CN104924313B - There is teach-by-doing teaching mechanical arm system and the method for learning by imitation mechanism - Google Patents
There is teach-by-doing teaching mechanical arm system and the method for learning by imitation mechanism Download PDFInfo
- Publication number
- CN104924313B CN104924313B CN201510244111.1A CN201510244111A CN104924313B CN 104924313 B CN104924313 B CN 104924313B CN 201510244111 A CN201510244111 A CN 201510244111A CN 104924313 B CN104924313 B CN 104924313B
- Authority
- CN
- China
- Prior art keywords
- mechanical arm
- teaching
- target object
- paw
- sensor
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 230000007246 mechanism Effects 0.000 title claims abstract description 15
- 230000006399 behavior Effects 0.000 claims abstract description 51
- 230000009471 action Effects 0.000 claims abstract description 47
- 238000001514 detection method Methods 0.000 claims abstract description 36
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 16
- 230000008569 process Effects 0.000 claims abstract description 14
- 210000002569 neuron Anatomy 0.000 claims description 14
- 230000004927 fusion Effects 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 12
- 230000001133 acceleration Effects 0.000 claims description 11
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 claims description 9
- 229910052744 lithium Inorganic materials 0.000 claims description 9
- 230000033001 locomotion Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 7
- 210000004247 hand Anatomy 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 230000003542 behavioural effect Effects 0.000 claims description 6
- 230000005540 biological transmission Effects 0.000 claims description 6
- 230000005611 electricity Effects 0.000 claims description 5
- 230000007935 neutral effect Effects 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- YVGGHNCTFXOJCH-UHFFFAOYSA-N DDT Chemical compound C1=CC(Cl)=CC=C1C(C(Cl)(Cl)Cl)C1=CC=C(Cl)C=C1 YVGGHNCTFXOJCH-UHFFFAOYSA-N 0.000 claims description 2
- 238000002939 conjugate gradient method Methods 0.000 claims description 2
- 238000011217 control strategy Methods 0.000 claims 1
- 241000282414 Homo sapiens Species 0.000 description 5
- 241001465754 Metazoa Species 0.000 description 5
- 238000004891 communication Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000019771 cognition Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- TVEXGJYMHHTVKP-UHFFFAOYSA-N 6-oxabicyclo[3.2.1]oct-3-en-7-one Chemical compound C1C2C(=O)OC1C=CC2 TVEXGJYMHHTVKP-UHFFFAOYSA-N 0.000 description 1
- 241000282412 Homo Species 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 239000003990 capacitor Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000010408 film Substances 0.000 description 1
- 238000013100 final test Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 210000000653 nervous system Anatomy 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000010408 sweeping Methods 0.000 description 1
- 229910052715 tantalum Inorganic materials 0.000 description 1
- GUVRBAGPIYLISA-UHFFFAOYSA-N tantalum atom Chemical compound [Ta] GUVRBAGPIYLISA-UHFFFAOYSA-N 0.000 description 1
- 239000010409 thin film Substances 0.000 description 1
Landscapes
- Manipulator (AREA)
- Steering Control In Accordance With Driving Conditions (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
There is teach-by-doing teaching mechanical arm system and the method for learning by imitation mechanism, described system includes the study module of the action executing module, the sensing module of multisensor composition and controller composition of mechanical arm and steering wheel composition.The present invention adopts the independently-powered mode of module, on this basis using teach-by-doing off-line teaching, by one group of action detection device being made up of gyro sensor and accelerometer, perceptually module is carried on the robotic arm, the status information of each connecting rod in motor process for the collection machinery arm, then by these Information application learning by imitation algorithms, instruct mechanical arm system learning by imitation teaching behavior.The present invention takes hierarchical classification control, improve specific aim and the information transfer efficiency of control system, the present invention can know the purpose of teaching behavior by study, still can complete imitation task when changing mechanical arm initial attitude or target object place orientation, have higher degree of intelligence.
Description
Technical field
The present invention relates to a kind of teach-by-doing teaching mechanical arm system with learning by imitation mechanism and method, belong to intelligent machine
Device people's category.
Background technology
Robot is developed so far all many-sides affecting modern humans' life considerablely, and people are to robot
Intelligence require improve constantly in case its preferably be the mankind service, robot learning be raising robot automtion degree one
Plant effective method.In nature, imitation is that human or animal grasps the motor skill the most effective learning style.And people
It is the bionical prototype of robot with animal, it is feasible for therefore the mechanism of imitation being applied to robot.Human or animal's is many
Technical ability or behavior in its nervous system cognitive process progressively formation and development get up, by imitating human or animal, make
The cognition of robot and behavior are more nearly human or animal, thus producing autonomous behavior, this process claims from learning method
For learning by imitation.
Learning by imitation make robot pass through observe demonstrator action just can be with Fast Learning to useful action.This
With the method exchanging learning new knowledge and solve problem of environment so that learning by imitation and traditional isolated robot learning
Method is compared, and learning by imitation has a lot of advantages：(1) learning efficiency is high, directly obtains effective information, has evaded huge studying space
Search problem；(2) lift intelligence, make robot constantly lift behavioral competence under surroundings；(3) practical, after study
Behavior can be done directly on working environment；(4) enhance interactive capability, make man machine relation more friendly；(5) decrease
The complexity of programming.Therefore, learning by imitation is all with a wide range of applications in military field or civil field.Mould
Imitative study is not only behavioral science and the research emphasis of neuroscience, also has very for artificial intelligence and machine learning
Important Research Significance.
At present, based on industrial robot, such robot can only base for the robot that teaching action can be reproduced
In teaching experience by programming to it, complete part deliberate action, degree of intelligence is relatively low, do not possess the ability of learning by imitation.Shen
Yang Junyou of positive polytechnical university et al. applies to learning by imitation mechanism on mechanical arm, controls graph model by gathering visual information
Complete the imitation of teaching action, apply in learning by imitation and some breakthroughs are obtained on actual robot.But used by this robot
Equipment cost is high, and the visual information of collection will be imitated and do not possess real-time, completed imitation task by further data processing
Time-consuming longer.
