CN112571418B - Four-footed robot motion state clustering and autonomous decision-making method - Google Patents

Four-footed robot motion state clustering and autonomous decision-making method Download PDF

Info

Publication number
CN112571418B
CN112571418B CN202011473046.7A CN202011473046A CN112571418B CN 112571418 B CN112571418 B CN 112571418B CN 202011473046 A CN202011473046 A CN 202011473046A CN 112571418 B CN112571418 B CN 112571418B
Authority
CN
China
Prior art keywords
robot
motion
clustering
time domain
time
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202011473046.7A
Other languages
Chinese (zh)
Other versions
CN112571418A (en
Inventor
张帅帅
朱其刚
尹燕芳
刘明
荣学文
范永
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong University of Science and Technology
Original Assignee
Shandong University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shandong University of Science and Technology filed Critical Shandong University of Science and Technology
Priority to CN202011473046.7A priority Critical patent/CN112571418B/en
Publication of CN112571418A publication Critical patent/CN112571418A/en
Application granted granted Critical
Publication of CN112571418B publication Critical patent/CN112571418B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00—Programme-controlled manipulators
    • B25J9/16—Programme controls
    • B25J9/1628—Programme controls characterised by the control loop
    • B25J9/1653—Programme controls characterised by the control loop parameters identification, estimation, stiffness, accuracy, error analysis
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00—Programme-controlled manipulators
    • B25J9/16—Programme controls
    • B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
    • B25J9/1661—Programme controls characterised by programming, planning systems for manipulators characterised by task planning, object-oriented languages
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00—Programme-controlled manipulators
    • B25J9/16—Programme controls
    • B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
    • B25J9/1664—Programme controls characterised by programming, planning systems for manipulators characterised by motion, path, trajectory planning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Mechanical Engineering (AREA)
  • Robotics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Manipulator (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)

Abstract

A four-footed robot motion state clustering and autonomous decision-making method comprises the following steps: (1) recording a corresponding stepping time sequence when each foot of the robot steps for time domain feature extraction, recording the corresponding change of trunk roll angles when the robot is in different stable states, performing time domain feature extraction, and clustering the time domain features respectively; (3) establishing a comprehensive evaluation function of the movement performance of the quadruped robot; (4) performing an experiment of the quadruped robot through a complex terrain according to a comprehensive evaluation function to obtain relevant parameters in autonomous decision making of the robot; (5) and realizing the autonomous decision of control parameters according to the determined relevant parameters in the autonomous decision and the real-time stepping time and trunk roll angle information in the robot motion process, and autonomously generating the motion adaptive to the current walking terrain and the motion state of the robot. The method comprehensively utilizes the step time and the current stability clustering analysis result to carry out autonomous decision making, realizes the adaptivity of the quadruped robot to the terrain, and ensures the real-time performance of control.

