Brain-like processor-oriented seven-degree-of-freedom motion control method and system

Abstract

The invention discloses a seven-degree-of-freedom motion control method and system for a brain-like processor. The method comprises the steps that S1, the expected joint position, speed and acceleration of a mechanical arm are converted into a pulse sequence through an incremental encoder; s2, inputting the pulse sequence into a liquid state machine (LSM), and extracting spatial-temporal characteristics of a motion track by using a dynamic reservoir of the LSM; s3, jointly inputting the output characteristics of the liquid state machine LSM and the real torque value at the previous moment into a multi-layer perceptron MLP, and predicting the current joint torque; and S4, the predicted torque is adjusted through a feedback controller, and trajectory tracking is executed. According to the method, the parameter adjustment requirement is reduced, the method can be directly deployed in a brain-like processor, the power consumption is low, the calculation efficiency is high, meanwhile, the complex nonlinear relation in the high degree of freedom is learned through cyclic connection, and high-precision motion control is achieved. The method has wide application prospects in the fields of industrial robots, intelligent manufacturing, medical rehabilitation and the like.

Description

Technical Field

[0001] The present invention relates to the field of robot motion control technology, and in particular to a seven-degree-of-freedom motion control method and system for a brain-like processor. Background Art

[0002] For an N-DOF (degrees of freedom) manipulator, its inverse dynamics can be modeled as: ; in , , are joint position, velocity and acceleration respectively. is the inertia matrix. is the Coriolis term, which depends on and .at last, is the force or torque due to gravity acting on the system, is the joint torque vector.

[0003] Inverse dynamics modeling of a seven-degree-of-freedom robotic arm presents significant challenges, as precise parameters are difficult to obtain and the system is susceptible to external interference. In recent years, data-driven approaches have gained increasing popularity. Locally weighted projection regression (LWPR), while offering high real-time performance, requires manual adjustment of many parameters. Kernel methods such as Gaussian process regression (GPR) offer high accuracy but are computationally complex, making them difficult to implement in real time.

[0004] In contrast, the Liquid State Machine (LSM) extracts features from the dynamic evolution of the reservoir and only processes the output layer when performing predictions, reducing the need for parameter adjustment and achieving high computational efficiency. Compared to traditional artificial neural networks (ANNs), SNNs naturally possess the ability to process spatiotemporal information and can better utilize the temporal continuity between the robot's motion trajectory data. Summary of the Invention

[0005] To address the challenges of the existing technology, the present invention aims to provide a seven-degree-of-freedom motion control method for brain-inspired processors. This method reduces the need for parameter adjustment, can be directly deployed on brain-inspired processors, and exhibits low power consumption and high computational efficiency. Another object of the present invention is to provide a seven-degree-of-freedom motion control system for brain-inspired processors that implements the aforementioned method.

[0006] To achieve the above objectives, the present invention provides a seven-degree-of-freedom motion control method for a brain-inspired processor, comprising: S1. Convert the desired joint position, velocity, and acceleration of the robotic arm into a pulse sequence using an incremental encoder. S2. Input the pulse sequence into the liquid state machine LSM and use its dynamic reservoir to extract the spatiotemporal characteristics of the motion trajectory; S3. Input the output features of the liquid state machine (LSM) and the true torque value at the previous moment into the multi-layer perceptron (MLP) to predict the current joint torque. S4. Adjust the predicted torque through the feedback controller and perform trajectory tracking.

[0007] Furthermore, the liquid state machine (LSM) is used to project the input robot arm motion trajectory data into a high-dimensional space to extract and retain the features of the motion trajectory.

[0008] Furthermore, the liquid state machine LSM contains 100±20 pulse neurons, the ratio of excitatory to inhibitory neurons is 4:1, and the neurons are distributed in a three-dimensional cube topology. and The connection probability between them satisfies: ; in, is the Euclidean distance between neurons, is the attenuation coefficient, e is a mathematical constant, and C is a hyperparameter.

[0009] Furthermore, the incremental encoder determines the pulse emission according to the difference between the current moment and the previous moment, and the difference calculation method is: ; in, is the motion information of the robot arm at time t, is the motion information of the robotic arm at time t-1, >0, it stimulates excitatory neuronal pulses. When <0, inhibitory neuron pulses are stimulated.