Content of the invention
In order to overcome the above-mentioned deficiencies of the prior art, the invention provides a kind of teach-by-doing with learning by imitation mechanism is shown
Religion mechanical arm system and method, by the direct drag operation mechanical arm of improved teach-by-doing teaching mode, by a kind of motion detection
Device completes the learning by imitation to teaching action after perceiving teaching behavior act information, so that mechanical arm system not only can be imitated again
Existing teaching action, additionally it is possible to know the purpose of teaching behavior by study, has higher degree of intelligence.Reduce traditional handss simultaneously
The labor intensity of handle teaching mode, reduces sensor cost, overcomes and gathers what teaching behavior brought using camera technique
The loaded down with trivial details problem of image procossing, improves the learning efficiency of mechanical arm system.
Mechanical arm system proposed by the present invention, is a kind of mechanical arm system with learning by imitation mechanism, and main inclusion is held
Row module, sensing module and study module.System performing module includes sixdegree-of-freedom simulation and paw；Sensing module includes moving
Make detection means and target object detection means；Study module, as the brain of control system, completes sensing data process, rudder
The control of machine controller and realize the functions such as learning by imitation algorithm.The present invention grips the dynamic of object on desktop with mechanical arm paw
As example, make the mechanical arm learning by imitation behavior, with the feasibility of checking system and method.
The technical solution adopted for the present invention to solve the technical problems is：
There is the teach-by-doing teaching mechanical arm system of learning by imitation mechanism, in this system, dsp controller (2), 7.4v lithium electricity
Pond (5), voltage converter (6), long U type connecting rod one (7) are fixedly mounted on mechanical arm system chassis (1)；51 single-chip microcomputers are minimum
System (3), steering engine controller (4) are connected with dsp controller (2) respectively；Described voltage converter (6) and 7.4v lithium battery (5)
Connect, 7.4v lithium battery (5) is connected with dsp controller (2)；The side of long U type connecting rod one (7) is provided with long U type connecting rod two
(8)；One end of long U type connecting rod two (8) is provided with ultrasonic sensor (18)；
The bottom of two long U type connecting rods connects composition H type connecting rod (9)；H type connecting rod (9) one end and long U type connecting rod two (8)
Side connect；Long U type connecting rod three (10) one end is connected with H type connecting rod (9) other end；Mechanical arm paw (11) and long U-shaped company
The other end of bar three (10) connects；
MG996R steering wheel one (12) is arranged on the junction of long U type connecting rod two (8) and long U type connecting rod one (7)；
MG996R steering wheel two (13) is arranged on the side junction of H type connecting rod (9) and long U type connecting rod two (8)；
MG996R steering wheel three (14) is arranged on one end junction of H type connecting rod (9) and long U type connecting rod three (10)；
MG996R steering wheel four (15) is arranged at the other end of long U type connecting rod three (10)；
MG996R steering wheel five (16) is arranged on the end of mechanical arm paw (11)；
MG996R steering wheel six (17) is arranged on the side of mechanical arm paw (11)；
Three-axis gyroscope sensor one (19) forms detection module one with accelerometer module one (22), and detection module one sets
Put in H type connecting rod (9) middle position；
Three-axis gyroscope sensor two (20) forms detection module two with accelerometer module two (23), and detection module two sets
Put in long U type connecting rod three (10) middle position；
Three-axis gyroscope sensor three (21) forms detection module three with accelerometer module three (24), and detection module three sets
Put the one end in mechanical arm paw (11)；
Detection module one, detection module two, detection module three form
Infrared distance sensor (25) is arranged on the centre position of MG996R steering wheel six (17)；
Touch sensor one (26), touch sensor two (27) are separately positioned on the paw up and down of mechanical arm paw (11)
Place.
During mechanical arm system work of the present invention, initially with teach-by-doing teaching method, mechanical arm is shown
Religion, carries out learning by imitation by after sensing module collection teaching information through control system, control machinery arm imitates teaching behavior act.
(1) teach-by-doing teaching and training data are processed
After mechanical arm system starts, the steering engine controller of system does not go up electricity, and that is, on mechanical arm, steering wheel is in power-off shape
State, by the way of off-line teaching, only sensing module is operated.Complete to grip the dynamic of object on desktop with handss driving machinery arm
Make, action detection device collection three handss during teaching of three three-axis gyroscope sensors and accelerometer module composition
The pose coordinate angular velocity of rotation of arm link and angular acceleration information, infrared distance sensor collection paw to target object away from
From information.Action detection device sends the teaching collecting behavior signal and distance signal to dsp controller by I/O port,
After the signal collecting is processed by dsp controller, obtain the status information of teaching behavior.
(2) mechanical arm system determines original state
For electricity on 51 single-chip minimum systems and steering engine controller, that is, on mechanical arm, steering wheel is in "on" position.System
First passing through single-chip microcomputer makes mechanical arm revert to residing attitude before teach-by-doing teaching, and then system is searched by the sensor carrying
Seek the orientation determining target object, the ultrasonic sensor that base steering wheel carries carries out rotation and determines target object place side
To the steering wheel in paw direction carries out rotating makes its infrared distance sensor carrying keep just to direction, machinery with target object
Arm keeps this attitude.
(3) mechanical arm system learning by imitation teaching behavior
Mechanical arm system enter the learning by imitation stage, build feedforward neural network, using the status information of teaching behavior as
The input of network, the action policy of manipulator motion is as the output of network.The action policy of acquisition is sent to by dsp controller
Joint angle is converted into pulse width signal with control machinery arm by Single-chip Controlling steering engine controller by 51 single-chip minimum systems
Articulation respective angles, make paw move to target object direction.Gathered by infrared distance sensor after manipulator motion
The range information of paw and target object simultaneously sends dsp controller to, if paw or is detected not more than 0cm with target object distance
To target object, then proceed learning by imitation, reset action strategy, produce servos control signal, circular flow, until
Meet the imitation termination condition (infrared distance sensor detects paw and target object distance as 0cm) setting, paw closes,
Gripping is detected whether to target object by the touch sensor of mechanical arm paw, thus completing teaching behavior act.
Compared with prior art, the present invention has advantages below：
(1) present invention adopts the independently-powered mode of module, is sensing module and performing module is independently powered, in this base
Using a kind of method of improved teach-by-doing off-line teaching on plinth, this method can overcome joint during teach-by-doing teaching
Counter-force effect and the shortcoming of high labor intensive, thus completing teaching task well, improve teaching efficiency.