Description

Four-footed robot motion state clustering and autonomous decision-making method
Technical Field
The invention relates to a method for clustering and autonomously deciding a motion state of a four-footed bionic robot, which can be used for clustering and analyzing the motion state of the four-footed bionic robot according to the motion state data of the four-footed bionic robot when the four-footed bionic robot walks on a terrain with higher complexity and autonomously deciding according to an analysis result so as to improve the terrain adaptability and the autonomy of the motion state of the four-footed bionic robot, and belongs to the field of robot control.
Background
The complex terrain environment has outstanding complex diversity and is a main reason for causing various movement difficulties and dangerous situations of the quadruped robot. The real-time motion state of the robot when walking on the complex terrain is accurately sensed, and the method is an important precondition for formulating a control strategy according with the current motion state and realizing stable walking.
In the aspect of research on the motion state perception of the quadruped robot, research on a motion state estimation method of a body of the quadruped robot (Shenyang university Master academic paper 2014) mainly aims at the motion state of the body of the quadruped robot, provides a body speed state calculation method based on a kinematics model, and realizes the perception of the speed state of the body of the quadruped robot; in the design of a four-footed bionic robot state information distributed parallel perception system (Master academic thesis of university of China, 2013), a state perception system design method with high real-time performance and reliability and a state perception core unit hardware design method of the four-footed bionic robot are researched, the structure and state information of the four-footed bionic robot are analyzed, the state perception system structure is established, and a method for solving the problems of state perception real-time performance and reliability is provided; in the Real-time foot attitude sensing for a human and not robot based on inertial sensors and sensors (Robotics and biometics, 2008.ROBIO 2008.IEEE International Conference on. IEEE,2009: 365-.
At present, although some research is carried out on the motion state perception aspect of a robot through partial results, a matched motion state real-time analysis method cannot be provided according to the typical characteristics of the motion state of the quadruped robot when the quadruped robot walks on a complex terrain, the motion state analysis result is not subjected to subsequent deep analysis, and the autonomous decision of the robot on motion parameters cannot be realized.
Disclosure of Invention
In order to solve the problem that the autonomous decision-making of the robot on the motion parameters cannot be realized in the prior art, the invention provides a method for clustering and analyzing the motion state of the quadruped robot by using a clustering algorithm and further realizing the autonomous decision-making when the quadruped robot walks with a static gait, so that the autonomy and terrain adaptability of the quadruped robot in a complex and unknown terrain environment are improved.
The invention discloses a four-footed robot motion state clustering and autonomous decision-making method, which specifically comprises the following steps:
(1) recording a corresponding step time sequence when each foot of the robot takes a step in the walking process of the robot on the terrain with different rugged degrees, extracting time domain characteristics of the sequence, and clustering the sequence into 3 classes by using a fuzzy clustering method according to the extracted time domain information;
(2) recording the corresponding change of the trunk roll angle when the robot has different stable states in the motion process, extracting time domain characteristics, and clustering the extracted time domain information into 3 classes by using a fuzzy clustering method according to the extracted time domain information;
(3) integrating energy consumption and stability evaluation indexes in the motion process of the robot, and establishing a comprehensive evaluation function J (t) of the motion performance of the quadruped robot;
(4) according to the comprehensive evaluation function, performing an experiment of the quadruped robot passing through a complex terrain respectively under different stepping time clustering and stability clustering states, and analyzing experiment data to obtain relevant parameters in autonomous decision making of the robot;
(5) and realizing the autonomous decision of control parameters according to the determined relevant parameters in the autonomous decision and the real-time stepping time and trunk roll angle information in the robot motion process, and autonomously generating the motion adaptive to the current walking terrain and the motion state of the robot.
The time domain feature extraction process of the stepping time in the robot motion process in the step (1) is as follows:
in the h-th stepping process of the quadruped robot, the time from foot lifting to touch the ground is thAnd storing the stepping time of each foot of the quadruped robot in the walking process into an array S, wherein the following formula is shown:
S＝(t1,t2,…,th)，
