A ship intelligent steering method and system fusing neuromorphic computing

Abstract

The application discloses a warship intelligent steering method and system fusing neuromorphic computing, and the method comprises the following steps: acquiring characteristic values of multiple sensor data and screening out abnormal characteristic values of each sensor data, and encoding the abnormal characteristic values into a set of neural pulse event values in the form of pulses, which are used for simulating the response of neurons to abnormal information; expanding the set of neural pulse event values in a rudder angle strategy space to form a neural response intensity which dynamically changes over time and is used for simulating the response sensitivity of neurons in each direction; projecting the neural response intensity to the rudder angle strategy space to form a rudder angle strategy atlas which dynamically changes over time, and extracting an optimal strategy rudder angle from the rudder angle strategy atlas as a final rudder angle of the warship, so that the intelligent steering of the warship is completed.

Description

Technical Field

[0001] This invention belongs to the field of ship intelligent steering technology, and more specifically, relates to a ship intelligent steering method and system that integrates neuromorphic computing. Background Technology

[0002] Existing intelligent ship steering systems mostly rely on rule-based control, classical PID algorithms, fuzzy logic, optimal control, or deep reinforcement learning to achieve automatic heading adjustments. Their core idea is to generate rudder angle commands based on heading errors, environmental disturbances (such as wind, waves, and currents), and target tracks, thereby achieving path tracking or disturbance-resistant steady-state control. In engineering practice, mainstream systems often depend on centralized decision-making frameworks, periodically collecting sensor data (such as gyroscope angular velocity, GPS heading, and current meter direction) and inputting it into the control algorithm, updating it at fixed time intervals (e.g., 50–100 ms). However, this type of architecture based on periodic sampling-processing-control has significant shortcomings under high-speed sea conditions, strong disturbances, or sudden events. On the one hand, because sensor inputs are typically sudden and non-stationary (such as surge impacts and transient rheological changes), sampling at a fixed frequency easily leads to the loss of critical dynamic information, resulting in a delayed response from the decision-making system. On the other hand, existing algorithms mostly simulate continuous control laws at the computational level, making it difficult to provide structural feedback on abrupt changes at the microsecond level. This is especially true in the initial steering phase, where insufficient perception of asymmetric disturbances often leads to overcorrection or delayed response. Meanwhile, while deep learning algorithms possess strong fitting capabilities, their high real-time performance and energy consumption requirements severely limit their deployment in shipboard embedded systems. Furthermore, existing methods generally employ full data processing, exhibiting poor sensitivity to information and failing to dynamically select the processing granularity based on the "anomaly" of the input signal, easily resulting in redundant calculations or ineffective control.

[0003] Therefore, in ship intelligent steering systems, there is an urgent need for a new method that can simulate the rapid response mechanism of biological neurons, possess event-driven perception characteristics, and achieve millisecond-level low-power inference, in order to adapt to complex and ever-changing navigation environments. Summary of the Invention

[0004] To address the above technical problems, this invention proposes a ship intelligent steering method integrating neuromorphic computing, comprising:

[0005] The feature values ​​of multiple sensor data are acquired and the abnormal feature values ​​of each sensor data are filtered out. These are then encoded as a set of neural impulse event values ​​in the form of impulses to simulate the neuron's response to abnormal information.

[0006] The set of neural impulse event values ​​is expanded in the rudder angle strategy space to form a neural response intensity that changes dynamically over time, which is used to simulate the response sensitivity of neurons in various directions.

[0007] The neural response intensity is projected onto the rudder angle strategy space to form a rudder angle strategy map that changes dynamically over time. The optimal strategy rudder angle is then extracted from the map as the final rudder angle of the ship, thereby completing the intelligent steering of the ship.

[0008] Furthermore, acquiring information from multiple sensors and filtering out abnormal information includes:

[0009]

[0010] Where, x k (t) represents the characteristic value of the k-th sensor at time t, σ k Let θ be the moving standard deviation of the eigenvalues ​​of the k-th sensor within the time window τ. k The offset threshold of the feature value of the k-th sensor. ζ is the moving average of the characteristic values ​​of the k-th sensor within the time window τ. k (t) represents the standardized offset of the feature value of the k-th sensor at time t.

[0011] Furthermore, the set of neural impulse event values ​​encoded in impulse form includes:

[0012]

[0013] Where Λ(t) is the set of neural impulse event values ​​at time t, N is the number of sensors, and λ k (t) represents the discretized pulse value of the characteristic value of the k-th sensor at time t, x k (t) represents the amplitude of the characteristic value of the k-th sensor at time t.