[0010] Furthermore, the number of pulses emitted is: ; in, is an adjustable threshold.

[0011] Furthermore, the multi-layer perceptron (MLP) receives the features extracted by the liquid state machine (LSM), takes the extracted features and the actual torque value applied to the robotic arm in the previous time step as input, performs regression based on these inputs, and outputs the predicted torque that generates the robotic arm state trajectory.

[0012] Furthermore, the input dimension of the multi-layer perceptron MLP is the output feature dimension of the liquid state machine LSM + 7-dimensional torque history value, and the output layer is 7 linear neurons corresponding to the seven-degree-of-freedom torque prediction.

[0013] Furthermore, the use of the particle swarm optimization (PSO) algorithm to search for optimal structural parameters of the liquid state machine (LSM) includes: 1) Connection probability from the input layer to the liquid layer; 2) Four types of connection probabilities within the liquid layer: excitation → excitation / inhibition → excitation / excitation → inhibition / inhibition → inhibition; 3) Number of neurons.

[0014] Furthermore, in step S1, the dimension of the input trajectory is 21 dimensions, including 7-dimensional joint angles, 7-dimensional joint velocities, and 7-dimensional joint accelerations.

[0015] A seven-degree-of-freedom motion control system for a brain-inspired processor that implements the above-mentioned control method comprises: 1) Incremental encoding module: includes 21 input channels, each with 2 spiking neurons; 2) Liquid State Machine (LSM) processing module: Contains a spiking neuron reservoir arranged in a three-dimensional cube; 3) Multi-layer perceptron (MLP) torque prediction module: fully connected neural network; 4) PD feedback controller; 5) Torque history value cache unit; Each module is asynchronously event-driven through the neuromorphic computing core of the brain-like processor.

[0016] This invention reduces the need for parameter adjustment and can be directly deployed on brain-inspired processors. It offers low power consumption and high computational efficiency. Furthermore, it achieves high-precision motion control by learning complex nonlinear relationships within high degrees of freedom through recurrent connections. This invention has broad application prospects in industrial robotics, intelligent manufacturing, medical rehabilitation, and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of single spike neuron activity; Figure 2 This is a schematic diagram of the overall control architecture of the present invention; Figure 3 This is the structure diagram of the pulse neural network; Figure 4 Schematic diagram of incremental pulse encoding. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0020] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0021] The following is combined with Figure 1-Figure 4 The specific embodiments of the present invention are described in detail. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0022] SNNs are a type of artificial neural network inspired by biological neural systems. Unlike traditional artificial neural networks, SNNs transmit information through spike trains. Similar to biological neurons, when the voltage of a spiking neuron exceeds a threshold, the neuron fires a spike, resetting its membrane voltage to its resting potential. Figure 1 The activity of a single neuron is shown. Before a neuron fires a spike, its membrane potential accumulates over time, influenced not only by the current input but also by the retention of past information. This property makes spiking neural networks naturally capable of processing spatiotemporal data. Compared to simple feedforward SNN motion control networks that control low degrees of freedom (two or three degrees of freedom), the recurrent connections used in this method are more capable of learning the complex nonlinear relationships found in high-degree-of-freedom robotic arms.

[0023] like Figure 2-Figure 4 As shown, the present invention provides a seven-degree-of-freedom motion control method for a brain-inspired processor, comprising: S1. Convert the desired joint position, velocity, and acceleration of the robotic arm into a pulse sequence using an incremental encoder. S2. Input the pulse sequence into the liquid state machine LSM and use its dynamic reservoir to extract the spatiotemporal characteristics of the motion trajectory; S3. Input the output features of the liquid state machine (LSM) and the true torque value at the previous moment into the multi-layer perceptron (MLP) to predict the current joint torque. S4. Adjust the predicted torque through the feedback controller and perform trajectory tracking.

[0024] Liquid State Machine LSM: Liquid State Machine LSM is a recurrent neural network based on pulse neurons. Figure 3 As shown in Figure 1, the liquid state machine (LSM) mainly consists of an input layer, a liquid layer, and a readout layer. The connections between the input layer and the liquid layer are random, and the input layer is only connected to the excitatory neurons in the liquid layer. The liquid layer is the main part of the liquid state machine (LSM), and the connections between the excitatory neurons and inhibitory neurons are also random, including connections between themselves and other neurons. The input current of the excitatory neuron k in the liquid layer can be obtained by the following formula: ; in, is the input current of the input layer at the current time step, is the output of the liquid layer itself at the previous time step. n is the number of neurons in the input layer, and m is the number of neurons in the liquid layer. is the weight from the input layer to the liquid layer, is the input of the current time step, is the connection weight between the liquid layer and itself, It is the final output of the liquid layer at the previous time step.

[0025] Due to the recurrent connections and nonlinear activations between spiking neurons, the liquid state machine (LSM) has the ability to extract features, dynamically memorize, and utilize past information, making it better able to process time series data.

[0026] The overall framework of the control system is as follows Figure 2 As shown in Figure 2, the trajectory generator generates the desired state of the robotic arm at the next moment (joint angles, velocities, and accelerations), converts this into pulse input via the incremental encoder, and feeds it into the neural network. The network's readout layer outputs the predicted torque.

[0027] The trajectory generator in this invention is implemented using the Ruckig library in ROS. The Ruckig trajectory generator uses a high-order polynomial interpolation algorithm to generate smooth trajectories that satisfy motion constraints in real time. Its core principle is to precisely connect any initial state to the target state using a combination of seven stages with varying jerk. Based on analytical calculations rather than numerical iteration, the present invention can complete trajectory planning in microseconds and achieve continuous state transitions through a state-space model, providing the complete desired state, including position, velocity, and acceleration, on demand according to the control cycle.

[0028] The feedback controller then adjusts the predicted torque based on the current and desired states to ensure accurate trajectory tracking of the robot arm. Finally, the actual torque applied to the robot arm at the previous moment is directly encoded and provided to the network to help predict the torque at subsequent moments.

[0029] The feedback controller is implemented using a PD controller. Proportional gain P is a key parameter in the PD controller, used to adjust the proportional relationship between the system output and the error. Generally speaking, increasing the P value speeds up the system's response, but setting it too high can cause system oscillation or even instability. The differential gain D is used to compensate for system damping and suppress oscillations, thereby reducing overshoot and stabilizing oscillations. Generally, increasing the differential gain D improves system stability, but setting it too high can cause the system to respond too slowly or even overshoot. By properly adjusting the proportional gain P and differential gain D, the PD controller can achieve optimal control in the robotic arm system, achieving precise position and posture control.

[0030] In the PID control of a robotic arm, to optimize the settings of the proportional gain (P) and the differential gain (D), a step-by-step strategy of adjusting P first and then D is adopted. Simulation and experimental verification are combined to evaluate tracking accuracy and steady-state performance. The recommended initial range for the proportional gain P is [1, 10]. Larger P values ​​can improve system response speed and reduce steady-state error, but excessively large P values ​​can lead to overshoot or persistent oscillation. Therefore, during the simulation phase, the P value should be gradually increased, observing the robotic arm's step response, and selecting a critical value that allows the system to quickly stabilize without significant overshoot. Subsequently, the P value is further fine-tuned during experimental verification to ensure that tracking accuracy meets requirements during actual motion. The differential gain D functions to suppress overshoot and oscillation. Its typical range is [0.005, 1]. Smaller D values ​​can avoid amplifying high-frequency noise, while appropriately increasing D can improve dynamic performance. When adjusting D, it should be gradually increased from a fixed P value, observing the damping effect of the system response to ensure smooth convergence during rapid robotic arm motion. Ultimately, the optimal combination of P and D should enable the robotic arm to exhibit good tracking accuracy, fast response, and stable and oscillation-free characteristics in both simulation and experiment.

[0031] Network structure: Spiking neural networks for inverse dynamics solutions such as Figure 3 As shown, the network consists of two parts: Liquid State Machine (LSM): A liquid state machine is used to project the input robot arm motion trajectory data into a high-dimensional space, extracting and retaining the features of the motion trajectory.

[0032] The Multi-Layer Perceptron (MLP) receives the features extracted by the LSM and uses the extracted features and the actual torque applied to the robot arm at the previous time step as input. It performs regression based on these inputs and outputs the predicted torque that generates the robot arm's state trajectory.