in order to effectively utilize the step time to evaluate the walking terrain, before the step time is subjected to cluster analysis, the time domain characteristics corresponding to the first k step times of the robot are extracted according to the data in the array S, and the method comprises the following steps:
(minimum value)hIk
hIk＝min(th-k+1,th-k+2,…,th)，
Maximum value-hMk
hMk＝max(th-k+1,th-k+2,…,th)，
③ Peak and Peak valueshPk
hPk＝hMk-hIk，
Average valuehAk
Figure GDA0003473634810000021
Standard deviation ofhSk
Figure GDA0003473634810000022
Synthesizing the extracted time domain characteristics of the first k stepping motions of the robot to form a characteristic matrixTC：
Figure GDA0003473634810000023
The time domain feature extraction process of the trunk roll angle in the robot motion process in the step (2) is as follows:
setting a sampling period, and recording the transverse roll angle of the robot trunk in the qth sampling period as alphaqAnd storing the transverse rolling angle of the trunk in the moving process of the quadruped robot into an array R, wherein the transverse rolling angle is shown as the following formula:
R＝(α1,α2,…,αq)，
according to data in the array R, extracting time domain features corresponding to attitude angles of the robot in the first R sampling periods, wherein the time domain features comprise:
(minimum value)qIr
qIr＝min(αq-r+1,αq-r+2,…,αq)，
Maximum value-qMr
qMr＝max(αq-r+1,αq-r+2,…,αq)，
③ average valueqAr
Figure GDA0003473634810000031
Standard deviation ofqSr
Figure GDA0003473634810000032
Synthesizing the extracted time domain characteristics corresponding to the attitude angle of the robot in the first r sampling periods to form a characteristic matrixRT：
Figure GDA0003473634810000033
The steps of using the fuzzy clustering method in the step (1) and the step (2) are as follows:
the fuzzy clustering comprises the following steps:
initializing a membership degree U ═ Uij]Matrix U(0)Wherein u isijThe membership degree of the jth sample data to the ith class;
② calculating the ith class center vector ci：
Figure GDA0003473634810000034
Where m is a membership factor, typically 2; and N is the total number of clusters.
Updating the membership degree matrix;
Figure GDA0003473634810000035
wherein C is a cluster number, xiFor the ith sample data, cjAnd ckClass j and class k center vectors, respectively.
If U(r+1)-UrIf | | < epsilon, the iteration is stopped. Where r is the number of iterations and epsilon is a constant value (a smaller value is preset).
The specific process of the step (3) is;
evaluation of energy consumption of the robot:
the acceleration of the robot trunk movement is used to measure the energy consumption in the gravity center adjustment stage, as follows:
Figure GDA0003473634810000041
wherein the content of the first and second substances,Bcyand (t) represents the motion track of the robot in the lateral direction during the adjustment of the center of gravity.
Evaluating the stability of the robot:
measuring the stability of the robot in the motion process by using the roll angle of the trunk posture, recording alpha (t) to represent the change of the trunk roll angle in the motion process of the robot, and then obtaining the corresponding stability evaluation functionαJ (t) is represented by the following formula:
αJ(t)＝∫|α(t)|dt，
the robot motion evaluation function J (t) is:
J(t)＝BJ(t)+αJ(t)。
the invention carries out cluster analysis on the stepping time of each foot in the walking process of the robot and indirectly evaluates the roughness of the walking terrain; and carrying out clustering analysis on the motion state by using the transverse rolling angle of the trunk so as to evaluate the stable state of the robot. And (4) performing autonomous decision making by comprehensively utilizing the stepping time and the current stability clustering analysis result, and realizing the adaptivity of the quadruped robot to the terrain. Has the following beneficial effects:
1. the robot can accurately evaluate the self motion state through the self motion state including the stepping time and the change of the trunk attitude angle, and can obtain the roughness of the current walking terrain and the influence of the roughness on the stability of the robot without depending on visual equipment;
2. the motion state clustering analysis method can reasonably classify the motion state, can effectively reduce the complexity of control strategy decision and ensure the real-time performance of a control system;
3. the motion planning process takes the stability and energy consumption of the robot motion into consideration, and the motion performance of the robot is optimized;
4. the autonomous decision making of the quadruped robot in a complex terrain environment can be realized, the autonomy of the quadruped robot is improved, and the intelligent level is further improved.
Drawings
Fig. 1 is a simulation model diagram of a twelve-degree-of-freedom quadruped bionic robot.
FIG. 2 shows the stability desire SmSchematic representation of (a).
Fig. 3 is a flow chart of a gait of the robot.
Fig. 4 is a rectangular plantar trajectory of a robot swing foot.
Fig. 5 is a diagram of an optimal parameter training process (simulation experiment process).
Fig. 6 is a diagram illustrating an example of a robot simulation experiment.
Fig. 7 is a robot gait control flow chart.
Detailed Description
