A method, system, device and medium for controlling rapid fear stress decision-making of a brain-like model

Abstract

The present invention discloses a method, system, device and medium for controlling rapid fear stress decision-making of a brain-like system, including: acquiring environmental perception information of an unmanned platform; identifying target data from the environmental perception information of the unmanned platform to obtain a target threat level label; determining the three-dimensional position and three-dimensional velocity of a target according to the environmental perception information of the unmanned platform, and predicting the target trajectory according to the three-dimensional position and three-dimensional velocity of the target; making a fear stress decision according to the target threat level label and the predicted target trajectory. The method, system, device and medium can improve the recognition accuracy and speed of the unmanned platform, and at the same time, the recognition power consumption is relatively low.

Description

Technical Field

[0001] The present invention belongs to the technical field of control, and relates to a decision-making control method, system, device and medium, and particularly relates to a brain-like rapid fear stress decision-making control method, system, device and medium. Background Art

[0002] Various unmanned platforms such as unmanned aerial vehicles may face various unpredictable threats in the working environment, and a rapidly approaching target is one type of typical threat. It is crucial for an unmanned system to have the ability to autonomously sense threats and quickly respond to threats to maintain its survival.

[0003] Frame imaging cameras are often used in the image sensing sensors of unmanned systems. The exposure time of frame imaging cameras ranges from 1 ms to 100 ms. The physical constraint of the exposure time makes it difficult for them to be competent for the perception of fast objects, and this limitation is also difficult to be effectively compensated by complex algorithms. An event camera is a bionic sensor different from traditional frame cameras. It does not capture images by measuring the absolute brightness at a constant rate, but asynchronously measures the brightness changes of each pixel and outputs an event stream, encoding the time, position and sign of the brightness changes. It has a wide imaging dynamic range and a high imaging frame rate, is applicable to scenes with more drastic light changes while ensuring an extremely high imaging speed, and thus is applicable to the recognition of high-speed threat targets.

[0004] Traditional machine vision technology can be used for threat target detection, and its detection process is based on image pixel values, etc. However, traditional machine vision has a slow calculation speed under resource-constrained conditions. An artificial neural network can relatively quickly complete the recognition of threat targets. However, the recognition of threat targets requires a large amount of resources, and the threat target recognition under the high energy consumption of ANN hinders the use of threat recognition on unmanned devices with strict energy consumption management. As the third-generation neural network, the low-power spiking neural network is a potential solution to realize threat recognition algorithms applicable to embedded and mobile terminals. In the network, due to the way of pulse transmission, the spiking neural network adopts an accumulation operation with less energy consumption. The strong biological similarity and low energy consumption make the spiking neural network have great application potential in low-energy consumption threat recognition. However, in a complex environment, it is difficult to ensure the survival rate of an unmanned system only relying on the spiking neural network.

[0005] Currently, although some technologies in the prior art use spiking neural networks for threat recognition, the recognition error of the target in actual use is relatively large. If the recognition accuracy is to be improved, the network scale needs to be increased, further increasing the power consumption. Therefore, it is an urgent problem to improve the recognition accuracy and speed and ensure low-power recognition when an unmanned system conducts threat recognition. Summary of the Invention

[0006] The purpose of the present invention is to overcome the above-mentioned disadvantages of the prior art, and provide a method, system, device and medium for controlling brain-like rapid fear stress decision-making, which can improve the recognition accuracy and speed of unmanned platforms, and at the same time have relatively low recognition power consumption.

[0007] To achieve the above object, the present invention discloses a method for controlling brain-like rapid fear stress decision-making, including:

[0008] Obtain the environmental perception information of the unmanned platform;

[0009] Perform target data recognition on the environmental perception information of the unmanned platform to obtain a target threat level label;

[0010] According to the environmental perception information of the unmanned platform, determine the three-dimensional position and three-dimensional speed of the target, and predict the target trajectory based on the three-dimensional position and three-dimensional speed of the target;

[0011] Make a fear stress decision based on the target threat level label and the predicted target trajectory.

