Experimental platform for constructing electrical stimulation-behavior mapping relation

Abstract

The invention relates to the field of biological robots, neural engineering and multi-sensor data fusion, and discloses an experimental platform for constructing an electrical stimulation-behavior mapping relation, which comprises a behavior identification and data marking module used for identifying and judging the behavior of a tortoise robot according to a state vector sequence and marking the behavior of the tortoise robot according to the state vector sequence. Obtaining a behavior label under the current electrical stimulation parameter; and associating and recording the state vector sequence, the behavior tag, the current electrical stimulation parameter and the video timestamp to generate a structured database. The problems of low manual observation efficiency, high subjectivity and difficulty in data association in traditional research are solved, and an efficient, objective, high-throughput and high-fidelity data acquisition and analysis platform is provided for systematically researching a complex mapping relation between nerve electrical stimulation and animal behaviors.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of bio-robotics, neuroengineering and multi-sensor data fusion, and particularly relates to an experimental platform for constructing an electrical stimulation-behavior mapping relationship. BACKGROUND

[0002] The research of "bio-robotics" using electrical stimulation to regulate animal behavior is a frontier direction of neuroengineering. In the research, a core task is to understand the precise correspondence between different electrical stimulation parameters (such as location, frequency, amplitude) and the specific behavior (such as forward, left turn, right turn) produced by the animal, i.e. to establish a "stimulation-response" model.

[0003] Currently, the construction of this model mainly relies on manual methods. Researchers need to manually set the electrical stimulation parameters, then observe and record the animal's behavioral responses by naked eye for a long time, and finally review the video for manual labeling. This method has the following serious defects: Low efficiency and low throughput: manual operation and observation are time-consuming and labor-intensive, making it difficult to complete systematic traversal testing of a large number of multi-dimensional stimulation parameter combinations in a short time, i.e. the experimental throughput is extremely low.

[0004] Subjectivity and irreproducibility: the judgment of behavior depends on the subjective experience of the observer, and the standards between different researchers are difficult to unify, resulting in poor reproducibility of experimental results.

[0005] Poor data quality and relevance: the time stamp recorded manually and the video frame, the triggering time of the stimulation signal are difficult to synchronize accurately, resulting in delays and errors in the data, affecting the training accuracy of the subsequent model.

[0006] In order to overcome these problems, some researches have tried to introduce automation technology. However, single sensing scheme has its own physical limitations that cannot be overcome: Limitations of pure visual scheme: it relies on external cameras and marker points, which are easily disturbed by environmental factors such as changes in light and marker points being blocked by limbs. At the same time, the frame rate of the camera is limited, making it difficult to capture the instantaneous and high-frequency response of the animal's muscles under electrical stimulation.

[0007] Limitations of pure inertial navigation scheme: although the inertial measurement unit can provide high-frequency dynamic data greater than 100Hz, its output needs to be integrated to obtain attitude and position, which will inevitably produce drift errors that accumulate over time, resulting in distorted long-term monitoring results.

[0008] Therefore, there is an urgent need in the art for an automated system that can integrate multi-source sensor information, achieve high-precision, high-robustness, and high-frequency dynamic capture of animal behavior, and automatically and accurately associate the behavior recognition results with the stimulation parameters, to realize high-throughput data acquisition and labeling. SUMMARY

[0009] The main purpose of the present application is to solve the technical problems of low experimental throughput, strong subjectivity of manual marking, poor data quality and difficulty in accurate correlation in the prior art. An experimental platform for constructing an electrical stimulation-behavior mapping relationship comprises: An electrical stimulator for outputting different electrical stimulation parameters; An electrical stimulation parameter traversal module for automatically controlling the electrical stimulator to output different electrical stimulation parameters according to a preset scheme; A turtle robot receiving electrical stimulation through an electrical stimulation electrode of the electrical stimulator, the turtle robot being provided with visual marker points on the body; A visual sensing unit for capturing two-dimensional absolute coordinates of the visual marker points and providing a video timestamp; An inertial sensing unit fixed to the turtle robot for collecting angular velocity and linear acceleration data thereof; A posture solving module for performing optimal state estimation on the motion posture of the turtle robot according to the data of the visual sensing unit and the inertial sensing unit, and outputting a state vector sequence; A behavior recognition and data marking module for recognizing and determining the behavior of the turtle robot according to the state vector sequence to obtain a behavior label under the current electrical stimulation parameter; correlating and recording the state vector sequence, the behavior label, the current electrical stimulation parameter and the video timestamp to generate a structured database.

[0010] The definition of the visual marker points is as follows: From front to back, visual marker points a, b, c and d are arranged on the central axis of the turtle robot body; the visual marker point a is located at the top end of the turtle's nose, the visual marker point c is located at the center of the shell, the visual marker point d is located at the rear end of the shell, and the visual marker point d is the reference point of the hip; Visual marker points b1, b2, b3 and b4 are respectively located at the elbow joints and finger joints of the left and right forelimbs; Points d1 and d2 are respectively located at the knee joints of the left and right hind limbs.

[0011] The posture solving module adopts an EKF algorithm to perform optimal estimation on the data of the visual sensing unit and the inertial sensing unit, and finally corrects the state of the visual marker points recorded by the visual sensing unit data to perform optimal state estimation on the motion posture of the turtle robot and output a state vector sequence.

[0012] The EKF algorithm is as follows: Prediction stage: using the data of the inertial sensing unit and the optimal state at the last time, the state at the current time is predicted through a nonlinear state transition function; Update phase: When data is received from the visual sensing unit, the predicted value is corrected by Kalman gain to obtain the optimal state estimate at the current moment; Correction phase: The visual sensing unit records the state of the visual markers, calculates the relative velocity of each limb end relative to its joint, reflects the intensity of the limb end swing through the magnitude of the vector, and then estimates the optimal state of the turtle robot's motion posture and outputs a state vector sequence.

[0013] The forward movement determination rule is: Current velocity norm Is it greater than the threshold? and angular velocity of heading Is it less than the threshold? Does diagonal limb coordination exist?

[0014] The rule for determining a left turn is: angular velocity of the heading. Is it greater than the threshold? Whether the intensity of movement on the right side of the body is significantly greater than that on the left side.

[0015] The rule for determining a stationary state is: both the velocity norm and the heading angular velocity are less than their respective thresholds.

[0016] Define the velocity of the left forelimb relative to the body's center of mass c as: The velocity of the right hind limb relative to the center of mass c is .when , When the projection component in the direction the turtle is moving is positively correlated with the height, it can be confirmed as moving forward.

[0017] By comparing the relative velocity norms of the extremities of both limbs with respect to their joints, that is, when > And the difference exceeds the threshold If the intensity of movement is significantly greater on the right side than on the left side, it indicates whether the intensity of movement on the right side is significantly greater than that on the left side.

[0018] This invention also relates to a method for constructing an electrical stimulation-behavior mapping database, the method comprising the following steps: S1: Visual markers are placed on the turtle robot and inertial sensing units are fixed, while a preset electrical stimulation parameter scheme is loaded on the computer terminal. S2: The electrical stimulator is instructed to begin applying stimulation. At the same time, the attitude calculation module estimates the optimal state of the turtle robot's motion posture based on the data from the visual sensing unit and the inertial sensing unit, and outputs a state vector sequence. S3: The behavior recognition and data labeling module continuously analyzes the optimal state estimation vector, recognizes the behavior according to the preset behavior judgment rule, and instantly generates a data record containing a timestamp, a current stimulation parameter, a state vector sequence and a behavior label; S4: After the preset duration of a set of stimulation parameters ends, the next set of parameters in the parameter list is automatically switched to and the monitoring and labeling steps are repeated until all parameters are traversed, and finally a complete electric stimulation-behavior mapping database is saved.

[0019] The present application has the following beneficial effects: The present application solves the problems of low efficiency, strong subjectivity and difficulty in data correlation in traditional manual observation, and provides an efficient, objective and high-throughput high-fidelity data acquisition and analysis platform for systematic research on the complex mapping relationship between neural electric stimulation and animal behavior.

[0020] The present application realizes high-throughput automated experiments, integrates electric stimulation traversal, behavior monitoring, recognition and data labeling in an automated process, greatly improves experimental efficiency, and makes it possible to systematically screen a large number of parameter combinations.

[0021] The present application guarantees the objectivity and high fidelity of behavior data, defines and recognizes behavior based on a multi-modal fusion algorithm and a quantitative kinematic operator, and eliminates human subjective judgment errors. The fused data have high frequency and high precision, and provide a high-quality data source for establishing an accurate model.