The study object of the invention is a bionic quadruped robot, which has twelve degrees of freedom as shown in figure 1. The four-footed bionic robot has various walking gaits and can walk stably on terrains with different complexities. On the terrain with higher complexity, the quadruped robot is generally usedThe static gait can effectively increase the stability of the human body. When the robot walks with a static gait, the stability of the robot is often increased by adjusting the trunk in the lateral direction, and in order to make the robot have enough desire to stabilize, a minimum stability margin (S) is generally set in the static gait planningm) As shown in fig. 2. The robot increases the stability margin in the walking process through the movement of the trunk in the lateral direction. However, if SmThe value of (a) is set to be large, and the movement amount Δ y of the trunk in the lateral direction (see fig. 2) is increased, thereby increasing the energy consumption of the robot.
Therefore, the invention provides a method for clustering motion states and autonomously deciding the motion states of the quadruped robot, so that the quadruped robot can autonomously adjust S according to the change of the complexity of the walking terrain and the self stability thereofmThe value of (a) is effective to improve the complex terrain adaptability and the autonomous ability, and a basic gait flow chart is shown in fig. 3.
When the robot walks on a terrain with high roughness, the change of the roughness of the terrain is an important factor influencing the stability of the robot, the step time of swinging feet of the robot can be directly influenced by the height of obstacles in the terrain and the fluctuation of the terrain, and when the obstacles with different heights exist in the terrain or the fluctuation is large, the corresponding step time of each foot of the robot can have large difference when the robot swings. Therefore, the walking time of the robot in the walking process can reflect the rugged degree of the walking terrain to a certain extent. According to the invention, the rugged degree of the walking terrain is indirectly evaluated according to the walking time of each foot in the walking process of the robot.
As shown in fig. 4, the rectangular plantar trajectory can help the quadruped robot smoothly pass through complex terrain without colliding with obstacles on the terrain in case of unknown terrain information. In the h-th stepping process of the quadruped robot, the time from foot lifting to touch the ground is thAnd storing the stepping time of each foot of the quadruped robot in the walking process into an array S, wherein the following formula is shown:
S＝(t1,t2,…,th)，
in order to accurately evaluate the complexity of the walking terrain by using the stepping time of the robot, before the clustering analysis is carried out on the motion state of the robot, the time domain characteristics of the robot in the previous k stepping motions are extracted according to the data in the array S, and the method comprises the following steps:
(minimum value)hIk
hIk＝min(th-k+1,th-k+2,…,th)，
Maximum value-hMk
hMk＝max(th-k+1,th-k+2,…,th)，
③ Peak and Peak valueshPk
hPk＝hMk-hIk，
Average valuehAk
Figure GDA0003473634810000051
Standard deviation ofhSk
Figure GDA0003473634810000052
Synthesizing the above feature information to form a feature matrixTC：
Figure GDA0003473634810000061
The purpose of the body gravity center adjustment is to ensure the stability of the robot in the stepping process. The roll angle of the trunk can represent the stability of the robot in the stepping process, wherein the closer the value of the roll angle of the trunk is to zero and the smaller the change is, the better the stability of the trunk in the moving process is. Therefore, the roll angle of the trunk posture is used for measuring the stability of the robot in the invention. Setting a sampling period and recording the qthThe transverse roll angle of the robot trunk in each sampling period is recorded as alphaqAnd storing the transverse rolling angle of the trunk in the moving process of the quadruped robot into an array R, wherein the transverse rolling angle is shown as the following formula:
R＝(α1,α2,…,αq)，
in order to effectively utilize the roll angle information of the trunk to evaluate the motion stability of the robot, before clustering analysis is carried out on the trunk attitude angle, according to data in an array R, time domain features corresponding to the attitude angle of the robot in the first R sampling periods are extracted, and the method comprises the following steps:
(minimum value)qIr
qIr＝min(αq-r+1,αq-r+2,…,αq)，
Maximum value-qMr
qMr＝max(αq-r+1,αq-r+2,…,αq)，
③ average valueqAr
Figure GDA0003473634810000062
Standard deviation ofqSr
Figure GDA0003473634810000063
Synthesizing the above feature information to form a feature matrixRT：
Figure GDA0003473634810000064
Time domain feature matrix according to the extracted stepping time sequence and the trunk roll angle change sequenceTC、αAnd C, respectively clustering the stepping time and the current stable state into 3 classes by using a fuzzy clustering method.
The fuzzy clustering comprises the following steps:
1. initializing membership degree U ═ Uij]Matrix U(0)Wherein u isijThe membership degree of the jth sample data to the ith class;
2. computing class i center vector ci；
Figure GDA0003473634810000071
Where m is a membership factor, typically 2, and N is the total number of clusters.
3. Updating a membership matrix;
Figure GDA0003473634810000072
wherein C is a cluster number, xiFor the ith sample data, cjAnd ckClass j and class k center vectors, respectively.
4. If | | | U(r+1)-UrIf | | < epsilon, the iteration is stopped. Where r is the number of iterations and epsilon is a preset smaller value.