(2) present invention is mounted in machine by a kind of by the action detection device that gyro sensor and accelerometer module form
On tool arm, as the behavior state information in the sensing module collection machinery arm motor process of system, then should to these information
With learning algorithm, instruct mechanical arm system learning by imitation.Device and side using this collection machinery arm motor behavior information
Method, can reduce the cost of sensor, overcome the image processing process of photographic head technology later stage complexity simultaneously, improve machinery
The efficiency of arm system learning by imitation.
(3) control system of the present invention takes the method that hierarchical classification controls, and carries out control activity relatively independently, passes through
Dsp controller carries out the process of data and the training of learning by imitation algorithm to sensing module collection, by 51 single-chip microcomputers
Mini system run action strategy, the motion of control machinery arm, the method improves specific aim and the information transmission effect of control system
Rate.
(4) learning by imitation mechanism is applied in mechanical arm system the present invention, has the mechanical arm system of learning by imitation mechanism
System not only can reproduce teaching action additionally it is possible to know the purpose of teaching behavior by study, is changing mechanical arm initial attitude
Or still can complete imitation task during the orientation of target object place, there is higher degree of intelligence.
Brief description
Fig. 1 .1 is the mechanical construction drawing of system involved in the present invention；
Fig. 1 .2 is the overall structure figure of system involved in the present invention；
Fig. 2 is the block diagram of system of system involved in the present invention；
Fig. 3 is system operation flow chart involved in the present invention；
Fig. 4 is the method flow diagram that teach-by-doing teaching and training data are processed；
Fig. 5 determines the method flow diagram of original state for mechanical arm system；
Fig. 6 be mechanical arm unite learning by imitation teaching behavior method flow diagram；
Fig. 7 by system learning by imitation algorithm used feedforward neural network；
Fig. 8 .1 is system body front view involved in the present invention；
Fig. 8 .2 is system top view involved in the present invention；
Fig. 8 .3 is system right view involved in the present invention；
In figure：1- mechanical arm system chassis, 2-DSP controller, 3-51 single-chip minimum system, 4- steering engine controller, 5-
7.4v lithium battery, 6- voltage converter, the long U type connecting rod of 7- one, the long U type connecting rod of 8- two, 9-H type connecting rod, the long U type connecting rod of 10-,
11- mechanical arm paw, 12-MG996R steering wheel one, 13-MG996R steering wheel two, 14-MG996R steering wheel three, 15-MG996R steering wheel
Four, 16-MG996R steering wheel five, 17-MG996R steering wheel six, 18- ultrasonic sensor, 19- three-axis gyroscope sensor one, 20-
Three-axis gyroscope sensor two, 21- three-axis gyroscope sensor three, 22- accelerometer module one, 23- accelerometer module two,
24- accelerometer module three, 25- infrared distance sensor, 26- touch sensor one, 27- touch sensor two.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.
Fig. 1 .1-1.2 is the structure chart of the teach-by-doing teaching mechanical arm system with learning by imitation mechanism.Including a bottom
Disk, the arm segment being made up of robot linkage, paw, steering wheel, controller and battery, due to mechanical arm entirety inertia relatively
Greatly, in order to make, mechanical arm system is stable in motor process not to be tilted, and controller and battery is fixed on chassis and connects arm segment
Opposite side with keep balance.
Fig. 2 is the composition frame chart of mechanical arm system.Mechanical arm system is made up of four parts, is sensing module, study respectively
Module, power module and performing module.Sensing module comprises three-axis gyroscope sensor, accelerometer, infrared distance measurement sensing
Device, ultrasonic sensor and touch sensor, study module is dsp controller, and power module comprises lithium battery and voltage turns
Change device, performing module comprises 51 single-chip minimum systems, steering engine controller and six steering wheels.
Dsp controller is arranged on the chassis of mechanical arm, is the core of control system.The main process task chip of dsp controller
Use the TMS320F28335 chip of TI company, the speed of service is 150MIPS.The D/A of the integrated 8bit-SPI of this controller
Conversion, 4 tunnel outputs；16 road AD input interfaces；2 road UART serial line interfaces, a road is RS232 interface；16 road PWM outputs, permissible
Meet the operation of the multi-sensor data process needed for the system and learning by imitation algorithm, will with the control reaching mechanical arm system
Ask.
51 single-chip minimum systems are arranged on dsp controller as double-decker, as output control in performing module
The controller of system strategy.The minimum system that the STC89C52 single-chip microcomputer being produced by STC Corporation is constituted, specification is 9cm × 7.6cm,
32 I/O ports of this system integration, 3 16 bit timing devices, and at power supply, adopt tantalum capacitor design, job stability is strong, meets conduct
Slave computer is controlling the requirement of steering engine controller.
Steering engine controller is arranged on dsp controller as double-decker, as in performing module directly control steering wheel
The executor rotating.32 road steering engine controllers at most can control 32 steering wheels simultaneously, fully meets multi-degree-of-freemechanical mechanical arm
Demand, wherein 8 tunnels carry overload protection, and this controller can be communicated with 51 single-chip microcomputers to realize Based Intelligent Control, and carries
PS2 communication interface, it is possible to achieve handle operates, abundant teaching form.
Steering wheel, as the joint connecting each connecting rod of mechanical arm, installs six from base to paw altogether.Steering wheel uses
MG996R simulates steering wheel, a size of 40.7mm × 19.7mm × 42.9mm, and maximum rotation angle is 180 °, and running voltage is
4.8V-7.2V, moment of torsion is 15kg × cm it is adaptable to joint of mechanical arm steering wheel.
Ultrasonic sensor is arranged on mechanical arm base connecting rod, for detecting the direction that target object is located, base rudder
Machine carries out 180 ° of rotations, and when ultrasonic sensor detects dead ahead and there is target object, steering wheel stops the rotation.Selected sensing
The detection range of device is 2cm-450cm, and precision up to 0.3cm, is set out range finding using IO, the square wave of 8 40kHz of transmission automatically；
Whether automatic detection has signal to return；There is signal to return, a high level is exported by IO, high level duration is exactly super
Sound wave is from the time being transmitted into return.Final test is apart from the S=high level time × velocity of sound/2.