[0012] The process of obtaining the environmental perception information of the unmanned platform is as follows:

[0013] Obtain event camera data of an object approaching rapidly through an event camera;

[0014] Obtain the depth information of an object approaching rapidly through a binocular camera;

[0015] Obtain the displacement information, angular velocity information and acceleration information of the unmanned platform through an attitude sensor.

[0016] The process of performing target data recognition on the environmental perception information of the unmanned platform to obtain a target threat level label is as follows:

[0017] Perform a coarsening operation on the event camera data;

[0018] Input the coarsened event camera data into a spiking neural network to obtain a target threat level label.

[0019] Before inputting the coarsened event camera data into the spiking neural network, it further includes:

[0020] Obtain an event camera data set with target classification labels;

[0021] Construct a multi-layer feedforward spiking neural network based on LIF spiking neurons;

[0022] Use the event camera data set with target classification labels to train the feedforward spiking neural network to obtain the spiking neural network.

[0023] The loss function during the training of the feedforward spiking neural network using the target classification label event camera dataset is as follows:

[0024]

[0025] where is the set of threatening samples.

[0026] Based on the three-dimensional position and three-dimensional velocity of the target, the target trajectory is predicted using the Kalman filtering method.

[0027] The specific operation of performing fear stress decision control according to the target threat level label and the predicted target trajectory is as follows:

[0028] Adjust the stress level according to the target threat level label, and determine whether the target poses a threat based on the predicted target trajectory;

[0029] When the adjusted stress level exceeds the preset threshold, a preparation instruction is issued;

[0030] When the adjusted stress level exceeds the preset threshold and the target poses a threat, an execution instruction is issued.

[0031] The present invention discloses a brain-inspired fast fear stress decision control system, including:

[0032] A perception module for obtaining environmental perception information of the unmanned platform;

[0033] A fast-path target recognition module for performing target data recognition on the environmental perception information of the unmanned platform to obtain a target threat level label;

[0034] A slow-path calculation module for determining the three-dimensional position and three-dimensional velocity of the target based on the environmental perception information of the unmanned platform, and predicting the target trajectory based on the three-dimensional position and three-dimensional velocity of the target;

[0035] A stress decision coupling module for performing fear stress decision according to the target threat level label and the predicted target trajectory.

[0036] The present invention discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the brain-inspired fast fear stress decision control method are implemented.

[0037] The present invention discloses a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the brain-inspired fast fear stress decision control method are implemented.

[0038] The present invention has the following beneficial effects:

[0039] When the brain-inspired rapid fear stress decision control method, system, device and medium of the present invention are specifically operated, two information processing channels, fast and slow, are established to process the environmental perception information, that is, the target data of the environmental perception information of the unmanned platform is identified to obtain the target threat level label; the target trajectory is predicted according to the environmental perception information of the unmanned platform, and then the fear stress decision is made according to the output information of the fast and slow information processing channels. It can not only use the fast response of the fast channel to prepare for action execution, ensure the rapid response of the unmanned platform under low power consumption, but also use the accurate calculation of the slow channel to obtain the target state information, realize the accurate response to the target, and ensure the recognition accuracy and speed of the unmanned platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is the flowchart of the method of the present invention;

[0041] Figure 2 is the system structure diagram of the present invention;

[0042] Figure 3 is the control principle diagram of the present invention.

[0043] Figure 4 is the schematic diagram of the double-channel coupling decision response model. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] In order to enable those skilled in the art of the present technology to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments, and are not intended to limit the scope of the present invention disclosure. In addition, in the following description, the description of known structures and technologies is omitted to avoid unnecessarily confusing the concepts disclosed in the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0045] The structural schematic diagrams according to the disclosed embodiments of the present invention are shown in the accompanying drawings. These figures are not drawn to scale, and for the purpose of clear expression, some details are enlarged and some details may be omitted. The shapes of various regions and layers shown in the figures and their relative sizes and positional relationships are only exemplary, and may actually deviate due to manufacturing tolerances or technical limitations. Those skilled in the art can design regions / layers with different shapes, sizes and relative positions according to actual needs.