Adjusting S according to the change of terrain roughness and self stabilitymCan realize the optimal movement of the robot on terrains with different complexity. Only will SmThe value of the parameter is determined within a reasonable range, so that the robot can be ensured not to influence the motion performance of the robot or even lose stability due to exceeding of the reasonable range in the parameter adjusting process, and the stability margin of the robot is effectively ensured.
SmThe determination formula is as follows:
Figure GDA0003473634810000073
wherein [ S ]min,Smax]Is SmBeta (beta is not less than 0) is an adjustment interval.
In order to realize the S pair of the quadruped robot in the motion processmIs required to determine Smin，SmaxAnd the value of β.
Determining S according to the result of clustering the step timemThe quadruped robot can adapt to the terrain; determining the value of beta according to the result of clustering the current trunk attitude angle to realize the S pair of the robot in the motion processmTo achieve optimal motion performance of the robot.
To determine the S corresponding to each state of the robotmin，SmaxAnd the value of beta, and provides an evaluation method of the movement performance of the robot. The stability is the basis that the robot passes through the rugged terrain, the energy consumption is an important scale for evaluating the performance of the foot type robot, and the stability of the movement and the optimization of the energy consumption must be considered in the moving process of the robot. Therefore, the motion performance of the robot is evaluated by integrating the energy consumption of the robot in the gravity center adjusting stage and the stability of the robot in the stepping stage.
1. Energy consumption evaluation of robots
The acceleration of the robot trunk movement is used to measure the energy consumption in the gravity center adjustment stage, as shown in the following formula:
Figure GDA0003473634810000074
wherein the content of the first and second substances,Bcyand (t) represents the motion track of the robot in the lateral direction during the adjustment of the center of gravity.
2. Evaluation of stability of robot
Measuring the stability of the robot in the motion process by using the roll angle of the trunk posture, recording alpha (t) to represent the change of the trunk roll angle in the motion process of the robot, and then obtaining the corresponding stability evaluation functionαJ (t) represented by the following formula:
αJ(t)＝∫|α(t)|dt，
thus, the robot motion evaluation function j (t) is:
J(t)＝BJ(t)+αJ(t)。
and respectively carrying out an experiment of the quadruped robot passing through a complex terrain under different stepping time clustering and stability clustering states according to the comprehensive evaluation function by the robot motion evaluation function J (t) given above, and analyzing experimental data to obtain related parameters in autonomous decision making of the robot. Fig. 5 shows a flow chart of obtaining an autonomous decision parameter by an iterative method in a simulation experiment. Before the experiment, the value of the iteration number n is set, and the value of the optimal evaluation function is set by making the iteration number counting variable i equal to 0opJ＝0，SmUpper and lower limits of valueopSmin＝0、op S max0 and optimum adjustment intervalopβ ═ 0. In the simulation environment, the ruggedness of the terrain is adjusted by setting the height, size, and position of each obstacle in the simulation environment shown in FIG. 6, and then S is randomly generatedmin、SmaxBeta, the robot adopts the group of parameters to perform cluster analysis through the stepping time and the trunk attitude angle in the motion process and calculate an evaluation function in real timeiValue of J, comparisoniJ andopj, obtaining the value of the optimal evaluation function in the current iteration, judging whether the current time i reaches the iteration time n, if not, continuing the iteration, if so, recording the current optimal stepping time and the stable state clustering result, and the optimal parameter corresponding to the current motion state clusteringopSmin、opSmax、opBeta is used as the reference. According to the process, S corresponding to each motion state cluster of the robot can be determined through simulation experimentsmin，SmaxAnd the value of β.
And realizing the autonomous decision of control parameters according to the determined relevant parameters in the autonomous decision and the real-time stepping time, trunk roll angle and other information in the robot motion process, and autonomously generating the motion adaptive to the current walking terrain and the robot motion state. Fig. 7 shows the control flow of the robot in the walking process. The robot firstly carries out clustering analysis according to the stepping time and the attitude angle information, and then carries out clustering analysis according to the clusteringAnalysis results and S corresponding to each clustering result obtained by experimentmin，SmaxAnd the value of beta, calculating SmThe values of the parameters are used for realizing the autonomous decision of the motion parameters, and finally, the quadruped robot realizes the autonomous adaptation on the complex terrain by using the parameters, thereby effectively improving the terrain adaptation capability and the intelligent level of the quadruped robot.

