized in that... In step S31, both the theoretical instantaneous safety decision duration TISDT and the theoretical standard safety decision duration TSSDT are related to the safety decision duration SDT. SDT represents the upper limit of the collision avoidance decision triggering time for the ship's navigator under the premise of ensuring navigational safety, that is, the maximum time to determine whether to make a collision avoidance decision. The setting of the safety decision duration SDT is related to the remaining distance DDV, inter-ship distance D, and TDV before another vessel TS intrudes into the ship's OS domain, as detailed below: The remaining distance DDV before another vessel's TS intrudes into the vessel's OS and the inter-ship distance D represent the urgency of the current situation. That is, the clearer the urgency of the current situation, the shorter the SDT will be, and vice versa. The greater the remaining time TDV before another vessel's TS intrudes into the vessel's OS, the more time the human operator has to make a safety decision, and the longer the SDT will be.

5. The method for modeling the triggering timing of ship collision avoidance decisions, which mimics the cognitive process of human society, as described in claim 4, is characterized in that... The specific calculation process for the Security Decision Time (SDT) includes: The remaining time TDV before another vessel's TS intrudes into the OS domain of this vessel, the remaining distance DDV before another vessel's TS intrudes into the OS domain of this vessel, and the distance D are limited to specific ranges and intervals. The unit of the remaining time TDV before another vessel's TS intrudes into the OS domain of this vessel is hours, and the unit of D is nautical miles. The remaining time TDV before another ship's TS intrudes into the ship's OS domain is divided into six sub-intervals. For each group of other vessels' TS before intruding into the vessel's OS domain, based on the survey and analysis results of the ship's drivers, the remaining distance DDV before intruding into the vessel's OS domain and the proportion of remaining time TDV before intruding into the vessel's OS domain under different ship-to-ship distances D are calculated respectively. The analytical expression for the Security Decision Duration (SDT) was obtained through data fitting, and the SDT was calculated based on the actual values ​​of the independent variables, as follows: in, express j+ TDV value at time 1 express j TDV value at time t, express j+ DDV value at time 1 express j The DDV value at time t. express j+ The distance between the two ships at time 1 express j The distance between the two ships at any given time.

6. The method for modeling the triggering timing of ship collision avoidance decisions, which mimics the cognitive process of human society, as described in claim 5, is characterized in that... Assume the numerical expression for the Security Decision Duration (SDT) is a special Gaussian smooth function, namely the bulge function, defined as: in, This represents the amplitude of the bulge function, i.e., the highest value at the center point. This represents the center radius of the bulge function. This represents the center point of the bulge function. This represents the minimum margin of the bulge function; Based on the numerical expression of the Security Decision Duration (SDT), the SDT calculation formula is obtained. ,as follows: when Then: when Then: in, and Represents the coordinates of the center point of the bulge function; This represents the proportion of the safety decision time SDT to the remaining time TDV before the other vessel's TS intrudes into the vessel's OS, under the most complex combination of the remaining distance DDV and the inter-ship distance D before the other vessel's TS intrudes into the vessel's OS. This represents the proportion of the minimum decision-making time for collision avoidance to the remaining time TDV before another vessel's TS intrudes into the vessel's OS domain; Represents the radius of the bulge function; , and These are the center and radius of the bulge function, respectively; Since the range of the inter-ship distance D is not consistent with the remaining distance DDV before other ships' TS intrudes into the OS domain of this ship, it is necessary to map the value of the inter-ship distance D from the interval [2,6] to the interval [0,1], specifically as follows: 。 7. The method for modeling the triggering timing of ship collision avoidance decisions, which mimics the cognitive process of human society, as described in claim 1, is characterized in that... Step S4 specifically includes: S41. Construct the decision boundary, which is a function of TDV. The design references the Sigmoid activation function, which has the characteristic of compressing a large numerical range into the interval (0, 1), and is defined as follows: in, To adjust the parameters of the boundary size, For the sensitivity of the boundary to TDV, The critical value at the baseline is the boundary. To limit the minimum value of the boundary; S42. Calculate the decision trigger result based on the accumulated information. The calculation formula is as follows: in, For the results triggered by intervention in the decision, This represents the total amount of information accumulated at the current moment. This represents the size of the information boundary at the current moment. As a scaling factor, and The options are to make a collision avoidance decision and not to make a collision avoidance decision. S43, Based on intervention scenarios The decision trigger result is calculated using the following formula: Among them, intervention scenarios include and , Defined as the total accumulated information duration at the current moment being greater than or equal to the current TISDT. ; Defined as time i Previous The total change of TISDT between the previous sampling time and the current time is less than or equal to the actual accumulated time. ; S44, When Intervention Scenarios satisfy and When, execute Otherwise execute .

Images