It is whole that three three-axis gyroscope sensors and three accelerometer modules are separately mounted to three execution in the middle of mechanical arm
On the connecting rod of body action, detect angular velocity and the angular acceleration of three connecting rod pose coordinates respectively, but gyroscope can produce drift
Shift error, and As time goes on can add up and become big, error can be made to become very big by integration, therefore can not individually make
Detect the sensor of teaching process with gyroscope as the system, so cooperation accelerometer module uses, joint-detection machinery
The movement angle of arm link.Three-axis gyroscope sensor main chip adopts L3G4200D, and running voltage is 3-5v, and measurement range is
250/500/2000dps, communication mode adopts IIC/SPI communication protocol.Accelerometer module master chip adopts MPU6050, work
Make voltage be 3-6v, a size of 15.24mm × 15.24mm × 2mm, measurement measuring range acceleration be ± 16g, angular velocity for ±
2000 °/s, communication mode adopts IIC/TTL communication protocol.
Infrared distance sensor is arranged on the steering wheel controlling paw direction, the distance of detection paw to target object.Red
Outer distance measuring sensor is using the GP2Y0A02YK0F infrared ray sensor of Japanese SHARP, analog signal output.The material of testee
Matter, ambient temperature and time of measuring do not interfere with the certainty of measurement of sensor.Sensor output voltage value corresponds to detection range,
Finding range is 20-150cm, running voltage 4.5v-5.5v.
Touch sensor is arranged at the gripping of mechanical arm paw, and whether detection paw grips object.Touch sensor
Using MEAS film piezo-electric sensor, for elasticity, tactile, the measurement of vibration and impact.Can produce when thin film sweeping action
That big voltage of little exchange (up to ± 90v).
Power module is arranged on the lower section of dsp controller, is the lithium battery of 7.4V using one piece of output voltage, and capacity is
2200mah, weight 116g, a size of 105mm × 33.5mm × 15mm.One electric pressure converter of configuration, 7.4V is converted into 5V
And 6.5V.
Mechanical arm system running flow chart of the present invention is as shown in Figure 3.First mechanical arm is carried out showing by doing and illustrating
Religion, is gathered teaching behavioural information and processed by sensing module, then determine mechanical arm initial shape according to target object location
State, system carries out learning by imitation and reproduces teaching behavior act.The method of described mechanical arm system learning by imitation includes following step
Suddenly：
Step 1, teach-by-doing teaching and training data are processed.
This stage power module is only dsp controller (2), three-axis gyroscope sensor one (19), three-axis gyroscope sensing
Device two (20), three-axis gyroscope sensor three (21) and accelerometer module one (22), accelerometer module two (23), acceleration
Meter module three (24) power supply, other components and parts are in off-position.Mechanical arm is taken with teach-by-doing off-line teaching mode, dragging machine
Tool arm grips target object, gathers teaching behavioural information by sensing module.Teach-by-doing teaching and the flow chart of training data process
As shown in figure 4, specifically including following steps：
Step 1.1, initializes mechanical arm system.
Configuration dsp controller 2 buffer status, initialize global and local variable, the position of initializing sensor and rudder
The controlled quentity controlled variable of machine.Power module is only dsp controller (2), three-axis gyroscope sensor and accelerometer module for power supply, Qi Tayuan
Device is in off-position.
Step 1.2, to mechanical arm by the way of teach-by-doing off-line teaching, pulling mechanical arm makes mechanical arm paw (11) press from both sides
Live with mechanical arm system chassis (1) target object at grade, keep paw during dragging to target object.Here mistake
Cheng Zhong, gathers place connecting rod pose coordinate by the action detection device that three-axis gyroscope sensor and accelerometer module form
Angular velocity and angular acceleration, are collected the distance of target object by infrared distance sensor.For ensureing the quality of teaching, make sensing
Device can accurately gather signal, pulls mechanical arm paw gripping target object by doing and illustrating with more slow speed, whole used
Time control is in 5s-10s, and takes 3-5 teaching behavior.
Step 1.3, the pose collecting coordinate angular velocity and angular acceleration are carried out data processing simultaneously by dsp controller (2)
Merged by adaptive weight fusion estimated algorithm, the distance collecting is changed into paw to target object distance, thus
To teaching behavior state information.
(1) data processing of three-axis gyroscope and accelerometer.
Gyroscope is used for measuring connecting rod angular velocity signal, is integrated by angular velocity, just can obtain angle value.Each three axle
The every 1ms of gyro sensor gathers a data, and every 10 values carry out a data processing, because the time is very short, ignore the time
The error that difference is brought, the angle calculation that three-axis gyroscope sensor obtains is：
angleAn=angleAn-1+gyron×dt
Wherein angleAnThe angle value sampling for gyroscope n-th, angleAn-1Sample for (n-1)th time for gyroscope
Angle value, gyronThe intermittent angle rate value sampling for gyroscope n-th, dt is the time used by teach-by-doing teaching process.
Accelerometer is used for measuring the linear acceleration of connecting rod, and the output valve of accelerometer and inclination angle are in non-linear relation,
Show as varies with sinusoidal function with the increase at inclination angle, with the method for three-axis gyroscope sensor acquisition data, accelerometer
The angle calculation that module obtains is：
AngleB=atan2 (y, z) × (180/3.14)
Defined in it, down, y-axis is forward for accelerometer module x-axis.The angle that angleB obtains for accelerometer, atan2
(y, z) is the radian of this inclination angle vertical direction.
(2) carry out data fusion using adaptive weight fusion estimated algorithm.
Take the signal angleA that three-axis gyroscope sensor acquisition arrivesnThe signal collecting with accelerometer module
AngleB, carries out data fusion using adaptive weight fusion estimated algorithm method, and it is without any priori of sensor measurement data
Knowledge, you can merge and the minimum data fusion value of variance, the variance after estimation is less than the variance of single sensor estimation and adopts
Averagely do the variance estimated with multisensor average.True angled state information s of each connecting rod of mechanical arm systemDComputing formula is such as
Under：
sD=W1×angleAn+W2×angleB
Wherein, W1For the optimal weighted factor of three-axis gyroscope sensor, W2For accelerometer module optimal weighting because
Son, therefore teaching behavior state information sD=(sD1,sD2,...,sDn).
(3) paw is to target object distance conversion.
If it is d that infrared distance sensor (25) detects with target object distance, because sensor emission mouth is apart from paw
Gripping center be 7cm, therefore actual paw to target object distance be L=d-7cm.The teaching of mechanical arm system and teaching number
Terminate according to the stage processing.
Step 2, mechanical arm system determines original state.