Claims (4)

1. A four-footed robot motion state clustering and autonomous decision-making method is characterized by comprising the following steps:
(1) recording a corresponding step time sequence when each foot of the robot takes a step in the walking process of the robot on the terrain with different rugged degrees, extracting time domain characteristics of the sequence, and clustering the sequence into 3 classes by using a fuzzy clustering method according to the extracted time domain information;
(2) recording the corresponding change of the trunk roll angle when the robot has different stable states in the motion process, extracting time domain characteristics, and clustering the extracted time domain information into 3 classes by using a fuzzy clustering method according to the extracted time domain information;
(3) energy consumption and stability evaluation indexes in the motion process of the robot are integrated, and a comprehensive evaluation function of the motion performance of the quadruped robot is established;
(4) according to the comprehensive evaluation function, performing an experiment of the quadruped robot passing through a complex terrain respectively under different stepping time clustering and stability clustering states to obtain related parameters in autonomous decision making of the robot;
(5) realizing the autonomous decision of control parameters according to the determined relevant parameters in the autonomous decision and the real-time stepping time and trunk roll angle information in the robot motion process, and autonomously generating the motion adaptive to the current walking terrain and the robot motion state;
the time domain feature extraction process of the stepping time in the robot motion process in the step (1) is as follows:
in the h-th stepping process of the quadruped robot, the time from foot lifting to touch the ground is thAnd storing the stepping time of each foot of the quadruped robot in the walking process into an array S, wherein the following formula is shown:
S＝(t1,t2,…,th)，
according to the data in the array S, extracting time domain characteristics corresponding to the first k stepping times of the robot, wherein the time domain characteristics comprise:
(minimum value)hIk
hIk＝min(th-k+1,th-k+2,…,th)，
Maximum value-hMk
hMk＝max(th-k+1,th-k+2,…,th)，
③ Peak and Peak valueshPk
hPk＝hMk-hIk，
Average valuehAk
Figure FDA0003473634800000011
Standard deviation ofhSk
Figure FDA0003473634800000012
Synthesizing the extracted time domain characteristics of the first k stepping motions of the robot to form a characteristic matrixTC：
Figure FDA0003473634800000021
The time domain feature extraction process of the trunk roll angle in the robot motion process in the step (2) is as follows:
setting a sampling period, and recording the transverse roll angle of the robot trunk in the qth sampling period as alphaqThe transverse rolling angle of the trunk of the quadruped robot in the moving processStored in array R, as shown below:
R＝(α1,α2,…,αq)，
according to data in the array R, extracting time domain features corresponding to attitude angles of the robot in the first R sampling periods, wherein the time domain features comprise:
(minimum value)qIr
qIr＝min(αq-r+1,αq-r+2,…,αq)，
Maximum value-qMr
qMr＝max(αq-r+1,αq-r+2,…,αq)，
③ average valueqAr
Figure FDA0003473634800000022
Standard deviation ofqSr
Figure FDA0003473634800000023
Synthesizing the extracted time domain characteristics corresponding to the attitude angle of the robot in the first r sampling periods to form a characteristic matrixRT：
Figure FDA0003473634800000024
2. The method for clustering and autonomously deciding the motion state of a quadruped robot according to claim 1, wherein the fuzzy clustering method used in the steps (1) and (2) is as follows:
initializing a membership degree U ═ Uij]Matrix U(0)Wherein u isijThe membership degree of the jth sample data to the ith class;
② calculating the ith class center vector ci；
Figure FDA0003473634800000031
Wherein m is a membership factor, and N is the total number of clusters;
updating the membership degree matrix;
Figure FDA0003473634800000032
wherein C is a cluster number, xiFor the ith sample data, cjAnd ckRespectively are j-th class and k-th class central vectors;
if U(r+1)-UrStopping iteration if | is less than epsilon; where r is the number of iterations and ε is a constant value.
3. The method for clustering and autonomously deciding the motion state of the quadruped robot according to claim 1, wherein the specific process for establishing the comprehensive evaluation function in the step (3) is;
evaluation of energy consumption of the robot:
the acceleration of the robot trunk movement is used to measure the energy consumption in the gravity center adjustment stage, as follows:
Figure FDA0003473634800000033
wherein the content of the first and second substances,Bcy(t) represents the motion track of the robot in the lateral direction during the adjustment of the center of gravity;
evaluating the stability of the robot:
measuring the stability of the robot in the motion process by using the transverse rolling angle of the trunk posture, recording alpha (t) to represent the change of the trunk transverse rolling angle in the motion process of the robot, and evaluating the corresponding stabilityFunction(s)αJ (t) is represented by the following formula:
αJ(t)＝∫|α(t)|dt，
the robot motion evaluation function J (t) is:
J(t)＝BJ(t)+αJ(t)。
4. the method for clustering and autonomously deciding the motion state of the quadruped robot according to claim 1, wherein the determination process of the relevant parameters in the autonomous decision of the robot in the step (4) is as follows:
minimum stability margin S of robot in static gait planningmThe determination formula is as follows:
Figure FDA0003473634800000034
Sm∈[Smin,Smax]，
wherein [ S ]min,Smax]Is SmBeta (beta is more than or equal to 0) is an adjustment interval,
obtaining S corresponding to each state of the robot through a simulation experiment according to the robot motion evaluation function J (t) given in the step (3)min，SmaxAnd the value of β.
CN202011473046.7A 2020-12-15 2020-12-15 Four-footed robot motion state clustering and autonomous decision-making method Active CN112571418B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011473046.7A CN112571418B (en) 2020-12-15 2020-12-15 Four-footed robot motion state clustering and autonomous decision-making method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011473046.7A CN112571418B (en) 2020-12-15 2020-12-15 Four-footed robot motion state clustering and autonomous decision-making method