[0046] Embodiment 1

[0047] Reference Figure 1 , the brain-like rapid fear stress decision-making control method described in the present invention includes:

[0048] Obtain the environmental perception information of the unmanned platform;

[0049] Identify the target data from the environmental perception information of the unmanned platform to obtain the target threat level label;

[0050] Determine the three-dimensional position and three-dimensional velocity of the target according to the environmental perception information of the unmanned platform, and predict the target trajectory according to the three-dimensional position and three-dimensional velocity of the target;

[0051] Make a fear stress decision according to the target threat level label and the predicted target trajectory.

[0052] In this embodiment, the process of obtaining the environmental perception information of the unmanned platform is as follows:

[0053] Obtain the event camera data of the object approaching rapidly through the event camera;

[0054] Obtain the depth information of the object approaching rapidly through the binocular camera;

[0055] Obtain the displacement information, angular velocity information and acceleration information of the unmanned platform through the attitude sensor.

[0056] In this embodiment, the process of identifying the target data from the environmental perception information of the unmanned platform to obtain the target threat level label is as follows:

[0057] Perform a coarsening operation on the event camera data;

[0058] Input the coarsened event camera data into the spiking neural network to obtain the target threat level label, that is, the threat and non-threat labels.

[0059] In this embodiment, between inputting the coarsened event camera data into the spiking neural network, it also includes:

[0060] Obtain the event camera data set with target classification labels;

[0061] Construct a multi-layer feedforward spiking neural network based on LIF spiking neurons;

[0062] Use the event camera data set with target classification labels to train the feedforward spiking neural network to obtain the spiking neural network.

[0063] In this embodiment, the loss function in the process of using the event camera data set with target classification labels to train the feedforward spiking neural network is:

[0064]

[0065] Among them, is the set of threatening samples.

[0066] In this embodiment, based on the three-dimensional position and three-dimensional velocity of the target, the target trajectory is predicted by using the Kalman filtering method.

[0067] In this embodiment, the specific operation of performing fear stress decision control according to the target threat level label and the predicted target trajectory is as follows:

[0068] Adjust the stress level according to the target threat level label, and judge whether the target poses a threat according to the predicted target trajectory;

[0069] When the adjusted stress level exceeds the preset threshold, a preparation instruction is issued;

[0070] When the adjusted stress level exceeds the preset threshold and the target poses a threat, an execution instruction is issued.

[0071] Embodiment 2

[0072] Refer to Figure 2 , the brain-like fast fear stress decision control system described in this embodiment includes a perception module, a fast-path recognition module, a slow-path calculation module, a stress decision coupling module, and an execution module.

[0073] In this embodiment, the perception module provides environmental perception information of the unmanned system for the fast-path recognition module and the slow-path calculation module through sensors. The perception information includes the internal state information of the unmanned system and the external environment information of the unmanned system, and further includes: event camera data output by an event camera, depth information output by a binocular camera, angular velocity and angular acceleration information, displacement, velocity, and acceleration information.

[0074] The fast-path recognition module includes a data preprocessing module and a spiking neural network. Among them, the data preprocessing module processes the event camera data output by the perception module and then inputs it into the spiking neural network. Target recognition is performed through the spiking neural network to obtain the target threat level label, and then the target threat level label is output to the stress decision coupling module.

[0075] Specifically, the data preprocessing module performs a coarse-graining operation on the event camera data output by the perception module to match the input data dimension of the spiking neural network.

[0076] It should be noted that the coarse-graining operation divides the event pixel space into coarse-grained regions, and then slides along the time axis with a time window of a preset width in a set direction, calculates the mean value of the time coordinates of all event pulse time coordinates in each coarse-grained space region within the time window, and this mean value is used as the time coordinate of a spatio-temporal coarse-grained event pulse.

[0077] The construction process of the pulse neural network is as follows: The feedforward pulse neural network is trained with the event camera dataset with target classification labels to obtain a pulse neural network, enabling it to receive the event camera data after the coarse-graining operation, quickly perform target recognition and classification, and output the target threat level label.