Ultrasonic sensor (18) and infrared distance sensor (25) searching target object institute that this stage is carried by system
The position at place, determines the initial attitude of mechanical arm, makes system when changing mechanical arm initial attitude or target object place orientation,
Task still can be completed.Mechanical arm system determines the flow chart of original state as shown in figure 5, specifically including following steps：
Step 2.1, initializes mechanical arm system.
Configuration dsp controller (2) and the buffer status of 51 single-chip minimum systems (3), initialization global and local becomes
Amount, the controlled quentity controlled variable of the position of initializing sensor and steering wheel and steering engine controller (4).
Step 2.2, mechanical arm reverts to initial attitude.
51 single-chip minimum systems (3) pass through control program, the initial attitude before making mechanical arm revert to teach-by-doing teaching,
Made mechanical arm before learning by imitation teaching behavior, identical pose coordinate before each connecting rod holding and teach-by-doing teaching, to guarantee
Even if changing mechanical arm initial attitude system to remain to complete learning by imitation task.
Step 2.3, the determination in target object direction.
51 single-chip minimum systems (3) pass through control program, make steering wheel (12) carry out from left to right 180 ° rotation, search with
System is in the target object of same desktop, when the ultrasonic sensor (18) on the connecting rod of steering wheel (12) place detects dead ahead
Target object when, steering wheel (12) stops operating and is fixed on current location it is determined that target object place direction, to guarantee to be
System during change target object place orientation is made to remain to complete learning by imitation task.
The determination in step 2.4, paw and target object direction.
Because infrared distance sensor (25) is arranged on and mechanical arm paw parallel position, infrared emitting direction is paw
Pawl point direction, therefore 51 single-chip minimum systems (3) pass through control program, so that steering wheel (15) is rotated down and pass until infrared distance measurement
Just to target object, that is, paw and target object are in just to direction sensor, and mechanical arm system determines the stage of original state
Terminate.
Step 3, mechanical arm system learning by imitation teaching behavior.
This stage learns by imitation algorithm by dsp controller (2) acquisition action policy, 51 single-chip minimum systems (3)
Moved by action policy control machinery arm, imitated teaching behavior act, by infrared distance sensor (25) judgement be
No complete imitation task, circular flow, until infrared distance sensor detects after paw reaches target object, mechanical arm paw
Closure, completes the imitation of teaching behavior.The flow chart of mechanical arm system system learning by imitation teaching behavior is as shown in fig. 6, concrete
Comprise the following steps：
Step 3.1, initializes mechanical arm system.
Configuration dsp controller (2) and the buffer status of 51 single-chip minimum systems (3), initialization global and local becomes
Amount, the controlled quentity controlled variable of the position of initializing sensor and steering wheel and steering engine controller (4).
Step 3.2, uses learning by imitation algorithm to mechanical arm system, is obtained using the method for feedforward neural network and controls plan
Slightly, such as Fig. 7 builds a BP neural network.With the connecting rod moment shape being gathered in teaching behavior by gyroscope and accelerometer
State sD=(sD1,sD2,...,sDn) as neutral net input layer, input layer has n neuron；With h=(h1,h2,...,
hp) for hidden layer output, the connection weight of input layer and hidden layer is wj=(wj1,wj2,...,wjp), hidden layer neuron threshold
It is worth for bj=(bj1,bj2,...,bjp), hidden layer transmission function is f1(), hidden layer has p neuron；With the corresponding moment
Action policy a=(a1,a2,...,aq) as neutral net output layer, the connection weight of hidden layer and output layer is wk=
(wk1,wk2,...,wkq), output layer neuron threshold value is bk=(bk1,bk2,...,bkq), output layer transmission function is f2(),
Output layer has q neuron.
(1) choose t input sample, that is, the state of t connecting rod is as input sD(t)=(sD1(t),sD2(t),...,
sDn(t)), according to BP neural network principle, using FR conjugate gradient method training network, network hidden layer can be obtained and be output as：
(2) error partial derivative δ (t) of each neuron of output layer and output h (t) of each neuron of network hidden layer is utilized
Revise connection weight wk(t)：
Wherein wk N+1For the connection weight after updating, wk NFor the connection weight before updating, η is learning rate.
(3) action policy finally giving neutral net output t is expressed as：
Step 3.3, training is obtained action policy a (t) and is transferred to single-chip minimum system, by single-chip microcomputer by dsp controller
Control steering engine controller so that mechanical arm is moved, imitate paw to the teaching behavior act of target object.Mechanical arm has executed
After action, dsp controller with the range information of target object and is sent to by infrared distance sensor detection, that is, detection paw is
No arrival target object.If infrared distance sensor detects paw and more than 0cm or is not detected by mesh with target object distance
Mark object, illustrates that the learning by imitation action of Single Chip Microcomputer (SCM) system execution fails to make paw reach target object or deviate target object,
Then this action policy failure, system comes back to step 3.2 and proceeds the new action policy of learning by imitation acquisition, until monolithic
Machine minimum system passes through to meet the action policy a'(t setting imitation termination condition) make mechanical arm paw reach object position
Put, that is, infrared sensor detect paw to target object range information L=0cm when, mechanical arm stop motion.
Step 3.4, starts to rotate now by Single-chip Controlling steering wheel (17), and rotation direction is so that mechanical arm paw is closed
Direction, illustrates handss when being attached to when paw grips the touch sensor one (26) of position, touch sensor two (27) detects signal
Pawl has clamped target object, and steering wheel (17) stops operating, and mechanical arm system realizes the gripping to target object, so far completes
Imitation to whole teaching behaviors, the stage of mechanical arm system learning by imitation teaching behavior terminates, thus whole mechanical arm system
End-of-job.