Publications (2)

Publication Number Publication Date
CN112571418A CN112571418A (en) 2021-03-30
CN112571418B true CN112571418B (en) 2022-02-25

Family

ID=75136189

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011473046.7A Active CN112571418B (en) 2020-12-15 2020-12-15 Four-footed robot motion state clustering and autonomous decision-making method

Country Status (1)

Country Link
CN (1) CN112571418B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113190029B (en) * 2021-04-06 2023-02-28 北京化工大学 Adaptive gait autonomous generation method of four-footed robot based on deep reinforcement learning
CN113401244A (en) * 2021-06-21 2021-09-17 北京理工大学 Automatic jack jacking disaster relief system based on sufficient robot
CN116256972B (en) * 2022-11-29 2024-01-23 哈尔滨理工大学 Six-foot robot man-machine instruction combination optimization method

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4498036A (en) * 1983-08-15 1985-02-05 Pneumo Corporation Adaptive control system
CN103786806A (en) * 2014-01-20 2014-05-14 北京航空航天大学 Multifunctional leg-and-wheel combination robot and multi-movement-mode intelligent switching method thereof
CN104267720A (en) * 2014-10-24 2015-01-07 山东大学 Method for generating free gaits for four-footed bionic robot
CN106547206A (en) * 2016-06-29 2017-03-29 中南大学 A kind of multi-foot robot dynamic stability based on strength of one's legs algorithm for estimating quantitatively judges method
CN108237531A (en) * 2016-12-26 2018-07-03 电子科技大学中山学院 A kind of humanoid robot gait's Learning Control Method
CN110328670A (en) * 2019-08-27 2019-10-15 山东科技大学 The quiet gait planning method of quadruped robot based on landform fuzzy self-adaption
CN110842921A (en) * 2019-11-19 2020-02-28 山东大学 Gait planning method for climbing and crossing of large-gradient terrain or high-obstacle of four-footed robot
CN111168684A (en) * 2020-01-22 2020-05-19 西北工业大学深圳研究院 On-orbit assembly sequence planning method for large-scale spatial structure
CN111338338A (en) * 2020-02-20 2020-06-26 山东科技大学 Robot speed self-adaptive control method based on road surface characteristic cluster analysis