[0033] The connection weights between input neurons and excitatory neurons in the liquid state machine (LSM), as well as the connection weights between neurons in the LSM, are constant after initialization and do not participate in network training. This reduces training complexity and accelerates model convergence. The connections between the LSM and the multilayer perceptron (MLP) are feedforward and updated according to the backpropagation rule.

[0034] Incremental encoder: The encoding method of the input trajectory adopts incremental encoding, such as Figure 4 The incremental encoder determines the pulse emission based on the difference between the current moment and the previous moment.

[0035] ; in, is the input at time t (the motion information of the robotic arm at time t), It is the input at time t-1 (the motion information of the robotic arm at time t-1).

[0036] For a 7-DOF manipulator, the input trajectory has 21 dimensions (including 7-dimensional joint angles, 7-dimensional joint velocities, and 7-dimensional joint accelerations). Each dimension of input corresponds to two spiking neurons. When the difference is positive, the excitatory neuron pulses; otherwise, the inhibitory neuron pulses. The network has hyperparameters ( is an adjustable threshold) to control the number of pulses emitted : ; The incremental encoder captures the dynamic changes of the input robot motion trajectory data, making it easier for the network to capture complex changes and trends in the robot motion trajectory.

[0037] In a spiking neural network, a timestep refers to the unit of time during which the network performs an iterative update. This can be understood as the neural network receiving and processing inputs and generating outputs at discrete points in time. Spiking neural networks discretize inputs and neuron states, representing neuronal activity as discrete pulse signals to simulate the information transfer process between neurons. In this invention, a single input is processed in multiple timesteps. This means the network receives the current input at multiple discrete points in time.

[0038] Liquid State Machine LSM structure and parameter search:

[0039] The Liquid State Machine (LSM) uses approximately 100 spiking neurons. We randomly select some of them as inhibitory neurons and the rest as excitatory neurons. The LSM neurons form a cube structure. We define neurons according to the following formula: and Distance between: ; in, is the x-coordinate of the ith neuron in the liquid layer, is the x-coordinate of the j-th neuron in the liquid layer, is the y coordinate of the i-th neuron in the liquid layer, is the y coordinate of the jth neuron in the liquid layer, is the z coordinate of the ith neuron in the liquid layer, is the z coordinate of the jth neuron in the liquid layer.

[0040] This distance is then used to calculate the neuron according to the following formula and The probability of connection between .

[0041] ; in, is the Euclidean distance between neurons, is the attenuation coefficient, e is a mathematical constant, and C is a hyperparameter. As the formula shows, the smaller the distance between neurons, the higher the probability of connection between them. This aspect also improves the biological plausibility of the network. The parameter search range is shown in Table 1: Table 1: Parameter search range ; Where n is the number of neurons in the liquid layer, is the connection probability between the input layer neurons and the liquid layer neurons, is the connection probability between excitatory neurons in the liquid layer and excitatory neurons. is the connection probability between excitatory neurons in the liquid layer and inhibitory neurons. is the connection probability between inhibitory neurons and excitatory neurons in the liquid layer. is the connection probability between inhibitory neurons in the liquid layer and inhibitory neurons.

[0042] Because different liquid state machine (LSM) structures have different feature extraction capabilities, we use a particle swarm optimization (PSO) algorithm for the search. Key search parameters include the number of neurons, the connection probability between input layer neurons and the liquid layer, and the four connection probabilities between excitatory and inhibitory neurons in the liquid layer.

[0043] The particle swarm optimization (PSO) algorithm utilizes the information sharing of individuals in the group to make the movement of the entire group evolve from disorder to order in the problem-solving space, thereby obtaining the optimal solution.

[0044] PSO is initialized as a swarm of random particles (random solutions). It then iterates to find the optimal solution. In each iteration, the particles update themselves by tracking two "extreme values" (pbest and gbest). After finding these two optimal values, the particles update their velocity and position using the following formula.

[0045] ;

[0046] Where i=1,2,...,N, N is the total number of particles in this group; is the velocity of the particle, is a random number between (0,1), is the current position of the particle, and is the learning factor. (Individual Optimum): represents the historical optimal position found by the i-th particle itself during the iteration process (that is, the solution with the best fitness of the particle in all previous iterations). (Global Optimum): represents the optimal position found by all particles in the entire particle swarm during the iteration process (i.e., the best solution found by the entire swarm so far).

[0047] The present invention uses a particle swarm optimization (PSO) algorithm to automatically optimize hyperparameters. The optimization process follows the following flow: First, the PSO algorithm initializes a set of hyperparameter combinations and calculates the error index of the torque prediction model based on the current parameter configuration. Subsequently, the PSO algorithm dynamically adjusts the particle position and velocity according to the error evaluation results to generate a new generation of hyperparameter combinations. Through iterative optimization, the algorithm continuously reduces the prediction error and eventually converges to a hyperparameter configuration that achieves the optimal torque prediction accuracy.

[0048] A multilayer perceptron (MLP) is a feedforward neural network consisting of an input layer, one or more hidden layers (in this article, we use a single hidden layer), and an output layer. An MLP is a fully connected neural network, where neurons in each layer are connected to every neuron in the next layer. The following is the calculation formula for a perceptron: ; Where X is the input of the layer, W is the weight of the layer, B is the bias of the layer, T is the activation function of the layer (Tanh function is used in this invention), and y is the output of the layer.

[0049] The present invention also provides a seven-degree-of-freedom motion control system for a brain-inspired processor that implements the above-mentioned control method, comprising: 1) Incremental encoding module: includes 21 input channels, each with 2 spiking neurons; 2) Liquid State Machine (LSM) processing module: Contains a spiking neuron reservoir arranged in a three-dimensional cube; 3) Multi-layer perceptron (MLP) torque prediction module: fully connected neural network; 4) PD feedback controller; 5) Torque history value cache unit; Each module is asynchronously event-driven through the neuromorphic computing core of the brain-like processor.

[0050] Example: The number of LSM neurons is 174, with excitability accounting for 80%, and 42 input neurons. The PSO search hyperparameters are configured as follows:

[0051] The MLP (Multi-Layer Perceptron) network adopts a classic three-layer fully connected architecture. The input layer receives a 181-dimensional composite feature vector consisting of features extracted from a 174-dimensional LSM and 7-dimensional actual joint torque values. The hidden layer has 128 neurons, using the Tanh activation function for nonlinear transformation. The output layer has 7 neurons corresponding to the torque predictions for each joint of the robotic arm. The model is trained using a batch strategy with step_size=50 and an Adam optimizer (learning rate lr=0.001, weight decay coefficient weight_decay=0.01) for 300 epochs. A step learning rate scheduler (StepLR) is also introduced, which decays the learning rate by multiplying it by 0.1 (gamma=0.1) after every 50 epochs (step_size=50).

[0052] The technical problem of the present invention is that the physical modeling of the inverse dynamics of the robotic arm is limited by the difficulty in obtaining parameters and external interference. The previous data-driven methods each have their own advantages, but they have limitations such as high demand for manual parameter adjustment, high computational complexity, high resource consumption, and low control freedom. The main technical solution of the present invention is to propose a motion control method and system for brain-like processors. The motion control method for brain-like processors of the present invention combines a liquid state machine with a multi-layer perceptron to predict the torque required for the movement of the robotic arm. This method reduces the need for parameter adjustment and can be directly deployed on brain-like processors. It has low power consumption and high computational efficiency. At the same time, it learns complex nonlinear relationships in high degrees of freedom through cyclic connections to achieve high-precision motion control. The present invention has broad application prospects in the fields of industrial robots, intelligent manufacturing, medical rehabilitation, etc.

[0053] Any process or method described in the flowchart of the present invention or in other ways herein can be understood as representing a module, segment or portion of code including one or more executable instructions for implementing specific logical functions or process steps, which can be implemented in any computer-readable medium for use by an instruction execution system, device or apparatus. Computer-readable media can be any medium that stores, communicates, propagates or transmits a program for use by an execution system, device or apparatus, including read-only memory, magnetic disk or optical disk, etc.

[0054] Throughout this specification, reference to terms such as "embodiment" and "example" indicates that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, those skilled in the art may combine or integrate different embodiments or examples described in this specification, as well as features therein, without creating any inconsistency.

[0055] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed above with reference to the preferred embodiment, it is not intended to limit the present invention. Therefore, any simple modifications, equivalent variations, and modifications to the above embodiment that do not depart from the technical solution of the present invention and are based on the technical essence of the present invention shall fall within the scope of protection of the technical solution of the present invention.

Claims

1. A seven-degree-of-freedom motion control method for brain-inspired processors, characterized by: The method comprises: S1. Convert the desired joint position, velocity, and acceleration of the robotic arm into a pulse sequence using an incremental encoder. S2. Input the pulse sequence into the liquid state machine LSM and use its dynamic reservoir to extract the spatiotemporal characteristics of the motion trajectory; S3. Input the output features of the liquid state machine (LSM) and the true torque value at the previous moment into the multi-layer perceptron (MLP) to predict the current joint torque. S4. Adjust the predicted torque through the feedback controller and perform trajectory tracking.

2. The seven-degree-of-freedom motion control method for a brain-inspired processor according to claim 1, characterized in that: The liquid state machine (LSM) is used to project the input robot arm motion trajectory data into a high-dimensional space to extract and retain the features of the motion trajectory.

3. The seven-degree-of-freedom motion control method for a brain-inspired processor according to claim 2, characterized in that: The liquid state machine LSM contains 100±20 spiking neurons, the ratio of excitatory to inhibitory neurons is 4:1, and the neurons are distributed in a three-dimensional cube topology. and The connection probability between them satisfies: ； in, is the Euclidean distance between neurons, is the attenuation coefficient, e is a mathematical constant, and C is a hyperparameter.

4. The seven-degree-of-freedom motion control method for a brain-inspired processor according to claim 1, characterized in that: The incremental encoder determines the pulse emission according to the difference between the current moment and the previous moment. The difference calculation method is: ； in, is the motion information of the robot arm at time t, is the motion information of the robotic arm at time t-1, >0, it stimulates excitatory neuronal pulses. When <0, inhibitory neuron pulses are stimulated.

5. The seven-degree-of-freedom motion control method for a brain-inspired processor according to claim 4, characterized in that: The number of pulses emitted is: ； in, is the threshold.

6. The seven-degree-of-freedom motion control method for a brain-inspired processor according to claim 1, characterized in that: The multi-layer perceptron (MLP) receives the features extracted by the liquid state machine (LSM), takes the extracted features and the actual torque value applied to the robotic arm in the previous time step as input, performs regression based on these inputs, and outputs the predicted torque that generates the robotic arm state trajectory.

7. The seven-degree-of-freedom motion control method for a brain-inspired processor according to claim 6, characterized in that: The input dimension of the multi-layer perceptron MLP is the output feature dimension of the liquid state machine LSM + the 7-dimensional torque history value, and the output layer is 7 linear neurons corresponding to the seven-degree-of-freedom torque prediction.

8. The seven-degree-of-freedom motion control method for a brain-inspired processor according to claim 1, characterized in that: The particle swarm optimization (PSO) algorithm is used to search for the optimal structural parameters of the liquid state machine (LSM), including: 1) Connection probability from the input layer to the liquid layer; 2) Four types of connection probabilities within the liquid layer: excitation → excitation / inhibition → excitation / excitation → inhibition / inhibition → inhibition; 3) Number of neurons.

9. The seven-degree-of-freedom motion control method for a brain-inspired processor according to claim 1, characterized in that: In step S1 , the dimension of the input trajectory is 21 dimensions, including 7-dimensional joint angles, 7-dimensional joint velocities, and 7-dimensional joint accelerations.

10. A seven-degree-of-freedom motion control system for a brain-inspired processor, used to implement the seven-degree-of-freedom motion control method for a brain-inspired processor according to any one of claims 1 to 9, characterized in that: The system comprises: 1) Incremental encoding module: includes 21 input channels, each with 2 spiking neurons; 2) Liquid State Machine (LSM) processing module: Contains a spiking neuron reservoir arranged in a three-dimensional cube; 3) Multi-layer perceptron (MLP) torque prediction module: fully connected neural network; 4) PD feedback controller; 5) Torque history value cache unit; Each module is asynchronously event-driven through the neuromorphic computing core of the brain-like processor.

Images