Claims (2)
1. there is the teach-by-doing teaching mechanical arm system of learning by imitation mechanism it is characterised in that：In this system, dsp controller
(2), 7.4v lithium battery (5), voltage converter (6), long U type connecting rod one (7) are fixedly mounted on mechanical arm system chassis (1)；
51 single-chip minimum systems (3), steering engine controller (4) are connected with dsp controller (2) respectively；Described voltage converter (6) with
7.4v lithium battery (5) connects, and 7.4v lithium battery (5) is connected with dsp controller (2)；The side of long U type connecting rod one (7) is provided with
Long U type connecting rod two (8)；One end of long U type connecting rod two (8) is provided with ultrasonic sensor (18)；
The bottom of two long U type connecting rods connects composition H type connecting rod (9)；H type connecting rod (9) one end and the side of long U type connecting rod two (8)
Face connects；Long U type connecting rod three (10) one end is connected with H type connecting rod (9) other end；Mechanical arm paw (11) and long U type connecting rod three
(10) the other end connects；
MG996R steering wheel one (12) is arranged on the junction of long U type connecting rod two (8) and long U type connecting rod one (7)；
MG996R steering wheel two (13) is arranged on the side junction of H type connecting rod (9) and long U type connecting rod two (8)；
MG996R steering wheel three (14) is arranged on one end junction of H type connecting rod (9) and long U type connecting rod three (10)；
MG996R steering wheel four (15) is arranged at the other end of long U type connecting rod three (10)；
MG996R steering wheel five (16) is arranged on the end of mechanical arm paw (11)；
MG996R steering wheel six (17) is arranged on the side of mechanical arm paw (11)；
Three-axis gyroscope sensor one (19) forms detection module one with accelerometer module one (22), and detection module one is arranged on
H type connecting rod (9) middle position；
Three-axis gyroscope sensor two (20) forms detection module two with accelerometer module two (23), and detection module two is arranged on
Long U type connecting rod three (10) middle position；
Three-axis gyroscope sensor three (21) forms detection module three with accelerometer module three (24), and detection module three is arranged on
One end of mechanical arm paw (11)；
Infrared distance sensor (25) is arranged on the centre position of MG996R steering wheel six (17)；
Touch sensor one (26), touch sensor two (27) are separately positioned at the paw up and down of mechanical arm paw (11)；
There is the teach-by-doing teaching mechanical arm system method of learning by imitation mechanism, during mechanical arm system work, initially with handling
Handss teaching method carries out teaching to mechanical arm, carries out learning by imitation by after sensing module collection teaching information through control system,
Control machinery arm imitates teaching behavior act；
(1) teach-by-doing teaching and training data are processed
After mechanical arm system starts, the steering engine controller of system does not go up electricity, and that is, on mechanical arm, steering wheel is in off-position, adopts
With the mode of off-line teaching, only sensing module is operated；With handss driving machinery arm complete grip desktop on object action, three
The action detection device collection of individual three-axis gyroscope sensor and accelerometer module composition three arms during teaching connect
The pose coordinate angular velocity of rotation of bar and angular acceleration information, infrared distance sensor gathers paw to the distance letter of target object
Breath；Action detection device sends the teaching collecting behavior signal and distance signal to dsp controller by I/O port, and DSP is controlled
After the signal collecting is processed by device processed, obtain the status information of teaching behavior；
(2) mechanical arm system determines original state
For electricity on 51 single-chip minimum systems and steering engine controller, that is, on mechanical arm, steering wheel is in "on" position；System is first
Mechanical arm is made to revert to residing attitude before teach-by-doing teaching by single-chip microcomputer, the sensor that then system is passed through to carry is searched really
Set the goal the orientation of object, and the ultrasonic sensor that base steering wheel carries carries out rotation and determines target object place direction, handss
The steering wheel in pawl direction carries out rotating makes its infrared distance sensor carrying keep just to direction with target object, and mechanical arm keeps
This attitude；
(3) mechanical arm system learning by imitation teaching behavior
Mechanical arm system enters the learning by imitation stage, builds feedforward neural network, using the status information of teaching behavior as network
Input, the action policy of manipulator motion is as the output of network；It is single that dsp controller sends the action policy of acquisition to 51
Joint angle is converted into pulse width signal with the joint on control machinery arm by Single-chip Controlling steering engine controller by piece machine minimum system
Rotate respective angles, so that paw is moved to target object direction；After manipulator motion, paw is gathered by infrared distance sensor
With the range information of target object and send dsp controller to, if paw and target object distance more than 0cm or can't detect mesh
Mark object, then proceed learning by imitation, reset action strategy, produces servos control signal, circular flow, until meeting
The imitation termination condition setting is that infrared distance sensor detects paw and target object distance as 0cm, and paw closes, by machine
Touch sensor at tool arm paw detects whether that target object is arrived in gripping, thus completing teaching behavior act.
2. the teach-by-doing teaching mechanical arm system with learning by imitation mechanism according to claim 1 it is characterised in that：First
First teach-by-doing teaching is carried out to mechanical arm, gathered teaching behavioural information and processed by sensing module, then according to object
Body position determines mechanical arm original state, and system carries out learning by imitation and reproduces teaching behavior act；Described mechanical arm system mould
The method of imitative study comprises the following steps：
Step 1, teach-by-doing teaching and training data are processed；
This stage power module is only dsp controller (2), three-axis gyroscope sensor one (19), three-axis gyroscope sensor two
(20), three-axis gyroscope sensor three (21) and accelerometer module one (22), accelerometer module two (23), accelerometer mould
Block three (24) is powered, and other components and parts are in off-position；Mechanical arm is taken with teach-by-doing off-line teaching mode, pulls mechanical arm
Gripping target object, gathers teaching behavioural information by sensing module；Specifically include following steps：
Step 1.1, initializes mechanical arm system；
Configuration dsp controller (2) buffer status, initialize global and local variable, the position of initializing sensor and steering wheel