[0078] Specifically, the process of training the preset shallow feedforward pulse neural network on the event camera dataset with target classification labels includes:

[0079] 1) Establish an event camera dataset with target classification labels;

[0080] 2) Construct a multi-layer feedforward pulse neural network based on LIF pulse neurons;

[0081] 3) Input the event camera data with classification labels into the multi-layer feedforward pulse neural network;

[0082] 4) Calculate the difference between the multi-layer feedforward pulse neural network and the event camera data representing the data label, and update the loss function;

[0083] 5) Update the connection weights inside the multi-layer feedforward pulse neural network according to the loss function;

[0084] 6) Repeat inputting the event camera data with classification labels and the connection weights inside the multi-layer feedforward pulse neural network until the loss function is less than the expectation;

[0085] 7) Evaluate the multi-layer feedforward pulse neural network to obtain a pulse neural network.

[0086] Specifically, the dynamic model of the LIF pulse neuron is:

[0087]

[0088] Among them, -(v i -V r ) is the leakage term, and the leakage term is used to ensure that the neuron gradually tends to rest when there is no input drive; is the external drive term, p = [p1, p2,..., p n T is the external drive signal of the processed event camera, and p changes dynamically with time; is the SNN internal coupling term, δ(t - τ​kl ) represents the l-th spike discharge event of neuron k, and W ik ∈R m×m is the synaptic connection matrix.

[0089] In this embodiment, the spiking neural network adopts a classical architecture, which is composed of a convolutional layer, a spiking neuron layer and an integration layer. The neurons use classical spiking neurons - integrate-and-fire neurons, and Spike max-pooling is selected for pooling. A structure of 6 convolutional layers - 2 fully connected layers - 1 voting layer is used. The size of all convolutional layers is set to 3, stride = 1, padding = 1, and the number of convolutional channels is 128. After each convolutional layer, a batch normalization layer is added. All pooling layers are set to kernel size = 2 and stride = 2. Before each fully connected layer, a dropout layer is added and remains constant throughout the duration, which can ensure that some neurons are continuously inhibited at each time stage. The output layer contains two neurons, which perform binary classification to determine whether the target approximation is far from a threat.

[0090] To ensure absolute accuracy during target detection, the training strategy of the spiking neural network needs to be adjusted to ensure the precision rate. In the present invention, during the training process of the spiking neural network, the loss function MSE used is:

[0091]

[0092] where is the set of threatening samples, β > 0, then the contribution of threatening samples to the loss function is greater, meeting the requirement of high precision rate.

[0093] In this embodiment, the F1-score is used to evaluate the trained model. Among them, the F1-score combines the precision and recall to reflect the performance of the model. The specific form is:

[0094]

[0095] The slow-path calculation module receives the environmental perception information output by the perception module, filters, corrects, time-aligns and clusters and segments the environmental perception information, then calculates the three-dimensional position and velocity, obtains the three-dimensional position, three-dimensional velocity data and trajectory prediction data of the target, and sends them to the stress decision coupling module, and establishes a repulsive force field model during the target approaching movement and outputs it to the coupling decision module.

[0096] It should be noted that in the present invention, the Kalman filtering method is adopted for target trajectory prediction. It should be noted that the Kalman filter is a highly efficient recursive filter that can use the estimated value of the previous moment and the observed value of the current moment to iteratively update the estimation of the state variable, and then predict the trajectory position of the next moment. The state equation and the observation equation for the Kalman filtering dynamic trajectory prediction are as follows:

[0097] x k =A k x k-1 +u k +w k

[0098] z k =C k x k +v k

[0099] Among them, the state at time k - 1 is x k-1 , and its corresponding covariance is The state at time k is x k , z k is the observed value at time k. By predicting the state x k at time k and the observation variable v k , the observed value z k is calculated.

[0100] The iterative process is as follows:

[0101] Prediction state stage:

[0102]

[0103]

[0104] Update the Kalman gain K:

[0105]

[0106] Update the posterior probability distribution based on the Kalman gain K:

[0107]

[0108]

[0109] Iterate according to the input information to achieve the prediction of the target trajectory.

[0110] In this embodiment, the repulsive force field model during the target approaching motion is:

[0111]

[0112] Among them, γ and η0 are adjustable parameters, and η i is the distance from the agent to the i-th obstacle, and is the moving speed of the obstacle.

[0113] The stress decision coupling module processes the received data and outputs an execution instruction to the execution module. Among them, the stress decision coupling module receives the target threat level label output by the fast-path target recognition module, integrates and makes a decision on this information, and outputs an instruction to the execution module; the stress decision coupling module receives the information output by the slow-path calculation module, integrates and makes a decision on this information, and outputs an instruction to the execution module.

[0114] The process that the stress decision coupling module receives the target threat level label of the fast-path target recognition module, integrates and makes a decision on this information, and outputs an instruction to the execution module includes:

[0115] 2a) Receiving the target threat level label output by the fast-path target recognition module;

[0116] 2b) Sending a preparation instruction to the execution module;

[0117] 2c) The execution module increases the system fear stress level parameter according to the received preparation instruction.

[0118] The process of receiving the output information of the slow-path calculation module, integrating and making a decision on the information, and outputting an instruction to the execution module includes:

[0119] 3a) Receiving the output information of the slow-path calculation module;

[0120] 3b) When the target trajectory does not coincide with itself, the system fear stress level parameter is reduced;

[0121] When the system fear level exceeds the fear level threshold F th , an avoidance instruction is sent to the execution module;

[0122] Preferably, when the fast-path target recognition module discovers a threat, the fear level is increased to F(t + t fast ):

[0123] F(t + t fast ) = F(t) + αδ fast (t)

[0124] Among them, t fast is the working period of the fast path, and α is a parameter controlling the importance of fast-path decision-making.

[0125] After the slow-path calculation module discovers a false alarm, it suppresses the system fear level to F(t + t slow ):

[0126] F(t + t slow ) = max(F(t) - βδ fast (t), 0)

[0127] where t slow is the duty cycle of the slow path, β is a parameter controlling the importance of slow path decision-making, and α < β.

[0128] When the stress decision coupling module does not receive dual-path stimulation, the fear level decays exponentially:

[0129]

[0130] where τ is the time constant and γ is the decay factor.

[0131] Preferably, the stress decision coupling module has five parameters: α, β, τ, γ, F th . Manually adjusting the parameters depends on experience and takes a lot of time. In this embodiment, the parameters are autonomously learned through a reinforcement learning method. As Figure 3 shown, the drone in the reinforcement learning environment operates according to a preset reward rule and adjusts the parameters in the stress decision coupling module according to the actual operation situation, and finally converges the parameters to meet the requirements according to different tasks.

[0132] As Figure 4 shown, when the system receives the output information of the fast path recognition module and the slow path calculation module, it will change the stress level in the stress decision coupling module. There are three basic states of the stress level:

[0133] State Ⅰ: A short-term change in the stress level due to the passing of a slow object or other object, but as time goes by and no new stimulation appears, the stress level of the stress module decreases;

[0134] State Ⅱ: Since the target object is flying towards the unmanned vehicle platform, the fast path continuously outputs high-threat level information stimuli, causing the stress level in the stress decision coupling module to continuously increase until it reaches the threshold F th , and the stress decision coupling module outputs an instruction;

[0135] State Ⅲ: Since the target object is flying towards the unmanned vehicle platform, the fast path continuously outputs high-threat level information stimuli, but the slow path calculation module determines that the target is not threatening after calculating the input information and outputs information to the stress decision coupling module to inhibit the stress level.

[0136] In three states, the stress decision coupling module issues instructions. In state I, the system issues a preparation instruction; in state II, when the system receives the information input from the fast path, it first issues a preparation instruction and increases the stress level. When the stress value exceeds the threshold F th then it issues an execution instruction; in state III, when the system first issues a preparation instruction and increases the stress level, it does not issue an execution instruction. The execution module receives the output instruction from the stress decision coupling module. When receiving the preparation instruction, the execution module calibrates its own state and makes execution preparations; when receiving the execution instruction, the execution module performs an avoidance execution action based on its own state information on the basis of execution preparations.

[0137] Embodiment III

[0138] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the brain-like fast fear stress decision control method are implemented. Among them, the memory may include internal memory, such as high-speed random access memory, and may also include non-volatile memory, such as at least one disk memory, etc.; the processor, network interface, and memory are interconnected through an internal bus, and this internal bus can be an Industry Standard Architecture bus, a Peripheral Component Interconnect standard bus, an Extended Industry Standard Architecture bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory is used to store programs. Specifically, the program may include program code, and the program code includes computer operation instructions. The memory may include internal memory and non-volatile memory and provide instructions and data to the processor.

[0139] Embodiment IV

[0140] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the brain-like fast fear stress decision control are implemented. Specifically, the computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may include random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include read-only memory (ROM), hard disk, flash memory, optical disc, magnetic disk, etc.

[0141] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0142] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0143] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0144] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are performed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0145] 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 above embodiments, those of ordinary skill in the art should understand that: still, the specific implementation manners of the present invention can be modified or equivalently replaced, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A method for controlling brain-like rapid fear stress decision-making, characterized in that, Including: Obtain the environmental perception information of the unmanned platform; Perform target data recognition on the environmental perception information of the unmanned platform to obtain a target threat level label; Determine the three-dimensional position and three-dimensional velocity of the target according to the environmental perception information of the unmanned platform, and predict the target trajectory according to the three-dimensional position and three-dimensional velocity of the target; Make a fear stress decision according to the target threat level label and the predicted target trajectory; The process of obtaining the environmental perception information of the unmanned platform is as follows: Obtain event camera data of approaching objects through an event camera; Obtain depth information of fast approaching objects through a binocular camera; Obtain displacement information, angular velocity information and acceleration information of the unmanned platform through an attitude sensor; The specific operation of making a fear stress decision control according to the target threat level label and the predicted target trajectory is as follows: Adjust the stress level according to the target threat level label, and judge whether the target poses a threat according to the predicted target trajectory; When the adjusted stress level exceeds the preset threshold, issue a preparation instruction; When the adjusted stress level exceeds the preset threshold and the target poses a threat, issue an execution instruction.

2. The method for controlling brain-like rapid fear stress decision-making according to claim 1, characterized in that, The process of performing target data recognition on the environmental perception information of the unmanned platform to obtain a target threat level label is as follows: Perform a coarsening operation on the event camera data; Input the coarsened event camera data into a spiking neural network to obtain a target threat level label; The coarsening operation divides the event pixel space into coarsened regions, and then slides along the time axis with a time window of a preset width in a set direction, and calculates the mean value of the time coordinates of all event pulses in each coarsened spatial region within the time window, and this mean value is used as the time coordinate of a spatio-temporal coarsened event pulse.

3. The method for controlling brain-like rapid fear stress decision-making according to claim 1, characterized in that, Before inputting the coarsened event camera data into the spiking neural network, it also includes: Obtain an event camera data set with target classification labels; Construct a multi-layer feedforward spiking neural network based on LIF spiking neurons; Use the event camera data set with target classification labels to train the feedforward spiking neural network to obtain the spiking neural network.

4. The method for controlling brain-like rapid fear stress decision-making according to claim 3, characterized in that, The loss function in the process of using the event camera data set with target classification labels to train the feedforward spiking neural network is: Among them, is the set of threatening samples.

5. The method for controlling brain-like rapid fear stress decision-making according to claim 1, characterized in that, Predict the target trajectory based on the three-dimensional position and three-dimensional velocity of the target by using the Kalman filtering method.

6. A brain-like rapid fear stress decision control system, characterized in that, Including: A perception module for obtaining the environmental perception information of the unmanned platform; A fast-path target recognition module for performing target data recognition on the environmental perception information of the unmanned platform to obtain a target threat level label; A slow-path calculation module for determining the three-dimensional position and three-dimensional velocity of the target according to the environmental perception information of the unmanned platform, and predicting the target trajectory according to the three-dimensional position and three-dimensional velocity of the target; A stress decision coupling module for making a fear stress decision according to the target threat level label and the predicted target trajectory; The specific operation of making a fear stress decision control according to the target threat level label and the predicted target trajectory is: Adjust the stress level according to the target threat level label, and determine whether the target poses a threat based on the predicted target trajectory; When the adjusted stress level exceeds a preset threshold, a preparation instruction is issued; When the adjusted stress level exceeds the preset threshold and the target poses a threat, an execution instruction is issued.

7. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, the steps of the brain-inspired rapid fear stress decision-making control method according to any one of claims 1-5 are implemented.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, the steps of the brain-inspired rapid fear stress decision-making control method according to any one of claims 1-5 are implemented.

Images