A method for evaluating the similarity between manta ray-like robots and manta ray motion modes

Abstract

The present invention relates to a method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray. The method comprises selecting the end points of the manta ray-like robot's pectoral fin rays as i feature points, and selecting feature points at the end points of the manta ray's pectoral fin rays in the same proportion as the manta ray. A motion capture device is used to record the motion of the manta ray and the manta ray-like robot. A DTW algorithm is used to evaluate the similarity of the motion postures of the manta ray-like robot and the manta ray. A method for evaluating the similarity of the motion performance of the manta ray-like robot and the manta ray is established, decomposing the motion into three basic modes: forward swimming, pitching, and yaw. The similarity of the motion performance of the manta ray-like robot is calculated, and the similarity between the manta ray-like robot and the manta ray is evaluated. The present invention establishes a similarity evaluation system between the manta ray-like robot and the manta ray, providing a direction for optimizing the biomimetic motion control of the manta ray-like robot.

Description

Technical Field

[0001] The present invention belongs to the technical field of bionic robots and relates to a method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray, and specifically to a method for evaluating the similarity of bionic motion of a manta ray-like robot. The method can quantitatively evaluate the motion similarity between the manta ray-like robot and the manta ray, and points out a direction for the motion optimization control strategy of the manta ray-like robot. Background Art

[0002] Manta rays are a typical fish that propuls itself by swinging their pectoral fins. Their streamlined, flattened bodies and pectoral fins with a large aspect ratio give them high propulsion efficiency, stability, and maneuverability, making them a good model for underwater robotics. For manta ray-like robots, a higher degree of fidelity translates to greater biocompatibility and stealth, broadening their application scope to areas such as deep-sea fish monitoring, marine aquariums, and underwater recreational manned vehicles. While mimicking manta ray motion, studying its mechanisms can also guide robots toward achieving greater maneuverability and stability.

[0003] In the research of manta ray-like robots, researchers from various countries have conducted extensive studies on everything from appearance and structure to motion control design to improve their performance. However, current manta ray-like robots still differ significantly from real manta rays in terms of appearance, structure, motion modes, and performance. Manta ray-like robots can visually mimic various manta ray movements, but relying on manual evaluation to determine whether the imitation is accurate lacks scientific validity. Previous studies have not quantitatively evaluated this similarity. The lack of a quantitative method for evaluating motion similarity makes it difficult to improve manta ray-like robots in terms of motion imitation, limiting the application of optimized control strategies for manta ray-like robots. While motion state is often intuitively represented by the robot's posture during movement—that is, its trajectory in space—biomimeticism doesn't simply mean mimicking the motion posture of a living being; it also involves influencing the robot's motion performance through this posture. Therefore, the present invention establishes a quantitative motion state similarity evaluation method to evaluate manta ray-like robots based on both their motion posture and performance. Summary of the Invention

[0004] Technical problems to be solved

[0005] In order to overcome the shortcomings of the prior art, the present invention proposes a method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray.

[0006] Technical Solution

[0007] A method for evaluating the similarity between a manta ray-like robot and a manta ray's motion modes, characterized by the following steps:

[0008] Step 1: Select the end points of the pectoral fin rays of the manta ray robot as i feature points, and select feature points at the end points of the pectoral fin rays of the manta ray in the same proportion;

[0009] Step 2: Motion state capture and data preprocessing: The motion of the manta ray and the manta ray-like robot is recorded using motion capture equipment;

[0010] The data lost during capture is processed using the Lagrange interpolation method:

[0011] If the coordinates of the manta ray feature point at time t are lost, the manta ray feature point sequence M is obtained by processing it using the Lagrange interpolation method. i ={q i1 ,q i2 ,q i3 ,...,q in}, where i is the number of selected pectoral fin feature points;

[0012] If the coordinates of the feature points of the manta ray robot at time t are lost, the feature point sequence of the manta ray robot is obtained by Lagrange interpolation method as Q i ={w i1 ,w i2 ,w i3 ,...,w im}, i is the number of selected pectoral fin feature points;

[0013] Step 3: Using the data processed in step 2, the DTW algorithm is used to evaluate the similarity between the manta ray robot and the manta ray's motion postures:

[0014]

[0015] Where A is the number of feature points, and m is the number of feature point coordinates in each feature point sequence;

[0016] Step 4: Establish a method to evaluate the similarity between the manta ray robot and the manta ray's motion performance:

[0017]

[0018] Among them: β1, β2, β3 are the weights of the basic attitudes of forward swimming, yaw, and pitch, ρ 11 is the weight of maneuverability in the forward swimming posture, ρ 12 is the weight of stability in the forward swimming posture, ρ 21 is the weight of maneuverability in yaw attitude, ρ 22 is the weight of stability in yaw attitude, ρ 31 is the weight of maneuverability in pitch attitude, ρ 32is the weight of stability in pitch attitude; speed, pitch, roll, and yaw represent speed, pitch angle, roll angle, and heading angle respectively. m_max Indicates the maximum forward swimming speed of the manta ray, speed q_max Indicates the maximum forward swimming speed of the manta ray robot; pitch m_t Indicates the range of the manta ray's pitch angle change over time, pitch q_t Indicates the range of the manta ray robot's pitch angle change over time; roll m_t Indicates the range of the manta ray's roll angle change over time, roll m_t Indicates the range of the manta ray robot's roll angle change over time; yaw m_max Indicates the maximum yaw speed of the manta ray, yaw q_max Indicates the maximum yaw speed of the manta ray robot; pitch m_max Indicates the maximum pitch angle of the manta ray, pitch q_max Indicates the maximum pitch angle of the manta ray robot;

[0019] Step 5: Motion state similarity evaluation, combined with motion posture similarity S ma (M i ,Q i )Evaluation method, motion performance similarity S sp (M i ,Q i ) evaluation method, and comprehensively establish the manta ray-like robot and manta ray similarity evaluation using the following formula:

[0020]

[0021] Among them, κ1 and κ2 are the weights of motion posture and motion performance in the total motion state similarity.

[0022] Three characteristic points of the end points of the pectoral fin rays are selected.

[0023] The capture modes of the motion capture device include: mechanical electric, optical and electromagnetic induction.

[0024] In step 2, if the coordinates of the manta ray feature point at time t are lost, construct the interpolation function at time t Then we can get the coordinate q at time t t After processing, the manta ray feature point sequence is M i ={q i1 ,q i2 ,q i3 ,...,q im}, where i is the number of selected pectoral fin feature points.

[0025] In step 2, if the coordinates of the feature points of the manta ray robot at time t are lost, construct the interpolation function at time t Then the coordinate w at time t can be obtained. After processing, the feature point sequence of the manta ray robot is Q i ={w i1 ,w i2 ,w i3 ,...,w im}.

[0026] The speed, pitch angle, roll angle and heading angle of the manta ray-like robot are measured by a posture sensor carried by the robot.

[0027] The manta ray speed, pitch angle, roll angle, and heading angle are obtained from the optically collected video in step S2.

[0028] The weights β1, β2, and β3 of the basic postures of forward swimming, yaw, and pitch are: if the motion state of the manta ray-like robot is only forward swimming, then β1=1, β2=β3=0; if the motion state of the manta ray-like robot is only yaw, then β2=1, β1=β3=0; if the motion state of the manta ray-like robot is only pitch, then β3=1, β1=β2=0.

[0029] The κ1 and κ2 are 0.5.

[0030] Beneficial effects

[0031] This invention proposes a method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray. The method selects the terminal points of the manta ray-like robot's pectoral fin rays as i feature points, and selects feature points at the terminal points of the manta ray's pectoral fin rays in the same proportion as the manta ray. The motions of the manta ray and the manta ray-like robot are recorded using a motion capture device. The motion postures of the manta ray-like robot and the manta ray are similarly evaluated using the DTW algorithm. A method for evaluating the similarity of the motion performance of the manta ray-like robot and the manta ray is established. The similarity is calculated for the motion performance of the manta ray-like robot and the manta ray, and the similarity between the manta ray-like robot and the manta ray is evaluated. This method establishes a similarity evaluation system between the manta ray-like robot and the manta ray, providing a direction for optimizing the biomimetic motion control of the manta ray-like robot.

[0032] The specific beneficial effects of the present invention are:

[0033] (1) When collecting manta ray data, the presence of a large number of other fish in the observation area can easily obstruct the manta ray, resulting in data loss during motion capture. Therefore, the present invention uses Lagrange interpolation to process the lost data. After linear interpolation of the motion trajectory of the feature points, the continuity requirement can be guaranteed, which provides conditions for subsequent motion similarity analysis.

[0034] (2) The manta ray robot and the manta ray have different movement speeds and synchronization. Their movements are almost similar, but they are not necessarily one-to-one corresponding in time series. In view of the difficulty in aligning the time trajectories of the manta ray robot and the manta ray, the DTW algorithm is used to calculate the similarity of movement postures. This algorithm takes into account the dynamic changes in time and can significantly reduce the problem of inaccurate similarity calculation caused by data collection errors.

[0035] (3) In view of the fact that the manta ray-like robot has multiple complex motion modes, this paper decomposes it into three basic modes: forward swimming, pitching, and yaw, and calculates the motion performance similarity of the manta ray-like robot to facilitate the motion performance similarity calculation.

[0036] (4) The present invention establishes a similarity evaluation system between the manta ray-like robot and the manta ray, which points out the direction for the optimization of the bionic motion control of the manta ray-like robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 :Flowchart of the similarity evaluation method for manta ray-like robot motion states

[0038] Figure 2 : Manta ray-like robot and manta ray feature point selection

[0039] Figure 3 : Comparison of the manta ray-like robot and manta ray feature point motion sequences DETAILED DESCRIPTION

[0040] The present invention will now be further described with reference to the embodiments and accompanying drawings:

[0041] Figure 1 The flowchart of the method for evaluating the motion similarity of a manta ray-like robot according to the present invention is shown, and includes the following steps:

[0042] S1: Motion posture feature point selection.

[0043] Manta rays rely on their broad pectoral fins to complete complex swimming patterns. Their motion posture is changed by the different movements of the pectoral fins. Therefore, for the quantitative analysis of motion posture similarity, this paper will use the trajectory relationship of the angles of the corresponding feature points on the pectoral fin edges to evaluate the similarity of the motion posture of the manta ray-like robot and the manta ray.

[0044] As for the selection of feature points, the present invention starts from the perspective of manta ray-like robot and selects the end point of the pectoral fin as the feature point. According to the proportional relationship, the feature points corresponding to the manta ray pectoral fin are selected, as shown in the following figure. Figure 2 shown.

[0045] S2: Motion state capture and data preprocessing.

[0046] Motion capture uses motion capture equipment to record the subject's movements for analysis and feature extraction. Common capture methods include mechanical, electric, optical, and electromagnetic induction. In the present invention, the biomimetic observation target is a manta ray, but observation sites and conditions are limited, and attaching sensors to the manta ray to capture its motion is difficult. Therefore, optical capture is the only method available for capturing manta ray motion.

[0047] Optical capture is used, but the presence of numerous other fish in the oceanarium obstructs the capture of manta ray motion, which can easily lead to data loss during motion capture. The high-definition camera in the observation equipment has a sampling frequency of 60Hz, with a sampling period of less than 0.04s. Manta ray movements within a sampling period are relatively small, so the present invention uses Lagrange interpolation to process lost data.

[0048] It is known that a function f(x) on the interval [a,b] is at n+1 different points f j The value y at j =f(x j )(j=0,1,2,...n), there exists an n-degree polynomial that satisfies the difference condition:

[0049]

[0050] in:

[0051]

[0052] Select n = 1; that is, linear interpolation, then:

[0053]

[0054]

[0055] If the coordinates of the manta ray feature point at time t are lost, the interpolation function at time t is constructed according to formula (4): Then we can get the coordinate q at time t t After processing, the manta ray feature point sequence is M i ={q i1 ,q i2 ,q i3 ,...,q im}, where i is the number of selected pectoral fin feature points.

[0056] In this embodiment: Get the coordinate q at time t t After processing, the manta ray feature point sequence is M i ={q i1 ,q i2 ,q i3 ,...,q im}(i=1,2,3), where i represents the three feature points of the pectoral fin.

[0057] If the coordinates of the feature points of the manta ray robot at time t are lost, the interpolation function at time t is constructed according to formula (4): Then the coordinate w at time t can be obtained. After processing, the feature point sequence of the manta ray robot is Q i ={w i1 ,w i2 ,w i3 ,...,w im}.

[0058] In this embodiment, the coordinate w at time t is obtained. After processing, the feature point sequence of the manta ray robot is Q i ={w i1 ,w i2 ,w i3 ,...,w im}(i=1,2,3).

[0059] The comparison of the pectoral fin feature point motion sequences of the manta ray and the manta ray-like robot is shown in the figure below. Figure 3 shown.

[0060] S3: Due to the random nature of animal motion, the manta ray robot and manta ray differ in speed and synchronization. While the time series of feature points may be roughly similar on the time axis and their movements are nearly identical, they are not one-to-one in the time series. Using traditional distance algorithms, calculations without considering dynamic temporal changes can lead to significant errors. The DTW algorithm can calculate the similarity between two time series of different lengths. Given the unique temporal trajectories of the manta ray robot and manta ray motion, this section uses the DTW algorithm to evaluate the similarity of the motion states of the manta ray robot and manta ray.

[0061] The motion posture S constructed using the DTW algorithm ma (M i ,Q i )The similarity calculation formula is as follows:

[0062]

[0063] Where A is the number of feature points and m is the number of feature point coordinates in each feature point sequence.

[0064] In this embodiment: A = 3. m is the number of feature point coordinates in each feature sequence.

[0065] S4: Establishment of motion performance similarity evaluation method.

[0066] Manta rays have multiple motion modes when swimming in the water, and their basic movements are forward swimming, yaw, and pitching. These three movements are the primitives for manta rays to complete highly maneuverable and highly stable multimodal motion. During forward swimming, the forward swimming speed is a manifestation of manta ray maneuverability, and the pitch angle change is a manifestation of stability; during yaw, the yaw speed is a manifestation of maneuverability, and the roll angle change is a manifestation of stability; during pitching, the maximum pitch angle is a manifestation of maneuverability, and the pitch angle change is a manifestation of stability. Therefore, based on the multimodal motion posture of manta rays, a method for evaluating the similarity of motion performance between the manta ray-like robot and the manta ray is established as follows:

[0067]

[0068] Among them: β1, β2, β3 are the weights of the basic attitudes of forward swimming, yaw, and pitch; ρ 11 is the weight of maneuverability in the forward swimming posture, ρ 12 is the weight of stability in the forward swimming posture; ρ 21 is the weight of maneuverability in yaw attitude, ρ 22 is the weight of stability in yaw attitude; ρ 31 is the weight of maneuverability in pitch attitude, ρ 32 is the weight of stability in pitch posture; speed, pitch, roll, yaw represent speed, pitch angle, roll angle, and heading angle respectively. The speed, pitch angle, roll angle, and heading angle of the manta ray robot are measured by the attitude sensor carried by the robot; the speed, pitch angle, roll angle, and heading angle of the manta ray are obtained from the optical video collected in step S2. m_max Indicates the maximum forward swimming speed of the manta ray, speed q_max Indicates the maximum forward swimming speed of the manta ray robot. pitch m_t Indicates the range of the manta ray's pitch angle change over time, pitch q_t Indicates the range of the manta ray robot's pitch angle change over time. m_t Indicates the range of the manta ray's roll angle change over time, roll m_t Indicates the range of the manta ray robot's roll angle change over time. m_max Indicates the maximum yaw speed of the manta ray, yaw q_max Indicates the maximum yaw speed of the manta ray robot. pitch m_max Indicates the maximum pitch angle of the manta ray, pitch q_max Indicates the maximum pitch angle of the manta ray robot.

[0069] If the motion state of the manta ray-like robot is only swimming forward, then β1=1, β2=β3=0; if the motion state of the manta ray-like robot is only yaw, then β2=1, β1=β3=0; if the motion state of the manta ray-like robot is only pitching, then β3=1, β1=β2=0.

[0070] S5: Establishment of motion state similarity evaluation method.

[0071] Combined with motion posture similarity S ma (M i ,Q i )Evaluation method, motion performance similarity S sp (M i ,Q i ) evaluation method, and comprehensively establish the manta ray-like robot and manta ray similarity evaluation method:

[0072]

[0073] where κ i The weight of the motion posture and motion performance in the total motion state similarity. In this case, it is set to 0.5, and can be adjusted according to the emphasis on posture similarity or performance similarity.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, reductions, equivalent substitutions, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for evaluating the similarity between a manta ray-like robot and a manta ray motion mode, characterized in that Here are the steps: Step 1: Select the end points of the pectoral fin rays of the manta ray robot as i feature points, and select feature points at the end points of the pectoral fin rays of the manta ray in the same proportion; Step 2: Motion state capture and data preprocessing: The motion of the manta ray and the manta ray-like robot is recorded using motion capture equipment; The data lost during capture is processed using the Lagrange interpolation method: If the coordinates of the manta ray feature point at time t are lost, the manta ray feature point sequence M is obtained by processing it using the Lagrange interpolation method. i ={q i1 ,q i2 ,q i3 ,...,q in }, where i is the number of selected pectoral fin feature points; If the coordinates of the feature points of the manta ray robot at time t are lost, the feature point sequence of the manta ray robot is obtained by Lagrange interpolation method as Q i ={w i1 ,w i2 ,w i3 ,...,w im }, i is the number of selected pectoral fin feature points; Step 3: Using the data processed in step 2, the DTW algorithm is used to evaluate the similarity between the manta ray robot and the manta ray's motion postures: Where A is the number of feature points, and m is the number of feature point coordinates in each feature point sequence; Step 4: Establish a method to evaluate the similarity between the manta ray robot and the manta ray's motion performance: Among them: β1, β2, β3 are the weights of the basic attitudes of forward swimming, yaw, and pitch, ρ 11 is the weight of maneuverability in the forward swimming posture, ρ 12 is the weight of stability in the forward swimming posture, ρ 21 is the weight of maneuverability in yaw attitude, ρ 22 is the weight of stability in yaw attitude, ρ 31 is the weight of maneuverability in pitch attitude, ρ 32 is the weight of stability in pitch attitude; speed, pitch, roll, and yaw represent speed, pitch angle, roll angle, and heading angle respectively. m_max Indicates the maximum forward swimming speed of the manta ray, speed q_max Indicates the maximum forward swimming speed of the manta ray robot; pitch m_t Indicates the range of the manta ray's pitch angle change over time, pitch q_t Indicates the range of the manta ray robot's pitch angle change over time; roll m_t Indicates the range of the manta ray's roll angle change over time, roll q_t Indicates the range of the manta ray robot's roll angle change over time; yaw m_max Indicates the maximum yaw speed of the manta ray, yaw q_max Indicates the maximum yaw speed of the manta ray robot; pitch m_max Indicates the maximum pitch angle of the manta ray, pitch q_max Indicates the maximum pitch angle of the manta ray robot; Step 5: Motion state similarity evaluation, combined with motion posture similarity S ma (M i ,Q i )Evaluation method, motion performance similarity S sp (M i ,Q i ) evaluation method, and comprehensively establish the manta ray-like robot and manta ray similarity evaluation using the following formula: Among them, κ1 and κ2 are the weights of motion posture and motion performance in the total motion state similarity.

2. The method for evaluating the similarity between the manta ray-like robot and the manta ray motion mode according to claim 1, wherein: Three characteristic points of the end points of the pectoral fin rays are selected.

3. The method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray according to claim 1, wherein: The capture modes of the motion capture device include: mechanical electric, optical and electromagnetic induction.

4. The method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray according to claim 1, wherein: In step 2, if the coordinates of the manta ray feature point at time t are lost, construct the interpolation function at time t Then we can get the coordinate q at time t t , after processing, the manta ray feature point sequence is M i ={q i1 ,q i2 ,q i3 ,...,q im }, where i is the number of selected pectoral fin feature points.

5. The method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray according to claim 1, wherein: In step 2, if the coordinates of the feature points of the manta ray robot at time t are lost, construct the interpolation function at time t Then the coordinate w at time t can be obtained. After processing, the feature point sequence of the manta ray robot is Q i ={w i1 ,w i2 ,w i3 ,...,w im }.

6. The method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray according to claim 1, wherein: The speed, pitch angle, roll angle and heading angle of the manta ray-like robot are measured by a posture sensor carried by the robot.

7. The method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray according to claim 1, wherein: The manta ray speed, pitch angle, roll angle, and heading angle are obtained from the optically collected video in step S2.

8. The method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray according to claim 1, wherein: The weights β1, β2, and β3 of the basic postures of forward swimming, yaw, and pitch are: if the motion state of the manta ray robot is only forward swimming, then β1=1, β2=β3=0; if the motion state of the manta ray robot is only yaw, then β2=1, β1=β3=0; if the motion state of the manta ray robot is only pitch, then β3＝1,β1＝β2＝0。 9. The method for evaluating the similarity of motion modes between a manta ray-like robot and a manta ray according to claim 1, wherein: The κ1 and κ2 are 0.5.

Images