[0022] The present application realizes strong correlation and structured archiving of data, automatically synchronizes stimulation parameters, behavior labels and spatiotemporal data, generates a structured database and a labeled video file, and greatly facilitates subsequent data retrieval, visual analysis and machine learning model training. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The present application provides an experimental platform for constructing an electric stimulation-behavior mapping relationship; Figure 2 The present application provides a turtle robot with visual marker points; Figure 3 The present application provides a flowchart of an EKF algorithm; Figure 4 The present application provides a schematic diagram of a behavior recognition and data labeling module.

[0024] The following reference signs are used: Electric stimulator 3; Turtle robot 2; Visual sensing unit 1; Inertial sensing unit 5; The computer processing terminal 4. DETAILED DESCRIPTION

[0025] The terms "first", "second", "third", "fourth" and the like in the description and in the claims of the present application, and above figures (if any) are used to distinguish similar objects, and are not necessarily used to describe a particular sequential or chronological order. It should be understood that the data thus used can be interchanged, where appropriate, so that the embodiments described herein can be carried out in sequences other than those illustrated or described herein. Furthermore, the terms "comprise" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a list of steps or units is not necessarily limited to those steps or units that are clearly listed, but can include other steps or units that are not clearly listed or inherent to such processes, methods, products, or apparatuses.

[0026] The experiment involved in the present patent has obtained ethical approval, and the approval number is: AP#25-121-CWC, and the approval date is 2025-08-05.

[0027] For ease of understanding, the specific process of the embodiments of the present application is described below. Please refer to Figures 1-4 The first embodiment of the experiment platform for constructing the electrical stimulation-behavior mapping relationship in the embodiments of the present application comprises: An electrical stimulation instrument for outputting different electrical stimulation parameters; An electrical stimulation parameter traversal module for automatically controlling the electrical stimulation instrument to output different electrical stimulation parameters according to a preset scheme; A turtle robot receiving electrical stimulation through the electrical stimulation electrode of the electrical stimulation instrument, wherein the turtle robot is provided with a visual marker point on the body; A visual sensing unit for capturing the two-dimensional absolute coordinates of the visual marker point and providing a video timestamp; An inertial sensing unit fixed to the turtle robot for collecting angular velocity and linear acceleration data thereof; A posture solving module for performing optimal state estimation on the motion posture of the turtle robot according to the data of the visual sensing unit and the inertial sensing unit, and outputting a state vector sequence; A behavior recognition and data labeling module for recognizing and determining the behavior of the turtle robot according to the state vector sequence to obtain a behavior label under the current electrical stimulation parameter; associating and recording the state vector sequence, the behavior label, the current electrical stimulation parameter, and the video timestamp to generate a structured database.

[0028] Specifically, the modules of the present application comprise: The electrical stimulation parameter traversal module: A programmable multi-channel electrical stimulator. Its control program can automatically and cyclically change the output stimulation parameters (such as channel, frequency, pulse width, amplitude) according to a pre-set scheme, and provide the current applied parameters to the data labeling module in real time.

[0029] Visual sensing unit: top camera, used to capture the two-dimensional absolute coordinates of the visual markers fixed on the turtle's body.

[0030] Inertial sensing unit: a miniature IMU fixed near the center of mass of the turtle's carapace, transmitting high-frequency angular velocity and linear acceleration data wirelessly.

[0031] Attitude solving module: This module is the core algorithm unit of the system, running on a computer terminal. Its core is a fusion algorithm based on extended Kalman filter (EKF), used to achieve optimal estimation of the turtle's attitude.

[0032] 3.1 Definition of turtle marker points: Referring to Figure 2 , to accurately capture the turtle's motion, the selection and use of visual markers are based on kinematic analysis principles: body axis and reference point: point c is selected as the origin of the turtle's body coordinate system, and this point is also the preferred placement position of the IMU to minimize the interference caused by limb movement. Point b is selected as the reference point in front of the body. Vector defines the main axis of the turtle's body coordinate system. Head movement point: point a is located at the top of the turtle's snout. The change in the direction of vector can accurately quantify the direction of the head. Forelimb movement points: points b1, b2, b3, and b4 are located at the elbow joints and finger joints of the left and right forelimbs, respectively. Hindlimb movement points: point d is located near the body axis as a reference point for the hips. Points d1 and d2 are located at the knee joints of the left and right hindlimbs, respectively.

[0033] 3.2 Turtle motion attitude state space: First, define a state vector (State Vector) that describes the turtle's motion attitude in a two-dimensional plane: where is the position, is the velocity, is the heading angle, is the heading angular velocity.

[0034] 3.3 Sensor data fusion: The EKF algorithm optimally estimates the data from both sensors through a "prediction-update" iterative loop, and finally obtains the final decision operator after rectifying the camera data.

[0035] Prediction phase (Process): Using the high-frequency data provided by the IMU and the optimal state at the previous time , the state at the current time is predicted through a nonlinear state transition function .

[0036] Update phase (Measurement): When receiving low-frequency, high-precision visual data from the camera , the predicted value is corrected through the Kalman gain to obtain the optimal state estimation (posterior estimation) at the current time . Behavior feedback and adaptive adjustment phase: The core innovation of the present application lies in that the EKF algorithm is not static, but dynamically coupled with the behavior recognition result. After each 'update phase' is completed, the automatic behavior recognition module outputs the current behavior label, such as'still'. The label is fed back to the attitude solving module to adjust the process noise covariance matrix Q before the next 'prediction phase' starts.

[0037] If the behavior label is'still', set the diagonal elements related to the change of speed and angular velocity in the Q matrix to a value Q_static. If the behavior label is 'forward' or 'turning', set these elements to a value Q_dynamic. Among them, Q_static is less than Q_dynamic.

[0038] This adjustment allows the filter to intelligently change the degree of trust in the motion model when the turtle behavior mode changes, thereby achieving better state estimation. Correction phase (Correction): The low-frequency, high-precision visual data of the camera also records the state of the turtle marker points; By calculating the relative velocity of each limb end (such as b1, b4) relative to its joint (b2, b3), for example , the length of this vector reflects the degree of limb end swing.

[0039] Behavior recognition and data labeling module: This module receives the high-precision state vector sequence and marker point vector sequence from the attitude solving module and performs the following operations: 4.1 Behavior recognition: Based on the time series of state vectors and marker point vectors of time series, by preset behavior characteristic operators. For example: Forward determination: the speed norm is greater than a threshold value , and the heading angular velocity is less than a threshold value . At the same time, the system calculates the relative motion relationship between the diagonal limbs in real time. The standard forward movement is the coordinated movement of the diagonal limbs. For example, the left front limb and the right rear limb should swing forward in coordination, while the right front limb and the left rear limb swing backward in coordination. This can be quantified by calculating the correlation of the diagonal limb velocity vectors. Define the velocity of the left front limb relative to the body center of mass c as , and the velocity of the right rear limb relative to the center of mass c as . When the projection components of the two velocity vectors in the forward direction of the turtle are highly positively correlated, it can be confirmed as forward.

[0040] Left turn determination: the heading angular velocity is greater than a threshold value . At the same time, the system calculates and compares the motion intensity of the left and right limbs in real time. When turning left, the motion intensity of the right limb (the driving side) is significantly greater than that of the left limb (the turning axis side). This can be quantified by comparing the relative speed norms of the two limb ends relative to their joints. That is, when > and the difference exceeds a threshold value , it can be confirmed as left turn.

[0041] Still determination: both the speed norm and the heading angular velocity are less than their respective threshold values.

[0042] 4.2 Data association and labeling: Once the behavior is identified, the system immediately associates the behavior label (such as "forward"), the current stimulation parameters from the electrical stimulation module, the video timestamp, and the complete state vector and the marker point vector , and records them in a structured form.

[0043] The specific embodiments of the present application also include a method for constructing an electrical stimulation-behavior mapping database, which comprises the following steps: S1: placing visual marker points on the turtle robot and fixing the inertial sensing unit, while loading the preset electrical stimulation parameter scheme on the computer terminal; S2: instructing the electrical stimulator to start applying stimulation, while the attitude solving module performs optimal state estimation of the motion attitude of the turtle robot according to the data of the visual sensing unit and the inertial sensing unit, and outputs a sequence of state vectors; S3: The behavior recognition and data labeling module continuously analyzes the optimal state estimation vector, identifies the behavior according to the preset behavior judgment rule, and generates a data record in real time, which contains the timestamp, the current stimulation parameter, the state vector sequence, and the behavior label; S4: After the preset duration of a set of stimulation parameters ends, the system automatically switches to the next set of parameters in the parameter list and repeats the monitoring and labeling steps until all parameters are completed, and finally saves a complete electric stimulation-behavior mapping database.

[0044] Specifically, the system workflow is as follows: S1 Preparation: Visual marker points are placed on the turtle's body according to the positions shown in Figure 1 , and the IMU is fixed. A preset electric stimulation parameter traversal scheme list is loaded on the computer terminal. At the same time, the initial state vector and the covariance of the EKF are loaded.

[0045] S2 Start automatic traversal and monitoring cycle: The electric stimulation parameter traversal module takes the first set of parameters from the scheme list and instructs the electric stimulator to start stimulation. Referring to the flowchart in Figure 3 , the attitude solving module receives IMU data at a high frequency and continuously predicts the turtle's motion state according to the EKF prediction stage formula. At the same time, referring to the flowchart in Figure 3 , the attitude solving module receives visual data captured by the camera at a lower frequency, i.e., the coordinates of all marker points. Once the data is received, the EKF update stage is immediately executed to obtain the optimal state estimation . Similarly, the camera also performs an additional correction stage, i.e., by calculating the marker point velocity from the video to obtain .

[0046] S3 Real-time behavior recognition and data labeling: Referring to the flowchart in Figure 4 , the data labeling module continuously analyzes the optimal state vector output by the attitude solving module and calculates the relative positions and velocities between all marker points and other kinematic parameters. According to the aforementioned enhanced judgment method, the behavior is determined, including "forward", "left turn", and "still". The system immediately generates a data record containing at least: {timestamp, current stimulation parameter, state vector, kinematic feature parameter, behavior label}. This record is appended to a structured log file.

[0047] S4 Cycle and end: After the preset duration of a set of stimulation parameters ends, the system checks whether the parameter list has been completely executed. If not, it returns to S2. If all parameters have been completed, the experiment ends, the complete log file is saved, and a complete "electric stimulation-behavior" mapping database is formed.

[0048] In summary, the present application realizes high-throughput automated experiments, integrates electrical stimulation traversal, behavior monitoring, recognition and data labeling into an automated process, greatly improves experimental efficiency, and makes it possible to systematically screen a large number of parameter combinations. The objectivity and high fidelity of the behavior data are ensured, and the behavior is defined and recognized based on a multi-modal fusion algorithm and a quantitative kinematic operator, eliminating human subjective judgment errors. The fused data have high frequency and high precision, providing a high-quality data source for establishing an accurate model. Strong correlation and structured archiving of data are realized, and the system automatically synchronizes the stimulation parameters, behavior labels and spatiotemporal data accurately to generate a structured database and a labeled video file, greatly facilitating subsequent data retrieval, visual analysis and machine learning model training. At the same time, a closed-loop optimization of perception and recognition is realized, improving the intelligence level and robustness of the system. The present application forms an intelligent closed loop by feeding back the high-level behavior recognition results to the bottom-level pose estimation algorithm. This enables the system to adaptively cope with the dynamic changes of animal behavior, and still maintains high-precision tracking performance at the moment of motion mode switching (such as from stillness to vigorous movement), greatly enhancing the robustness of the system.

[0049] The above examples are only used to illustrate the technical solutions of the present application, and not to limit it; although the present application has been described in detail with reference to the foregoing examples, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacements for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An experimental platform for constructing an electrical stimulation-behavior mapping relationship, characterized in that, The method comprises the following steps: An electric stimulator is used to output different electric stimulation parameters; An electric stimulation parameter traversal module is used to automatically control the electric stimulator to output different electric stimulation parameters according to a preset scheme; A turtle robot receives electric stimulation through an electric stimulation electrode of the electric stimulator, and visual marker points are arranged on the body of the turtle robot; A visual sensing unit is used to capture two-dimensional absolute coordinates of the visual marker points and provide video time stamps; An inertial sensing unit is fixed to the turtle robot and is used to collect angular velocity and linear acceleration data of the turtle robot; A posture solving module is used to perform optimal state estimation on the motion posture of the turtle robot according to data of the visual sensing unit and the inertial sensing unit, and output a state vector sequence; An action recognition and data labeling module is used to recognize and determine the action of the turtle robot according to the state vector sequence, and obtain an action label under the current electric stimulation parameter; The state vector sequence, the action label, the current electric stimulation parameter and the video time stamp are associated and recorded to generate a structured database.

2. The experimental platform for constructing an electrical stimulation-behavior mapping relationship according to claim 1, characterized in that, The definition of the visual marker points is as follows: From front to back, visual marker points a, b, c and d are arranged on the central axis of the body of the turtle robot; the visual marker point a is located at the top end of the nose of the turtle, the visual marker point c is located at the center of the shell, the visual marker point d is located at the rear end of the shell, and the visual marker point d is the reference point of the hip; The elbow joints and finger joints of the left and right forelimbs are visual marker points b1, b2, b3 and b4; The knee joints of the left and right hindlimbs are visual marker points d1 and d2.

3. The experimental platform for constructing an electrical stimulation-behavior mapping relationship according to claim 1, wherein, The posture solving module adopts an EKF algorithm to perform optimal estimation on the data of the visual sensing unit and the inertial sensing unit, and finally performs correction on the state of the visual marker points recorded by the visual sensing unit data to perform optimal state estimation on the motion posture of the turtle robot and output a state vector sequence.

4. The experimental platform for constructing an electrical stimulation-behavior mapping relationship according to claim 1, wherein, The EKF algorithm is as follows: Prediction stage: the data of the inertial sensing unit and the optimal state at the last time are used to predict the state at the current time through a nonlinear state transition function; Update stage: when the data of the visual sensing unit is received, the predicted value is corrected through Kalman gain to obtain the optimal state estimation at the current time; Correction stage: the data of the visual sensing unit record the state of the visual marker points, the relative velocity of each limb end relative to its joint is calculated, the swing intensity of the limb end is reflected through the length of the vector, and then the optimal state estimation is performed on the motion posture of the turtle robot and a state vector sequence is output; After each update stage is completed, the current action label is output, and the action label is fed back to the posture solving module to adjust the process noise covariance matrix Q before the next prediction stage starts: If the action label is static, the diagonal elements in the Q matrix related to the changes of velocity and angular velocity are set to a value Q_static. If the action label is forward or turning, the diagonal elements are set to a value Q_dynamic, wherein Q_static is less than Q_dynamic.

5. The experimental platform for constructing electrical stimulation-behavior mapping relationship according to claim 1, characterized in that, The behavior label is a decision rule for going forward: whether the current speed norm is greater than a threshold , and the heading angular velocity is less than a threshold ; Whether there is a coordinated movement of diagonal limbs.

6. The experimental platform for constructing electrical stimulation-behavior mapping relationship according to claim 1, characterized in that, The decision rule for the behavior label of left turn is: the heading angle velocity whether it is greater than a threshold value ; whether the motion intensity of the right limb is significantly greater than that of the left limb.

7. The experimental platform for constructing electrical stimulation-behavior mapping relationship according to claim 1, characterized in that, The static determination rule is that the speed norm and the heading angular velocity are both less than respective threshold values.

8. The experimental platform for constructing an electrical stimulation-behavior mapping relationship according to claim 1, wherein, The velocity of the left front leg relative to the body center of mass c is defined as , and the velocity of the right rear leg relative to the center of mass c is . When , the projection component in the turtle's advancing direction is highly positively correlated, it can be confirmed that it is advancing.

9. The experimental platform for constructing electrical stimulation-behavior mapping relationship according to claim 1, characterized in that, By comparing the relative speed norm of the end of the limbs with respect to their joints, i.e. when > and the difference exceeds a threshold then it is indicated whether the movement intensity of the right limb is significantly greater than the left limb.

10. The method of claim 1, wherein the method further comprises: The method comprises the following steps: S1: a turtle robot is laid with visual marker points and an inertial sensing unit is fixed, and a preset electrical stimulation parameter scheme is loaded on a computer terminal; S2: an electrical stimulator is instructed to start stimulation, and a posture solution module estimates an optimal state of motion posture of the turtle robot according to data of the visual sensing unit and the inertial sensing unit, and outputs a state vector sequence; S3: a behavior recognition and data labeling module continuously analyzes the optimal state estimation vector, recognizes behavior according to a preset behavior determination rule, and instantly generates a data record, which comprises a time stamp, a current stimulation parameter, a state vector sequence and a behavior label; S4: after a preset time length of a group of stimulation parameters ends, the next group of parameters in a parameter list is automatically switched to and the monitoring and labeling steps are repeated until all parameters are traversed, and finally a complete electrical stimulation-behavior mapping database is saved.

Images