[0014] Furthermore, the set of neural impulse event values ​​is expanded in the rudder angle strategy space to form a neural response intensity that dynamically changes over time, including:

[0015] The rudder angle strategy space is the currently selectable rudder angle command space, and the rudder angle command space includes multiple strategy rudder angles;

[0016] The strength of a neural response that varies dynamically over time includes:

[0017]

[0018] in, γ represents the neural response strength at time t for the policy rudder angle x′. i Let D be the i-th neural impulse event value in the set of neural impulse event values. i (x′) represents the angle difference between the current ship rudder angle and the strategy rudder angle x′ corresponding to the i-th neural impulse event value, Γ i(x′) is the diffusion scale of the strategy rudder angle x′ corresponding to the i-th neural impulse event value.

[0019] Furthermore, projecting the neural response intensity onto the rudder angle strategy space to form a rudder angle strategy map that dynamically changes over time includes:

[0020]

[0021] in, Let Ω be the rudder angle strategy map at time t, which includes all possible rudder angles θ′ of all ships. θ′ is the possible rudder angle θ′ of a ship, Ω is the rudder angle strategy space, and σ(θ′|x′) is the directional activation weight of the strategy rudder angle x′ on the possible rudder angle θ′ of a ship.

[0022] Furthermore, extracting the optimal strategy rudder angle as the ship's final rudder angle includes:

[0023]

[0024] Where, θ * (t) represents the optimal rudder angle at time t.

[0025] This invention also proposes a ship intelligent steering system integrating neuromorphic computing, comprising:

[0026] The encoding module is used to acquire feature values ​​from multiple sensor data and filter out abnormal feature values ​​from each sensor data, encoding them in the form of pulses as a set of neural pulse event values ​​to simulate the neuron's response to abnormal information.

[0027] The neural response intensity module is used to expand the set of neural impulse event values ​​in the rudder angle strategy space to form a neural response intensity that changes dynamically over time, which is used to simulate the response sensitivity of neurons in various directions.

[0028] The intelligent steering module projects the neural response intensity onto the rudder angle strategy space to form a rudder angle strategy map that changes dynamically over time, and extracts the optimal strategy rudder angle from it as the ship's final rudder angle, thereby completing the ship's intelligent steering.

[0029] Furthermore, acquiring information from multiple sensors and filtering out abnormal information includes:

[0030]

[0031] Where, x k (t) represents the characteristic value of the k-th sensor at time t, σ k Let θ be the moving standard deviation of the eigenvalues ​​of the k-th sensor within the time window τ. k The offset threshold of the feature value of the k-th sensor. ζ is the moving average of the characteristic values ​​of the k-th sensor within the time window τ. k (t) represents the standardized offset of the feature value of the k-th sensor at time t.

[0032] Furthermore, the set of neural impulse event values ​​encoded in impulse form includes:

[0033]

[0034] Where Λ(t) is the set of neural impulse event values ​​at time t, N is the number of sensors, and λ k (t) represents the discretized pulse value of the characteristic value of the k-th sensor at time t, x k (t) represents the amplitude of the characteristic value of the k-th sensor at time t.

[0035] Furthermore, the set of neural impulse event values ​​is expanded in the rudder angle strategy space to form a neural response intensity that dynamically changes over time, including:

[0036] The rudder angle strategy space is the currently selectable rudder angle command space, and the rudder angle command space includes multiple strategy rudder angles;

[0037] The strength of a neural response that varies dynamically over time includes:

[0038]

[0039] in, γ represents the neural response strength at time t for the policy rudder angle x′. i Let D be the i-th neural impulse event value in the set of neural impulse event values. i (x′) represents the angle difference between the current ship rudder angle and the strategy rudder angle x′ corresponding to the i-th neural impulse event value, Γ i (x′) is the diffusion scale of the strategy rudder angle x′ corresponding to the i-th neural impulse event value.

[0040] In summary, the technical solutions conceived by this invention have the following beneficial effects compared with the prior art:

[0041] This invention transforms multi-source sensor signals (such as gyroscope, current meter, AIS) into neural-like events, dynamically generates neural response intensity in the rudder angle strategy space, and constructs a rudder angle strategy map to achieve intelligent rudder angle decision reasoning with millisecond-level latency (<10ms).