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5854695B2 (en) * 2010-09-22 2016-02-09 キヤノン株式会社 Robot apparatus control method and robot apparatus

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4498036A (en) * 1983-08-15 1985-02-05 Pneumo Corporation Adaptive control system
CN103786806A (en) * 2014-01-20 2014-05-14 北京航空航天大学 Multifunctional leg-and-wheel combination robot and multi-movement-mode intelligent switching method thereof
CN104267720A (en) * 2014-10-24 2015-01-07 山东大学 Method for generating free gaits for four-footed bionic robot
CN106547206A (en) * 2016-06-29 2017-03-29 中南大学 A kind of multi-foot robot dynamic stability based on strength of one's legs algorithm for estimating quantitatively judges method
CN108237531A (en) * 2016-12-26 2018-07-03 电子科技大学中山学院 A kind of humanoid robot gait's Learning Control Method
CN110328670A (en) * 2019-08-27 2019-10-15 山东科技大学 The quiet gait planning method of quadruped robot based on landform fuzzy self-adaption
CN110842921A (en) * 2019-11-19 2020-02-28 山东大学 Gait planning method for climbing and crossing of large-gradient terrain or high-obstacle of four-footed robot
CN111168684A (en) * 2020-01-22 2020-05-19 西北工业大学深圳研究院 On-orbit assembly sequence planning method for large-scale spatial structure
CN111338338A (en) * 2020-02-20 2020-06-26 山东科技大学 Robot speed self-adaptive control method based on road surface characteristic cluster analysis

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
复杂地形环境中四足机器人行走方法研究;张帅帅;《中国博士学位论文全文数据库 信息科技辑》;21610105;第1-8,14-15页 *

Also Published As

Publication number Publication date
CN112571418A (en) 2021-03-30

Similar Documents

Publication Publication Date Title
CN112571418B (en) Four-footed robot motion state clustering and autonomous decision-making method
Yuan et al. A novel GRU-RNN network model for dynamic path planning of mobile robot
Levine et al. End-to-end training of deep visuomotor policies
Mac et al. Heuristic approaches in robot path planning: A survey
CN112356830B (en) Intelligent parking method based on model reinforcement learning
CN113031528B (en) Multi-legged robot non-structural ground motion control method based on depth certainty strategy gradient
CN111552183B (en) Six-legged robot obstacle avoidance method based on adaptive weight reinforcement learning
CN111673733B (en) Intelligent self-adaptive compliance control method of robot in unknown environment
CN111915246B (en) Parallel detection method for grain storage quantity of granary
Li et al. Sequential sensor fusion-based real-time LSTM gait pattern controller for biped robot
CN111382699A (en) Dynamic gesture recognition method based on particle swarm optimization LSTM algorithm
Wang et al. Vehicle lateral stability control based on stability category recognition with improved brain emotional learning network
CN110287896B (en) Human body behavior recognition method based on heterogeneous layered PSO and SVM
CN109227550A (en) A kind of Mechanical arm control method based on RBF neural
CN116594289A (en) Robot gesture pre-adaptation control method and device, electronic equipment and storage medium
Hu et al. Estimating biped gait using spline-based probability distribution function with Q-learning
Zhang et al. Direction-decision learning based pedestrian flow behavior investigation
Racz et al. Artificial neural network for mobile robot topological localization
CN106485283B (en) A kind of particle filter pedestrian target tracking based on Online Boosting
Thakur et al. Human joints analysis system: A machine learning approach
CN110990769B (en) Attitude migration algorithm system suitable for multi-degree-of-freedom robot
Ou et al. Support vector machine based approach for abstracting human control strategy in controlling dynamically stable robots
Yu et al. Walking simulation of biped robot on inclined plane based on gait recognition
Kumar et al. Robotic Terrain Classification with Loss Regularization
Saleh et al. Modeling and control of 3-omni wheel Robot using PSO optimization and Neural Network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant