Feedforward Control Method for Multirotor UAVs Based on Dynamic Cascaded Pulse Neural Network

Abstract

This invention belongs to the field of unmanned aerial vehicle (UAV) control technology, and particularly relates to a feedforward control method for multi-rotor UAVs based on a dynamic cascaded spiking neural network (RCN). The method includes: S1: constructing a network model of the RCN based on pulse signals; S2: constructing a cost function of the RCN based on the network model and pulse errors; S3: solving for the weights of the RCN; S4: setting a preset similarity threshold and determining the learning rules for the dynamic cascaded structure based on the preset similarity threshold; S5: obtaining the output signal of the RCN according to the weights and the learning rules of the dynamic cascaded structure, and implementing feedforward control of the multi-rotor UAV based on the output signal. This invention improves the adaptability and robustness of multi-rotor UAVs in complex flight environments.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) control technology, and particularly relates to a feedforward control method for multi-rotor UAVs based on a dynamic cascaded pulse neural network. Background Technology

[0002] In recent years, multi-rotor drones have shown broad application prospects in both military and civilian fields due to their flexible flight modes, high maneuverability, and vertical takeoff and landing capabilities. As the application of multi-rotor drones becomes more widespread, their flight environments are becoming increasingly complex, even requiring them to operate in complex terrains such as high-altitude mountains, jungles, narrow building clusters, and lakes. These flight environments often feature variable wind disturbances and complex uncertainties, which adversely affect the flight control of multi-rotor drones.

[0003] Composite control, based on feedback controllers, introduces feedforward control to further improve the dynamic response performance of the system, enabling it to suppress complex environmental disturbances in a timely manner without waiting for feedback errors to occur. Currently, due to the adaptive, self-organizing, and self-learning characteristics of artificial neural networks, many researchers are using them as feedforward controllers to realize composite control of unmanned aerial vehicles (UAVs).

[0004] As a third-generation neural network model with a near-biomimetic mechanism, spiking neural networks operate more like the mammalian brain. Compared to artificial neural networks, they possess stronger nonlinear computation, asynchronous event information processing, and self-learning capabilities, enabling more efficient human-like perception, cognition, and decision-making. However, spiking neural networks face the challenge of needing to predict network structure information and sample proportion distribution when dealing with data changes and unstable environments. Currently, they cannot be effectively applied in the feedforward control of multi-rotor UAVs. Summary of the Invention

[0005] In view of this, the present invention aims to provide a feedforward control method for multi-rotor UAVs based on dynamic cascaded spiking neural networks, in order to solve the problem that spiking neural networks cannot be well applied in the supervised feedforward control of multi-rotor UAVs. The present invention combines the idea of ​​dynamic cascaded structure to construct a new network framework, which overcomes the problem that network structure information, sample ratio distribution and other information must be prior knowledge, thereby improving the adaptability and robustness of multi-rotor UAVs in complex flight environments.

[0006] To achieve the above objectives, the technical solution created by this invention is implemented as follows:

[0007] A feedforward control method for a multi-rotor unmanned aerial vehicle (UAV) based on a dynamic cascaded pulse neural network includes the following steps:

[0008] S1: Acquire the pulse signal of the multi-rotor UAV, and construct a network model of dynamic cascaded pulse neural network based on the pulse signal. The pulse signal includes the expected pulse signal and the actual pulse signal.

[0009] S2: Cost function for constructing a dynamic cascaded spiking neural network based on the network model and pulse signal difference;

[0010] S3: Solving for the weights of the dynamic cascaded spiking neural network based on the initial preset parameters and cost function;

[0011] S4: Set a preset similarity threshold, and determine the dynamic cascade structure learning rule based on the preset similarity threshold. The dynamic cascade structure learning rule is used to dynamically adjust the number of hidden layer neurons, the number of hidden layers, and the weights of hidden layer neurons.

[0012] S5: Obtain the output signal of the dynamic cascaded spiking neural network based on the weights and dynamic cascaded structure learning rules, and realize feedforward control of the multi-rotor UAV based on the output signal.

[0013] Furthermore, in step S1, the network model of the dynamic cascaded spiking neural network... for:

[0014] ;

[0015] ;

[0016] ;

[0017] ;

[0018] ;

[0019] in, Indicates the resting potential. A transpose matrix representing the synaptic weights; , These are intermediate parameters representing the network model and have no physical meaning. To represent the intermediate parameters of the i-th synapse in the network model; This represents the weight of the i-th synapse; This represents the current core of the i-th synapse; Represents the unit step function; Indicates the current moment in the training process; Indicates the moment of the latest impulse emitted by the neuron; This represents a pulse signal; n represents the number of synapses. , , and These represent the weights of the first, second, and nth synapses, respectively. Indicates film capacitance; This represents the time constant of the neuron membrane.

[0020] Furthermore, in step S2, the cost function is:

[0021] ;

[0022] ;

[0023] ;

[0024] in, This represents the pulse signal difference between the desired pulse signal and the actual pulse signal. Indicates the desired membrane potential. Represents the cost function; Representing the network model; This represents the arrival time of the i-th desired pulse signal; This indicates the arrival time of the i-th actual pulse signal; Indicates the desired number of pulse signals; Indicates the number of actual pulse signals; This indicates the total training time.

[0025] Furthermore, in step S3, the weights of the dynamic cascaded spiking neural network are:

[0026] ;

[0027] ;

[0028] ;

[0029] ;

[0030] ;

[0031] ;

[0032] in, Indicates the expected weight; This indicates the increment for weight updates; Represents the cost function; express The approximate gradient; This represents the error function between the desired pulse signal and the actual pulse signal. Indicates the update rate coefficient; This represents an approximate impulse response signal; Indicates film capacitance; Indicates the current moment in the training process; This represents the time estimation constant of the neuron membrane; These are intermediate parameters representing approximate impulse response signals and have no physical meaning.

[0033] Furthermore, step S4 specifically includes the following steps:

[0034] S41: Set similarity threshold Threshold for the number of neurons in a single hidden layer and hidden layer preset values;

[0035] S42: Add neuron k to the hidden layer of the dynamic cascaded spiking neural network, and calculate the similarity between the weights of neuron k and the weights of the original neurons in the current hidden layer:

[0036] ;

[0037] in, This indicates that the newly added neuron k is compared with the existing i-th neuron. Weighted similarity; This represents the weight of the newly added neuron k; This represents the weight of the original i-th neuron in the dynamic cascaded spiking neural network; Represents a constant; This indicates the number of neurons in the current hidden layer;

[0038] S43: Determine similarity Is it less than the similarity threshold? If yes, proceed to step S44; otherwise, proceed to step S46.

[0039] S44: Determine whether the number of neurons in the hidden layer is less than or equal to the neuron count threshold. If so, add a new neuron in the hidden layer to record the new information and proceed to step S47; otherwise, proceed to step S45.

[0040] S45: Add a new hidden layer to the dynamic cascaded spiking neural network, add new neurons to the new hidden layer, replace the hidden layer in step S44 with the new hidden layer, and execute step S47.

[0041] S46: Merge the newly added neuron with the original neuron with the highest weight similarity and update the weight of the merged neuron.

[0042] S47: Replace neuron k with the newly added neuron k+1, and repeat steps S42~S46 until the number of hidden layers in the current dynamic cascaded spiking neural network is greater than the preset value of hidden layers, thus completing the dynamic adjustment of the number of hidden layer neurons, the number of hidden layers, and the weights of hidden layer neurons.

[0043] Furthermore, in step S46, the merged neuron weights are updated using the following formula:

[0044] ;

[0045] in, This indicates that neuron k was merged into the neuron with the highest similarity. The subsequent weights; This indicates the number of neurons in the current hidden layer; This represents the weight of neuron k.

[0046] Furthermore, in step S1, the desired pulse signal of the multi-rotor UAV is derived based on the desired trajectory, and the actual pulse signal is derived based on the actual flight trajectory of the multi-rotor UAV.

[0047] Furthermore, in step S2, the pulse signal difference is the difference between the desired pulse signal and the actual pulse signal.

[0048] Furthermore, in step S3, the initial preset parameters of the dynamic cascaded spiking neural network include the initial value of the number of hidden layer neurons, the initial value of the number of hidden layers, and the initial value of the weights of the hidden layer neurons, wherein the initial value of the number of hidden layers is set to 1.

[0049] Compared with the prior art, the present invention can achieve the following beneficial effects:

[0050] (1) The feedforward control method for multi-rotor UAVs based on dynamic cascaded spiking neural networks described in this invention combines the concept of dynamic cascading and proposes a novel dynamic cascaded spiking neural network, which can overcome the problem that the structural information and sample ratio distribution of spiking neural networks must be prior knowledge. The desired trajectory of the multi-rotor UAV and the output of the feedback controller are used as sample inputs for training, realizing supervised feedforward control of the dynamic cascaded spiking neural network, providing a new approach for feedforward control of multi-rotor UAVs.

[0051] (2) The feedforward control method for multi-rotor UAVs based on dynamic cascaded spiking neural networks described in this invention designs a supervisory weight and dynamic cascaded structure learning mechanism. By dynamically and adaptively increasing the hidden layer and hidden layer neurons to extract more external information centers, the multi-rotor UAV can adapt to the constantly changing unstable environment. When learning new information, it will avoid catastrophic interference or forgetting. Even if the input data changes continuously, it will not affect the normal operation of the dynamic cascaded spiking neural network, which greatly improves the stability and robustness of supervisory feedforward control. Attached Figure Description

[0052] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0053] Figure 1 A flowchart illustrating the feedforward control method for a multi-rotor UAV based on a dynamic cascaded pulse neural network, as described in an embodiment of the present invention.

[0054] Figure 2 A schematic diagram of the feedforward control structure of a multi-rotor UAV based on a dynamic cascaded pulse neural network, as described in an embodiment of the present invention;

[0055] Figure 3 A schematic diagram illustrating the dynamic evolution process of the hidden layer as described in the embodiments of the present invention;

[0056] Figure 4 A schematic diagram illustrating the dynamic evolution process of hidden layer neurons as described in an embodiment of the present invention;

[0057] Figure 5 A schematic diagram of longitude tracking of a twelve-rotor UAV under outdoor wind disturbance of level 4.

[0058] Figure 6 A schematic diagram of latitude tracking of a twelve-rotor UAV under outdoor wind disturbance of level 4.

[0059] Figure 7 A schematic diagram illustrating altitude tracking of a twelve-rotor UAV under outdoor wind conditions of force 4.

[0060] Figure 8 A schematic diagram of pitch angle tracking for a twelve-rotor UAV under outdoor wind disturbance of level 4.

[0061] Figure 9 A schematic diagram of roll angle tracking for a twelve-rotor UAV under outdoor level 4 wind disturbance;

[0062] Figure 10 This is a schematic diagram of yaw angle tracking of a twelve-rotor UAV under outdoor wind disturbance of level 4. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof.

[0064] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0065] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0066] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0067] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0068] like Figure 1-Figure 2As shown, this invention proposes a feedforward control method for a multi-rotor UAV based on a dynamic cascaded spiking neural network, specifically including the following steps: S1: Acquire the pulse signal of the multi-rotor UAV, and construct a network model of a dynamic cascaded spiking neural network based on the pulse signal. The pulse signal includes the expected pulse signal and the actual pulse signal; S2: Construct the cost function of the dynamic cascaded spiking neural network based on the difference between the network model and the pulse signal; S3: Solve the weights of the dynamic cascaded spiking neural network based on the initial preset parameters and the cost function; S4: Set a preset similarity threshold, and determine the dynamic cascaded structure learning rule based on the preset similarity threshold. The dynamic cascaded structure learning rule is used to dynamically adjust the number of hidden layer neurons, the number of hidden layers, and the weights of hidden layer neurons; S5: Obtain the output signal of the dynamic cascaded spiking neural network according to the weights of the dynamic cascaded spiking neural network and the dynamic cascaded structure learning rule, and realize feedforward control of the multi-rotor UAV based on the output signal.

[0069] This invention first establishes a dynamic cascaded spiking neural network model; determines the cost function of the dynamic cascaded spiking neural network, and then designs a supervised weight learning rule (the weights of the dynamic cascaded spiking neural network); determines the dynamic cascaded structure learning rule, and finally realizes supervised feedforward control of the dynamic cascaded spiking neural network, improving the dynamic control performance and disturbance rejection capability of multi-rotor UAVs in complex flight environments. The supervised feedforward control method for trajectory tracking of multi-rotor UAVs based on dynamic cascaded spiking neural networks provides a new approach for the research of trajectory tracking control of multi-rotor UAVs, and has certain theoretical reference significance and practical application value.

[0070] In some embodiments, in step S1, the network model of the dynamic cascaded spiking neural network is:

[0071] ;

[0072] ;

[0073] ;

[0074] ;

[0075] ;

[0076] in, Indicates the resting potential. A transpose matrix representing the synaptic weights; , These are intermediate parameters representing the network model and have no physical meaning. To represent the intermediate parameters of the i-th synapse in the network model; This represents the weight of the i-th synapse; This represents the current core of the i-th synapse; Represents the unit step function; Indicates the current moment in the training process; Indicates the moment of the latest impulse emitted by the neuron; The input pulse signal is represented by 'n' (if the input pulse is the desired pulse, the result is the desired pulse signal; if the input pulse is the actual pulse, the result is the actual pulse signal; there is no need to confirm the signal type here); 'n' represents the number of synapses. , , and These represent the weights of the first, second, and nth synapses, respectively. Indicates film capacitance; This represents the time constant of the neuron membrane.

[0077] In some embodiments, in step S2, the cost function is:

[0078] ;

[0079] ;

[0080] ;

[0081] in, This represents the pulse signal difference between the desired pulse signal and the actual pulse signal. Indicates the desired membrane potential. Represents the cost function; Representing the network model; This represents the arrival time of the i-th desired pulse signal; This indicates the arrival time of the i-th actual pulse signal; Indicates the desired number of pulse signals; Indicates the number of actual pulse signals; This indicates the total training time.

[0082] In some embodiments, in step S3, the weights of the dynamic cascaded spiking neural network are:

[0083] ;

[0084] ;

[0085] ;

[0086] ;

[0087] ;

[0088] ;

[0089] in, Indicates the expected weight; This indicates the increment for weight updates; Represents the cost function; express The approximate gradient; This represents the error function between the desired pulse signal and the actual pulse signal. Indicates the update rate coefficient; This represents an approximate impulse response signal; Indicates film capacitance; Indicates the current moment in the training process; This represents the time estimation constant of the neuron membrane; These are intermediate parameters representing approximate impulse response signals and have no physical meaning.

[0090] It should be noted that, due to the discontinuous transitions in the weight space, for simplicity, the cost function is minimized at each time step.

[0091] In some embodiments, step S4 specifically includes the following steps: S41: Setting a similarity threshold Threshold for the number of neurons in a single hidden layer and hidden layer preset values;

[0092] S42: Add neuron k to the hidden layer of the dynamic cascaded spiking neural network, and calculate the similarity between the weights of neuron k and the weights of the original neurons in the current hidden layer:

[0093] ;

[0094] in, This indicates that the newly added neuron k is compared with the existing i-th neuron. Weighted similarity; This represents the weight of the newly added neuron k; This represents the weight of the original i-th neuron in the dynamic cascaded spiking neural network; Represents a constant; This indicates the number of neurons in the current hidden layer;

[0095] S43: Determine similarity Is it less than the similarity threshold? If yes, proceed to step S44; otherwise, proceed to step S46.

[0096] S44: Determine whether the number of neurons in the hidden layer is less than or equal to the neuron count threshold. If so, add a new neuron in the hidden layer to record the new information and proceed to step S47; otherwise, proceed to step S45.

[0097] S45: Add a new hidden layer to the dynamic cascaded spiking neural network, add new neurons to the new hidden layer, replace the hidden layer in step S44 with the new hidden layer, and execute step S47.

[0098] S46: Merge the newly added neuron with the original neuron with the highest weight similarity and update the weight of the merged neuron.

[0099] S47: Replace neuron k with the newly added neuron k+1, and repeat steps S42~S46 until the number of hidden layers in the current dynamic cascaded spiking neural network is greater than the preset value of hidden layers, thus completing the dynamic adjustment of the number of hidden layer neurons, the number of hidden layers, and the weights of hidden layer neurons.

[0100] It should be noted that in step S44, if the similarity... Less than the similarity threshold Furthermore, the number of hidden layer neurons in the current dynamic cascaded spiking neural network is no greater than the neuron count threshold. This means that the information in the network differs significantly from the input sample. New neurons are added to the hidden layer to record new information, thereby determining the number of neurons in the hidden layer.

[0101] If similarity Less than the similarity threshold Furthermore, the number of neurons in the hidden layer is greater than the threshold for the number of neurons. New neurons are added to new hidden layers because neurons in the same hidden layer have similar functions, and adding too many neurons would lead to repetitive network training, which is detrimental to network stability. If the number of dynamically cascaded hidden layers exceeds a preset value, network learning optimization is terminated to avoid catastrophic network expansion, thus determining the number of hidden layers.

[0102] Furthermore, in step S45, when the input sample information is sufficiently similar to the information in the original neurons of the network, i.e., the similarity is equal to or greater than... Then the new neuron will be merged with the neuron with the highest weight similarity, thereby determining the weights of the hidden layer neurons.

[0103] Furthermore, similarity threshold Threshold for the number of neurons in a single hidden layer The default values ​​for the hidden layer are derived based on user needs.

[0104] In some embodiments, in step S46, the merged neuron weights are updated using the following formula:

[0105] ;

[0106] in, This indicates that neuron k was merged into the neuron with the highest similarity. The subsequent weights; This indicates the number of neurons in the current hidden layer; This represents the weight of neuron k.

[0107] In some embodiments, in step S1, the desired pulse signal of the multi-rotor UAV is derived based on the desired trajectory, and the actual pulse signal is derived based on the actual flight trajectory of the multi-rotor UAV.

[0108] In some embodiments, in step S2, the pulse signal difference is the difference between the desired pulse signal and the actual pulse signal.

[0109] In some embodiments, in step S3, the initial preset parameters of the dynamic cascaded spiking neural network include the initial value of the number of hidden layer neurons, the initial value of the number of hidden layers, and the initial value of the weights of the hidden layer neurons, wherein the initial value of the number of hidden layers is set to 1.

[0110] The following uses a twelve-rotor UAV as an example to illustrate the feedforward control method for multi-rotor UAVs based on dynamic cascaded pulse neural networks provided by this invention:

[0111] To verify the robustness of this invention, a trajectory tracking flight experiment of a twelve-rotor UAV was conducted in outdoor windy conditions at level 4. The maximum instantaneous wind speed in the flight environment was measured to be 11 m / s using an Omega anemometer. The code for the trajectory tracking supervised feedforward control method of this invention was written and input into the twelve-rotor airborne control main chip to complete the trajectory tracking flight experiment of the twelve-rotor UAV. The parameters of the dynamic cascaded pulse neural network were set as follows: The initial number of hidden layer neurons is 1, and the threshold of the hidden layer neurons is... .

[0112] like Figures 3-4 As shown, through the learning rule mechanism, the number of hidden layers eventually converges to 4, and the number of hidden neurons in each hidden layer converges to 30, 30, 30 and 22. Figures 5-7 The tracking results of a twelve-rotor UAV based on a dynamic cascaded pulse neural network in longitude, latitude, and altitude are described. Figures 8-10The attitude tracking results of a 12-rotor UAV are described, demonstrating excellent tracking performance. It is evident that, based on the feedforward control method provided by this invention, the trajectory tracking error of the 12-rotor UAV in all three directions is within ±0.8 meters under level 4 wind disturbance. Experiments show that the method designed in this invention can guarantee excellent trajectory tracking control performance for the 12-rotor UAV and can resist external level 4 wind disturbance, exhibiting strong robustness.

[0113] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0114] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A feedforward control method for a multi-rotor unmanned aerial vehicle based on a dynamic cascaded pulse neural network, characterized in that, Specifically, the steps include the following: S1: Acquire the pulse signal of the multi-rotor UAV, and construct a network model of dynamic cascaded pulse neural network based on the pulse signal. The pulse signal includes the expected pulse signal and the actual pulse signal. S2: Cost function for constructing a dynamic cascaded spiking neural network based on the network model and pulse signal difference; S3: Solving for the weights of the dynamic cascaded spiking neural network based on the initial preset parameters and cost function; S4: Set a preset similarity threshold, and determine the dynamic cascade structure learning rule based on the preset similarity threshold. The dynamic cascade structure learning rule is used to dynamically adjust the number of hidden layer neurons, the number of hidden layers, and the weights of hidden layer neurons. Step S4 specifically includes the following steps: S41: Set similarity threshold Threshold for the number of neurons in a single hidden layer and hidden layer preset values; S42: Add neuron k to the hidden layer of the dynamic cascaded spiking neural network, and calculate the similarity between the weights of neuron k and the weights of the original neurons in the current hidden layer: ； in, This indicates that the newly added neuron k is compared with the existing i-th neuron. Weighted similarity; This represents the weight of the newly added neuron k; This represents the weight of the original i-th neuron in the dynamic cascaded spiking neural network; Represents a constant; This indicates the number of neurons in the current hidden layer; S43: Determine similarity Is it less than the similarity threshold? If yes, proceed to step S44; otherwise, proceed to step S46. S44: Determine whether the number of neurons in the hidden layer is less than or equal to the neuron count threshold. If so, add a new neuron in the hidden layer to record the new information and proceed to step S47; otherwise, proceed to step S45. S45: Add a new hidden layer to the dynamic cascaded spiking neural network, add new neurons to the new hidden layer, replace the hidden layer in step S44 with the new hidden layer, and execute step S47. S46: Merge the newly added neuron with the original neuron with the highest weight similarity and update the weight of the merged neuron. In step S46, the merged neuron weights are updated using the following formula: ； in, This indicates that neuron k was merged into the neuron with the highest similarity. The subsequent weights; This indicates the number of neurons in the current hidden layer; This represents the weight of neuron k; S47: Replace neuron k with the newly added neuron k+1, and repeat steps S42~S46 until the number of hidden layers in the current dynamic cascaded spiking neural network is greater than the preset value of hidden layers, thus completing the dynamic adjustment of the number of hidden layer neurons, the number of hidden layers, and the weights of hidden layer neurons. S5: Obtain the output signal of the dynamic cascaded spiking neural network based on the weights and dynamic cascaded structure learning rules, and realize feedforward control of the multi-rotor UAV based on the output signal.

2. The feedforward control method for a multi-rotor UAV based on a dynamic cascaded pulse neural network according to claim 1, characterized in that, In step S1, the network model of the dynamic cascaded spiking neural network... for: ； ； ； ； ； in, Indicates the resting potential. A transpose matrix representing the synaptic weights; , These are intermediate parameters representing the network model and have no physical meaning. To represent the intermediate parameters of the i-th synapse in the network model; This represents the weight of the i-th synapse; This represents the current core of the i-th synapse; Represents the unit step function; Indicates the current moment in the training process; Indicates the moment of the latest impulse emitted by the neuron; This represents a pulse signal; n represents the number of synapses. , , and These represent the weights of the first, second, and nth synapses, respectively. Indicates film capacitance; This represents the time constant of the neuron membrane.

3. The feedforward control method for a multi-rotor UAV based on a dynamic cascaded pulse neural network according to claim 1, characterized in that, In step S2, the cost function is: ； ； ； in, This represents the pulse signal difference between the desired pulse signal and the actual pulse signal. Indicates the desired membrane potential. Represents the cost function; Representing the network model; This represents the arrival time of the i-th desired pulse signal; This indicates the arrival time of the i-th actual pulse signal; Indicates the desired number of pulse signals; Indicates the number of actual pulse signals; This indicates the total training time.

4. The feedforward control method for a multi-rotor UAV based on a dynamic cascaded pulse neural network according to claim 3, characterized in that, In step S3, the weights of the dynamic cascaded spiking neural network are: ； ； ； ； ； ； in, Indicates the expected weight; This indicates the increment for weight updates; Represents the cost function; express The approximate gradient; This represents the error function between the desired pulse signal and the actual pulse signal. Indicates the update rate coefficient; This represents an approximate impulse response signal; Indicates film capacitance; Indicates the current moment in the training process; This represents the time estimation constant of the neuron membrane; These are intermediate parameters representing approximate impulse response signals and have no physical meaning.

5. The feedforward control method for a multi-rotor UAV based on a dynamic cascaded pulse neural network according to claim 1, characterized in that, In step S1, the desired pulse signal of the multi-rotor UAV is derived from the desired trajectory, and the actual pulse signal is derived from the actual flight trajectory of the multi-rotor UAV.

6. The feedforward control method for a multi-rotor UAV based on a dynamic cascaded pulse neural network according to claim 1, characterized in that, In step S2, the pulse signal difference is the difference between the desired pulse signal and the actual pulse signal.

7. The feedforward control method for a multi-rotor UAV based on a dynamic cascaded pulse neural network according to claim 1, characterized in that, In step S3, the initial preset parameters of the dynamic cascaded spiking neural network include the initial value of the number of hidden layer neurons, the initial value of the number of hidden layers, and the initial value of the weights of the hidden layer neurons, wherein the initial value of the number of hidden layers is set to 1.

Images