[0042] This invention simulates the selective excitation mechanism of biological neurons in response to spatial orientation, and features enhanced orientation perception, adaptive diffusion response, and low-power inference, significantly improving the ship handling response speed under high sea states. Attached Figure Description

[0043] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention;

[0044] Figure 2 This is a system structure diagram of Embodiment 2 of the present invention. Detailed Implementation

[0045] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0046] The method provided by this invention can be implemented in a terminal environment that may include one or more of the following components: a processor, a storage medium, and a display screen. The storage medium stores at least one instruction, which is loaded and executed by the processor to implement the method described in the following embodiments.

[0047] A processor may include one or more processing cores. The processor uses various interfaces and lines to connect various parts of the terminal, and performs various functions and processes data by running or executing instructions, programs, code sets or instruction sets stored in the storage medium, and by calling data stored in the storage medium.

[0048] Storage media can include random access memory (RAM) or read-only memory (ROM). Storage media can be used to store instructions, programs, code, code sets, or instructions.

[0049] The display screen is used to show the user interface of each application.

[0050] In addition, those skilled in the art will understand that the structure of the terminal described above does not constitute a limitation on the terminal. The terminal may include more or fewer components, or combine certain components, or have different component arrangements. For example, the terminal may also include radio frequency circuits, input units, sensors, audio circuits, power supplies, and other components, which will not be described in detail here.

[0051] Example 1

[0052] like Figure 1 This embodiment proposes a ship intelligent steering method integrating neuromorphic computing, including:

[0053] Step 101: Acquire feature values ​​of multiple sensor data (e.g., gyroscope / current meter / AIS sensor data) and filter out abnormal feature values ​​of each sensor data, encoding them in the form of pulses as a set of neural pulse event values ​​to simulate the neuron's response to abnormal information;

[0054] Specifically, acquiring information from multiple sensors and filtering out abnormal information includes:

[0055]

[0056] Where, x k (t) represents the characteristic value of the k-th sensor at time t, σ k Let θ be the moving standard deviation of the eigenvalues ​​of the k-th sensor within the time window τ. k The offset threshold of the feature value of the k-th sensor. ζ is the moving average of the characteristic values ​​of the k-th sensor within the time window τ. k (t) represents the standardized offset of the feature value of the k-th sensor at time t.

[0057] Specifically, the set of neural impulse event values ​​encoded in impulse form includes:

[0058]

[0059] Where Λ(t) is the set of neural impulse event values ​​at time t, N is the number of sensors, and λ k (t) represents the discretized pulse value of the characteristic value of the k-th sensor at time t, x k (t) represents the amplitude of the characteristic value of the k-th sensor at time t.

[0060] Step 102: Expand the set of neural impulse event values ​​in the rudder angle strategy space to form a neural response intensity that changes dynamically over time, which is used to simulate the response sensitivity of neurons in various directions.

[0061] Specifically, the set of neural impulse event values ​​is expanded in the rudder angle strategy space to form a neural response intensity that dynamically changes over time, including:

[0062] The rudder angle strategy space is the currently selectable rudder angle command space, and the rudder angle command space includes multiple strategy rudder angles;

[0063] The strength of a neural response that varies dynamically over time includes:

[0064]

[0065] in, γ represents the neural response strength at time t for the policy rudder angle x′. i Let D be the i-th neural impulse event value in the set of neural impulse event values. i (x′) represents the angle difference between the current ship rudder angle and the strategy rudder angle x′ corresponding to the i-th neural impulse event value, Γ i(x′) is the diffusion scale of the strategy rudder angle x′ corresponding to the i-th neural impulse event value.

[0066] Preferably, this embodiment provides the following method for obtaining Γ. i (x′):

[0067] Γ i (x)=γ0+α·D i (x)

[0068] Where γ0 is the initial diffusion scale and α is the adjustment factor.

[0069] Preferably, γ0 and α can be set based on experience. For example, if the maneuver requires quick and focused control (such as sharp turn obstacle avoidance), set smaller γ0 and α (more focused response); if the sea conditions are complex and require a higher tolerance for error, set larger γ0 and α (more relaxed response); if the ship speed is high and the inertia is large, increase γ0.

[0070] Step 103: Project the neural response intensity onto the rudder angle strategy space to form a rudder angle strategy map that changes dynamically over time, and extract the optimal strategy rudder angle from it as the ship's final rudder angle, thereby completing the ship's intelligent steering.

[0071] Specifically, projecting the neural response intensity onto the rudder angle strategy space to form a rudder angle strategy map that dynamically changes over time includes:

[0072]

[0073] in, Let Ω be the rudder angle strategy map at time t, which includes all possible rudder angles θ′ of all ships. θ′ is the possible rudder angle θ′ of a ship, Ω is the rudder angle strategy space, and σ(θ′|x′) is the directional activation weight of the strategy rudder angle x′ on the possible rudder angle θ′ of a ship.

[0074] Preferably, in this embodiment, the directional activation weight σ(θ′|x′) of the strategy rudder angle x′ relative to the possible rudder angle θ′ of the ship is calculated using the following formula:

[0075]

[0076] Where, ρ L ρ represents the first activation threshold of the rudder angle θ′ that the ship may adopt relative to the strategic rudder angle x′. R Z1 is the second activation threshold for the possible rudder angle θ′ of the ship relative to the strategic rudder angle x′, and Z1 is the first normalization factor. Z2 is the second normalization factor.

[0077] For example, if the strategic rudder angle x′ is 15°, the available rudder angle θ′ of the ship may be 10°, and ρ L is 7°, satisfying the condition that if θ′<x′ and |θ′-x′|<ρ L , then:

[0078]

[0079] If the available rudder angle θ′ of the ship is 20°, ρ R is 10°, satisfying that if θ′≥x′ and |θ′-x′|<ρ R , then:

[0080]

[0081] Preferably, this embodiment provides an example of obtaining ρ L and ρ R , for example: ρ L =ρ R =maximum safe rudder angle change rate * unit time, and ρ L and ρ R can also be obtained through empirical values, which is not limited in this embodiment.

[0082] Specifically, extracting the optimal strategic rudder angle as the final rudder angle of the ship includes:

[0083]

[0084] wherein, θ * (t) is the optimal strategic rudder angle at time t.

[0085] Example 2

[0086] As shown in Figure 2 , this embodiment proposes an intelligent ship steering system integrating neuromorphic computing, comprising:

[0087] an encoding module, configured to acquire feature values of a plurality of sensor data (such as sensor data of gyroscope / flowmeter / AIS, etc.), screen out abnormal feature values of each sensor data, and encode the same into a set of neural spike event values in the form of pulses, which is used for simulating the response of neurons to abnormal information;

[0088] Specifically, acquiring a plurality of sensor information and screening out abnormal information comprises:

[0089]

[0090] wherein, x k (t) is a feature value of the k-th sensor at time t, σ k is the sliding standard deviation of feature values of the k-th sensor within a time window τ, θk The offset threshold for the feature value of the k-th sensor. ζ is the moving average of the characteristic values ​​of the k-th sensor within the time window τ. k (t) represents the standardized offset of the feature value of the k-th sensor at time t.

[0091] Specifically, the set of neural impulse event values ​​encoded in impulse form includes:

[0092]

[0093] Where Λ(t) is the set of neural impulse event values ​​at time t, N is the number of sensors, and λ k (t) represents the discretized pulse value of the characteristic value of the k-th sensor at time t, x k (t) represents the amplitude of the characteristic value of the k-th sensor at time t.

[0094] The neural response intensity module is used to expand the set of neural impulse event values ​​in the rudder angle strategy space to form a neural response intensity that changes dynamically over time, which is used to simulate the response sensitivity of neurons in various directions.

[0095] Specifically, the set of neural impulse event values ​​is expanded in the rudder angle strategy space to form a neural response intensity that dynamically changes over time, including:

[0096] The rudder angle strategy space is the currently selectable rudder angle command space, and the rudder angle command space includes multiple strategy rudder angles;

[0097] The strength of a neural response that varies dynamically over time includes:

[0098]

[0099] in, γ represents the neural response strength at time t for the policy rudder angle x′. i Let D be the i-th neural impulse event value in the set of neural impulse event values. i (x′) represents the angle difference between the current ship rudder angle and the strategy rudder angle x′ corresponding to the i-th neural impulse event value, Γ i (x′) is the diffusion scale of the policy rudder angle x′ corresponding to the i-th neural impulse event value.

[0100] Preferably, this embodiment provides the following method for obtaining Γ. i (x′):

[0101] Γ i (x)=γ0+α·D i (x)

[0102] Where γ0 is the initial diffusion scale and α is the adjustment factor.

[0103] Preferably, γ0 and α can be set based on experience. For example, if the maneuver requires quick and focused control (such as sharp turn obstacle avoidance), set smaller γ0 and α (more focused response); if the sea conditions are complex and require a higher tolerance for error, set larger γ0 and α (more relaxed response); if the ship speed is high and the inertia is large, increase γ0.

[0104] The intelligent steering module projects the neural response intensity onto the rudder angle strategy space to form a rudder angle strategy map that changes dynamically over time, and extracts the optimal strategy rudder angle from it as the ship's final rudder angle, thereby completing the ship's intelligent steering.

[0105] Specifically, projecting the neural response intensity onto the rudder angle strategy space to form a rudder angle strategy map that dynamically changes over time includes:

[0106]

[0107] in, Let Ω be the rudder angle strategy map at time t, which includes all possible rudder angles θ′ of all ships. θ′ is the possible rudder angle θ′ of a ship, Ω is the rudder angle strategy space, and σ(θ′|x′) is the directional activation weight of the strategy rudder angle x′ on the possible rudder angle θ′ of a ship.

[0108] Preferably, in this embodiment, the directional activation weight σ(θ′|x′) of the strategy rudder angle x′ relative to the rudder angle θ′ that the ship may adopt is calculated using the following formula:

[0109]

[0110] Where, ρ L ρ represents the first activation threshold of the rudder angle θ′ that the ship may adopt relative to the strategic rudder angle x′. R Z1 is the second activation threshold for the possible rudder angle θ′ of the ship relative to the strategic rudder angle x′, and Z1 is the first normalization factor. Z2 is the second normalization factor.

[0111] For example, if the strategic rudder angle x′ is 15°, the ship might use a rudder angle θ′ of 10°, ρ L The angle is 7°, satisfying the condition that if θ′<x′ and |θ′-x′|<ρ L ,but:

[0112]

[0113] If the rudder angle θ′ that the ship may use is 20°, ρ RThe angle is 10°, satisfying && if θ′≥x′ and |θ′-x′|<ρ R ,but:

[0114]

[0115] Preferably, this embodiment provides a method for obtaining ρ L and ρ R Examples, such as: ρ L =ρ R = Maximum safe rudder angle change rate * unit time, or ρ can be obtained through empirical values. L and ρ R This embodiment is not limited.

[0116] Specifically, extracting the optimal rudder angle as the ship's final rudder angle includes:

[0117]

[0118] Where, θ * (t) represents the optimal rudder angle at time t.

[0119] Example 3

[0120] This invention also proposes a storage medium storing multiple instructions for implementing the aforementioned ship intelligent steering method incorporating neuromorphic computing.

[0121] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0122] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following method steps: Step 101, acquiring feature values ​​of multiple sensor data (e.g., gyroscope / current meter / AIS sensor data) and filtering out abnormal feature values ​​of each sensor data, encoding them in the form of pulses as a set of neural pulse event values ​​to simulate the response of neurons to abnormal information;

[0123] Specifically, acquiring information from multiple sensors and filtering out abnormal information includes:

[0124]

[0125] Where, x k (t) represents the characteristic value of the k-th sensor at time t, σ k Let θ be the moving standard deviation of the eigenvalues ​​of the k-th sensor within the time window τ. k The offset threshold for the feature value of the k-th sensor. ζ is the moving average of the characteristic values ​​of the k-th sensor within the time window τ. k (t) represents the standardized offset of the feature value of the k-th sensor at time t.

[0126] Specifically, the set of neural impulse event values ​​encoded in impulse form includes:

[0127]

[0128] Where Λ(t) is the set of neural impulse event values ​​at time t, N is the number of sensors, and λ k (t) represents the discretized pulse value of the characteristic value of the k-th sensor at time t, x k (t) represents the amplitude of the characteristic value of the k-th sensor at time t.

[0129] Step 102: Expand the set of neural impulse event values ​​in the rudder angle strategy space to form a neural response intensity that changes dynamically over time, which is used to simulate the response sensitivity of neurons in various directions.

[0130] Specifically, the set of neural impulse event values ​​is expanded in the rudder angle strategy space to form a neural response intensity that dynamically changes over time, including:

[0131] The rudder angle strategy space is the currently selectable rudder angle command space, and the rudder angle command space includes multiple strategy rudder angles;

[0132] The strength of a neural response that varies dynamically over time includes:

[0133]

[0134] in, γ represents the neural response strength at time t for the policy rudder angle x′. i Let D be the i-th neural impulse event value in the set of neural impulse event values. i (x′) represents the angle difference between the current ship rudder angle and the strategy rudder angle x′ corresponding to the i-th neural impulse event value, Γ i (x′) is the diffusion scale of the policy rudder angle x′ corresponding to the i-th neural impulse event value.

[0135] Preferably, this embodiment provides the following method for obtaining Γ. i (x′):

[0136] Γ i (x)=γ0+α·D i (x)

[0137] Where γ0 is the initial diffusion scale and α is the adjustment factor.

[0138] Preferably, γ0 and α can be set based on experience. For example, when the control requires rapid concentration (such as sharp turning for obstacle avoidance), smaller γ0 and α are set to achieve more concentrated response; when the sea conditions are complex and a higher fault tolerance rate is required, larger γ0 and α are set to achieve a more relaxed response; when the ship speed is high and the inertia is large, γ0 is increased.

[0139] Step 103: projecting the neural response intensity into a rudder angle strategy space, forming a rudder angle strategy map that dynamically changes with time, and extracting the optimal strategy rudder angle therefrom as the final rudder angle of the ship, thereby completing intelligent ship steering.

[0140] Specifically, projecting the neural response intensity into the rudder angle strategy space to form a dynamically changing rudder angle strategy map over time includes:

[0141]

[0142] Wherein, is a rudder angle strategy map including all possible rudder angles θ' that can be adopted by the ship at time t, θ' is a possible rudder angle θ' that can be adopted by the ship, Ω is the rudder angle strategy space, and σ(θ'|x') is the directional activation weight of the strategy rudder angle x' for the possible rudder angle θ' that can be adopted by the ship.

[0143] Preferably, in this embodiment, the directional activation weight σ(θ'|x') of the strategy rudder angle x' for the possible rudder angle θ' that can be adopted by the ship is calculated by the following formula:

[0144]

[0145] Wherein, ρ L is a first activation threshold of the possible rudder angle θ' that can be adopted by the ship relative to the strategy rudder angle x', ρ R is a second activation threshold of the possible rudder angle θ' that can be adopted by the ship relative to the strategy rudder angle x', Z1 is a first normalization factor, Z2 is a second normalization factor,

[0146] For example, if the strategy rudder angle x' is 15°, the possible rudder angle θ' that can be adopted by the ship is 10°, and ρ L is 7°, which satisfies that θ'<x' and |θ'-x'|<ρ L , then:

[0147]

[0148] If the possible rudder angle θ' that can be adopted by the ship is 20°, and ρ R is 10°, which satisfies that θ'≥x' and |θ'-x'|<ρ R , then:

[0149]

[0150] Preferably, this embodiment provides a method for obtaining ρ L and ρ R Examples, such as: ρ L =ρ R = Maximum safe rudder angle change rate * unit time, or ρ can be obtained through empirical values. L and ρ R This embodiment is not limited.

[0151] Specifically, extracting the optimal rudder angle as the ship's final rudder angle includes:

[0152]

[0153] Where, θ * (t) represents the optimal rudder angle at time t.

[0154] Example 4

[0155] This invention also proposes an electronic device, including a processor and a storage medium connected to the processor. The storage medium stores multiple instructions, which can be loaded and executed by the processor to enable the processor to execute the aforementioned intelligent ship steering method incorporating neuromorphic computing.

[0156] Specifically, the electronic device in this embodiment can be a computer terminal, which may include one or more processors and a storage medium.

[0157] The storage medium can be used to store software programs and modules, such as the ship intelligent steering method integrating neuromorphic computing in this embodiment of the invention. The corresponding program instructions / modules are executed by the processor through running the software programs and modules stored in the storage medium, thereby performing various functional applications and data processing, thus realizing the aforementioned ship intelligent steering method integrating neuromorphic computing. The storage medium may include high-speed random access storage media, and may also include non-volatile storage media, such as one or more magnetic storage systems, flash memory, or other non-volatile solid-state storage media. In some instances, the storage medium may further include storage media remotely configured relative to the processor, which can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0158] The processor can call the information and application stored in the storage medium through the transmission system to execute the following method steps: Step 101, acquire the feature values ​​of multiple sensor data (e.g., gyroscope / current meter / AIS sensor data) and filter out the abnormal feature values ​​of each sensor data, and encode them in the form of a set of neural impulse event values ​​to simulate the response of neurons to abnormal information;

[0159] Specifically, acquiring information from multiple sensors and filtering out abnormal information includes:

[0160]

[0161] Where, x k (t) represents the characteristic value of the k-th sensor at time t, σ k Let θ be the moving standard deviation of the eigenvalues ​​of the k-th sensor within the time window τ. k The offset threshold of the feature value of the k-th sensor. ζ is the moving average of the characteristic values ​​of the k-th sensor within the time window τ. k (t) represents the standardized offset of the feature value of the k-th sensor at time t.

[0162] Specifically, the set of neural impulse event values ​​encoded in impulse form includes:

[0163]

[0164] Where Λ(t) is the set of neural impulse event values ​​at time t, N is the number of sensors, and λ k (t) represents the discretized pulse value of the characteristic value of the k-th sensor at time t, x k (t) represents the amplitude of the characteristic value of the k-th sensor at time t.

[0165] Step 102: Expand the set of neural impulse event values ​​in the rudder angle strategy space to form a neural response intensity that changes dynamically over time, which is used to simulate the response sensitivity of neurons in various directions.

[0166] Specifically, the set of neural impulse event values ​​is expanded in the rudder angle strategy space to form a neural response intensity that dynamically changes over time, including:

[0167] The rudder angle strategy space is the currently selectable rudder angle command space, and the rudder angle command space includes multiple strategy rudder angles;

[0168] The strength of a neural response that varies dynamically over time includes:

[0169]

[0170] in, γ represents the neural response strength at time t for the policy rudder angle x′. i Let D be the i-th neural impulse event value in the set of neural impulse event values. i (x′) represents the angle difference between the current ship rudder angle and the strategy rudder angle x′ corresponding to the i-th neural impulse event value, Γ i (x′) is the diffusion scale of the policy rudder angle x′ corresponding to the i-th neural impulse event value.

[0171] Preferably, this embodiment provides the following method for obtaining Γ. i (x′):

[0172] Γ i (x)=γ0+α·D i (x)

[0173] Where γ0 is the initial diffusion scale and α is the adjustment factor.

[0174] Preferably, γ0 and α can be set based on experience. For example, if the maneuver requires quick and focused control (such as sharp turn obstacle avoidance), set smaller γ0 and α (more focused response); if the sea conditions are complex and require a higher tolerance for error, set larger γ0 and α (more relaxed response); if the ship speed is high and the inertia is large, increase γ0.

[0175] Step 103: Project the neural response intensity onto the rudder angle strategy space to form a rudder angle strategy map that changes dynamically over time, and extract the optimal strategy rudder angle from it as the ship's final rudder angle, thereby completing the ship's intelligent steering.

[0176] Specifically, projecting the neural response intensity onto the rudder angle strategy space to form a rudder angle strategy map that dynamically changes over time includes:

[0177]

[0178] in, Let Ω be the rudder angle strategy map at time t, which includes all possible rudder angles θ′ of all ships. θ′ is the possible rudder angle θ′ of a ship, Ω is the rudder angle strategy space, and σ(θ′|x′) is the directional activation weight of the strategy rudder angle x′ on the possible rudder angle θ′ of a ship.

[0179] Preferably, in this embodiment, the directional activation weight σ(θ′|x′) of the strategy rudder angle x′ relative to the possible rudder angle θ′ of the ship is calculated using the following formula:

[0180]

[0181] Where, ρ L ρ represents the first activation threshold of the rudder angle θ′ that the ship may adopt relative to the strategic rudder angle x′. RZ1 is the second activation threshold for the possible rudder angle θ′ of the ship relative to the strategic rudder angle x′, and Z1 is the first normalization factor. Z2 is the second normalization factor.

[0182] For example, if the strategic rudder angle x′ is 15°, the ship might use a rudder angle θ′ of 10°, ρ L The angle is 7°, satisfying the condition that if θ′<x′ and |θ′-x′|<ρ L ,but:

[0183]

[0184] If the rudder angle θ′ that the ship may use is 20°, ρ R The angle is 10°, satisfying && if θ′≥x′ and |θ′-x′|<ρ R ,but:

[0185]

[0186] Preferably, this embodiment provides a method for obtaining ρ L and ρ R Examples, such as: ρ L =ρ R = Maximum safe rudder angle change rate * unit time, or ρ can be obtained through empirical values. L and ρ R This embodiment is not limited.

[0187] Specifically, extracting the optimal rudder angle as the ship's final rudder angle includes:

[0188]

[0189] Where, θ * (t) represents the optimal rudder angle at time t.

[0190] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0191] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0192] In the several embodiments provided by this invention, it should be understood that the disclosed technical content can be implemented in other ways. The system embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between units or modules, and may be electrical or other forms.

[0193] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0194] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0195] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, optical disks, and other media capable of storing program code.

[0196] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A ship intelligent steering method integrating neuromorphic computing, characterized in that, include: The feature values ​​of multiple sensor data are acquired and the abnormal feature values ​​of each sensor data are filtered out. These are then encoded as a set of neural impulse event values ​​in the form of impulses to simulate the neuron's response to abnormal information. The set of neural impulse event values ​​encoded in impulse form includes: in, For time The set of neural impulse event values, For the number of sensors, For time Time Discretized pulse values ​​of the characteristic values ​​of each sensor For time Time The amplitude of the characteristic values ​​of each sensor; The set of neural impulse event values ​​is expanded in the rudder angle strategy space to form a neural response intensity that changes dynamically over time, which is used to simulate the response sensitivity of neurons in various directions. Expanding the set of neural impulse event values ​​in the rudder angle policy space forms a neural response intensity that dynamically changes over time, including: The rudder angle strategy space is the currently selectable rudder angle command space, and the rudder angle command space includes multiple strategy rudder angles; The strength of a neural response that varies dynamically over time includes: in, For time Strategy Rudder The intensity of the neural response, The set of neural impulse event values ​​is the first One neural impulse event value, Current ship rudder angle and the The strategy rudder angle corresponding to each neural impulse event value The angle difference between them For the first The strategy rudder angle corresponding to each neural impulse event value The diffusion scale; The neural response intensity is projected onto the rudder angle strategy space to form a rudder angle strategy map that changes dynamically over time. The optimal strategy rudder angle is then extracted from this map as the ship's final rudder angle, thereby completing intelligent ship steering.

2. The ship intelligent steering method integrating neuromorphic computing as described in claim 1, characterized in that, Acquiring information from multiple sensors and filtering out abnormal information includes: in, For time Time The characteristic values ​​of each sensor For time window Inner The sliding standard deviation of the characteristic values ​​of each sensor For the first The offset threshold of the feature values ​​of each sensor For time window Inner The moving average of the characteristic values ​​of each sensor time Time The standardized offset of the feature values ​​of each sensor.

3. The intelligent ship steering method integrating neuromorphic computing as described in claim 1, characterized in that, Projecting the neural response intensity onto the rudder angle strategy space to form a rudder angle strategy map that dynamically changes over time includes: in, For time This includes all possible rudder angles used by ships. Rudder angle strategy diagram, Rudder angles that ships may adopt , For the rudder angle strategy space, As the rudder of strategy Rudder angles that ships may adopt The direction of activation weights.

4. The ship intelligent steering method integrating neuromorphic computing as described in claim 3, characterized in that, Extracting the optimal rudder angle from these factors as the ship's final rudder angle includes: in, For time The optimal rudder angle for the current strategy.

5. A shipboard intelligent steering system integrating neuromorphic computing, characterized in that, include: The encoding module is used to acquire feature values ​​from multiple sensor data and filter out abnormal feature values ​​from each sensor data, encoding them in the form of pulses as a set of neural pulse event values ​​to simulate the neuron's response to abnormal information. The set of neural impulse event values ​​encoded in impulse form includes: in, For time The set of neural impulse event values, For the number of sensors, For time Time Discretized pulse values ​​of the characteristic values ​​of each sensor For time Time The amplitude of the characteristic values ​​of each sensor; The neural response intensity module is used to expand the set of neural impulse event values ​​in the rudder angle strategy space to form a neural response intensity that changes dynamically over time, which is used to simulate the response sensitivity of neurons in various directions. Expanding the set of neural impulse event values ​​in the rudder angle policy space forms a neural response intensity that dynamically changes over time, including: The rudder angle strategy space is the currently selectable rudder angle command space, and the rudder angle command space includes multiple strategy rudder angles; The strength of a neural response that varies dynamically over time includes: in, For time Strategy Rudder The intensity of the neural response, The set of neural impulse event values ​​is the first One neural impulse event value, Current ship rudder angle and the The strategy rudder angle corresponding to each neural impulse event value The angle difference between them For the first The strategy rudder angle corresponding to each neural impulse event value The diffusion scale; The intelligent steering module projects the neural response intensity onto the rudder angle strategy space to form a rudder angle strategy map that changes dynamically over time, and extracts the optimal strategy rudder angle from it as the ship's final rudder angle, thereby completing the ship's intelligent steering.

6. The ship intelligent steering system integrating neuromorphic computing as described in claim 5, characterized in that, Acquiring information from multiple sensors and filtering out abnormal information includes: in, For time Time The characteristic values ​​of each sensor For time window Inner The sliding standard deviation of the characteristic values ​​of each sensor For the first The offset threshold of the feature values ​​of each sensor For time window Inner The moving average of the characteristic values ​​of each sensor time Time The standardized offset of the feature values ​​of each sensor.

Images