Controlled quentity controlled variable；Power module is only dsp controller (2), three-axis gyroscope sensor and accelerometer module for power supply, other units device
Part is in off-position；
Step 1.2, to mechanical arm by the way of teach-by-doing off-line teaching, pull mechanical arm make mechanical arm paw (11) clamp with
Mechanical arm system chassis (1) target object at grade, keeps paw to target object during dragging；In this process
In, the angle of place connecting rod pose coordinate is gathered by the action detection device that three-axis gyroscope sensor and accelerometer module form
Speed and angular acceleration, are collected the distance of target object by infrared distance sensor；For ensureing the quality of teaching, make sensor
Signal can accurately be gathered, mechanical arm paw gripping target object, whole institute's used time are pulled by doing and illustrating with more slow speed
Between control in 5s-10s, and take 3-5 teaching behavior；
Step 1.3, the pose collecting coordinate angular velocity and angular acceleration are carried out data processing and pass through by dsp controller (2)
Adaptive weight fusion estimated algorithm is merged, and the distance collecting is changed into paw to target object distance, thus being shown
Religion behavior state information；
(1) data processing of three-axis gyroscope and accelerometer；
Gyroscope is used for measuring connecting rod angular velocity signal, is integrated by angular velocity, just can obtain angle value；Each three axis accelerometer
The every 1ms of instrument sensor gathers a data, and every 10 values carry out a data processing, because the time is very short, ignore time difference
The error brought, the angle calculation that three-axis gyroscope sensor obtains is：
angleAn=angleAn-1+gyron×dt
Wherein angleAnThe angle value sampling for gyroscope n-th, angleAn-1The angle sampling for (n-1)th time for gyroscope
Value, gyronThe intermittent angle rate value sampling for gyroscope n-th, dt is the time used by teach-by-doing teaching process；
Accelerometer is used for measuring the linear acceleration of connecting rod, and the output valve of accelerometer and inclination angle are in non-linear relation, with
The increase at inclination angle and show as varies with sinusoidal function, with the method for three-axis gyroscope sensor acquisition data, accelerometer module
The angle calculation obtaining is：
AngleB=atan2 (y, z) × (180/3.14)
Defined in it, down, y-axis is forward for accelerometer module x-axis；The angle that angleB obtains for accelerometer, atan2 (y, z)
Radian for this inclination angle vertical direction；
(2) carry out data fusion using adaptive weight fusion estimated algorithm；
Take the signal angleA that three-axis gyroscope sensor acquisition arrivesnThe signal angleB collecting with accelerometer module, adopts
Adaptive weight fusion estimated algorithm method carries out data fusion, and it is without any priori of sensor measurement data, you can melt
Close out the minimum data fusion value of variance, the variance after estimation is less than the variance of single sensor estimation and adopts multisensor equal
Value averagely does the variance estimated；True angled state information s of each connecting rod of mechanical arm systemDComputing formula is as follows：
sD=W1×angleAn+W2×angleB
Wherein, W1For the optimal weighted factor of three-axis gyroscope sensor, W2For the optimal weighted factor of accelerometer module, therefore
Teaching behavior state information sD=(sD1,sD2,...,sDn)；
(3) paw is to target object distance conversion；
If it is d that infrared distance sensor (25) detects with target object distance, because sensor emission mouth grips apart from paw
Center be 7cm, therefore actual paw to target object distance be L=d-7cm；At the teaching of mechanical arm system and training data
The stage of reason terminates；
Step 2, mechanical arm system determines original state；
Residing for ultrasonic sensor (18) that this stage is carried by system and infrared distance sensor (25) searching target object
Position, determines the initial attitude of mechanical arm, makes system when changing mechanical arm initial attitude or target object place orientation, still
Task can be completed；
Specifically include following steps：
Step 2.1, initializes mechanical arm system；
Configuration dsp controller (2) and the buffer status of 51 single-chip minimum systems (3), initialize global and local variable, just
The position of beginningization sensor and the controlled quentity controlled variable of steering wheel and steering engine controller 4；
Step 2.2, mechanical arm reverts to initial attitude；
51 single-chip minimum systems (3) pass through control program, the initial attitude before making mechanical arm revert to teach-by-doing teaching, make machine
, before learning by imitation teaching behavior, identical pose coordinate before each connecting rod holding and teach-by-doing teaching, even if to guarantee for tool arm
Change mechanical arm initial attitude system to remain to complete learning by imitation task；
Step 2.3, the determination in target object direction；
51 single-chip minimum systems (3) pass through control program, make MG996R steering wheel one (12) carry out 180 ° of rotations from left to right, search
Seek the target object being in same desktop with system, the ultrasonic sensor (18) on MG996R steering wheel one (12) place connecting rod
Detect dead ahead target object when, MG996R steering wheel one (12) stops operating and is fixed on current location it is determined that target
Object place direction, even if remain to complete learning by imitation task with system when guaranteeing to change target object place orientation；
The determination in step 2.4, paw and target object direction；
Because infrared distance sensor (25) is arranged on and mechanical arm paw parallel position, infrared emitting direction is paw pawl point
Direction, therefore 51 single-chip minimum systems (3) pass through control program, so that MG996R steering wheel four (15) is rotated down until infrared survey
Away from sensor just to target object, that is, paw and target object are in just to direction, and mechanical arm system determines original state
Stage terminates；
Step 3, mechanical arm system learning by imitation teaching behavior；
This stage learns by imitation algorithm by dsp controller (2) acquisition action policy, and 51 single-chip minimum systems (3) pass through
Action policy control machinery arm is moved, and imitates teaching behavior act, has been determined whether by infrared distance sensor (25)
Become imitation task, circular flow, after infrared distance sensor detects paw arrival target object, mechanical arm paw closes
Close, complete the imitation of teaching behavior；Specifically include following steps：
Step 3.1, initializes mechanical arm system；
Configuration dsp controller (2) and the buffer status of 51 single-chip minimum systems (3), initialize global and local variable, just
The position of beginningization sensor and the controlled quentity controlled variable of steering wheel and steering engine controller (4)；
Step 3.2, uses learning by imitation algorithm to mechanical arm system, and the method using feedforward neural network obtains control strategy,
With connecting rod moment state s being gathered in teaching behavior by gyroscope and accelerometerD=(sD1,sD2,...,sDn) as god
Input layer through network, input layer has n neuron；With h=(h1,h2,...,hp) for hidden layer output, input layer with hidden
Connection weight containing layer is wj=(wj1,wj2,...,wjp), hidden layer neuron threshold value is bj=(bj1,bj2,...,bjp), imply
Layer transmission function is f1(), hidden layer has p neuron；Action policy a=(a with the corresponding moment1,a2,...,aq) conduct
The output layer of neutral net, hidden layer is w with the connection weight of output layerk=(wk1,wk2,...,wkq), output layer neuron threshold
It is worth for bk=(bk1,bk2,...,bkq), output layer transmission function is f2(), output layer has q neuron；
(1) choose t input sample, that is, the state of t connecting rod is as input sD(t)=(sD1(t),sD2(t),...,sDn
(t)), according to BP neural network principle, using FR conjugate gradient method training network, network hidden layer can be obtained and be output as：
(2) error partial derivative δ (t) of each neuron of output layer and output h (t) of each neuron of network hidden layer is utilized to revise
Connection weight wk(t)：
Wherein wk N+1For the connection weight after updating, wk NFor the connection weight before updating, η is learning rate；
(3) action policy finally giving neutral net output t is expressed as：
Step 3.3, training is obtained action policy a (t) and is transferred to single-chip minimum system, by Single-chip Controlling by dsp controller
Steering engine controller makes mechanical arm be moved, and imitates paw to the teaching behavior act of target object；Mechanical arm has executed action
Afterwards, dsp controller with the range information of target object and is sent to by infrared distance sensor detection, that is, whether detection paw arrives
Reach target object；If infrared distance sensor detects paw and more than 0cm or is not detected by object with target object distance
Body, illustrates that the learning by imitation action of Single Chip Microcomputer (SCM) system execution fails to make paw to reach target object or deviate target object, then this
Action policy failure, system comes back to step 3.2 and proceeds learning by imitation and obtains new action policy, until single-chip microcomputer
Mini system passes through to meet the action policy a'(t setting imitation termination condition) make mechanical arm paw reach target object location, that is,
Infrared sensor detect paw to target object range information L=0cm when, mechanical arm stop motion；
Step 3.4, starts to rotate now by Single-chip Controlling steering wheel 17, and rotation direction is to make mechanical arm paw closing direction,
Paw is described when being attached to when paw grips the touch sensor one (26) of position, touch sensor two (27) detects signal
Clamp target object, MG996R steering wheel six (17) stops operating, mechanical arm system realizes the gripping to target object, so far completes
Imitation to whole teaching behaviors, stage of mechanical arm system learning by imitation teaching behavior terminates, thus whole mechanical arm system
System end-of-job.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510244111.1A CN104924313B (en) | 2015-05-13 | 2015-05-13 | There is teach-by-doing teaching mechanical arm system and the method for learning by imitation mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510244111.1A CN104924313B (en) | 2015-05-13 | 2015-05-13 | There is teach-by-doing teaching mechanical arm system and the method for learning by imitation mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104924313A CN104924313A (en) | 2015-09-23 |
| CN104924313B true CN104924313B (en) | 2017-03-01 |
Family
ID=54111893
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510244111.1A Active CN104924313B (en) | 2015-05-13 | 2015-05-13 | There is teach-by-doing teaching mechanical arm system and the method for learning by imitation mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104924313B (en) |
Families Citing this family (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105700481A (en) * | 2016-03-23 | 2016-06-22 | 北京光年无限科技有限公司 | Intelligent robot motion generation method and system |
| CN105751238A (en) * | 2016-05-06 | 2016-07-13 | 马宁 | Adjustable mechanical arm |
| CN106112952B (en) * | 2016-06-27 | 2018-09-04 | 北京工业大学 | Both arms installation transfer robot |
| CN106272432B (en) * | 2016-09-28 | 2019-03-01 | 哈尔滨云控机器人科技有限公司 | A kind of six degree of freedom manipulator |
| CN106378769B (en) * | 2016-10-14 | 2018-11-09 | 昆明理工大学 | A kind of remote control manipulator |
| CN107170345B (en) * | 2017-04-11 | 2019-07-19 | 广东工业大学 | The teaching method and device based on machine vision and gyroscope towards industrial robot |
| JP6546618B2 (en) * | 2017-05-31 | 2019-07-17 | 株式会社Ｐｒｅｆｅｒｒｅｄ Ｎｅｔｗｏｒｋｓ | Learning apparatus, learning method, learning model, detection apparatus and gripping system |
| CN107336241B (en) * | 2017-08-08 | 2020-10-16 | 智造未来（北京）机器人系统技术有限公司 | Manipulator control device, manipulator control method, and manipulator device |
| CN107331279A (en) * | 2017-08-16 | 2017-11-07 | 嘉兴锐视智能科技有限公司 | Teaching apparatus and system |
| CN107717983A (en) * | 2017-08-21 | 2018-02-23 | 北京精密机电控制设备研究所 | A kind of apery mechanical arm teaching control method |
| CN108145686B (en) * | 2017-12-07 | 2023-12-15 | 浙江理工大学 | Flip type cylindrical material intelligent logistics trolley |
| CN109965472A (en) * | 2017-12-28 | 2019-07-05 | 沈阳新松机器人自动化股份有限公司 | A kind of walk supporting device |
| CN108563160A (en) * | 2018-01-23 | 2018-09-21 | 济南大学 | A kind of Remote Experimental Platform of diminution ratio model cootrol |
| CN108255107A (en) * | 2018-01-23 | 2018-07-06 | 济南大学 | A kind of Flexible Manufacture platform of low development cost |
| CN110298877A (en) * | 2018-03-21 | 2019-10-01 | 北京猎户星空科技有限公司 | A kind of the determination method, apparatus and electronic equipment of object dimensional pose |
| CN108376509A (en) * | 2018-04-09 | 2018-08-07 | 重庆鲁班机器人技术研究院有限公司 | Teaching mechanical arm system and its control method |
| CN108538179A (en) * | 2018-04-13 | 2018-09-14 | 湖南城市学院 | A kind of intelligent machine manufacturing operator production training simulation system |
| CN108896658B (en) * | 2018-05-14 | 2021-01-26 | 湖南湖大艾盛汽车技术开发有限公司 | Ultrasonic automatic detection method based on PLC |
| CN109159115B (en) * | 2018-08-17 | 2020-11-13 | 哈尔滨工业大学 | Science popularization-oriented tactile communication device |
| CN109050915B (en) * | 2018-08-30 | 2022-07-01 | 徐州工程学院 | Mechanical and electrical integration system for arms of flying robot |
| CN109093599A (en) * | 2018-09-07 | 2018-12-28 | 中科新松有限公司 | A kind of crawl teaching method and system |
| CN109500815B (en) * | 2018-12-03 | 2023-06-02 | 日照市越疆智能科技有限公司 | Robot for front gesture judgment learning |
| CN109760050A (en) * | 2019-01-12 | 2019-05-17 | 鲁班嫡系机器人（深圳）有限公司 | Robot behavior training method, device, system, storage medium and equipment |
| CN114080304B (en) * | 2019-08-22 | 2024-05-14 | 欧姆龙株式会社 | Control device, control method, and control program |
| CN110774269B (en) * | 2019-11-26 | 2020-11-20 | 上海节卡机器人科技有限公司 | Robot drag teaching method and device, electronic device and storage medium |
| CN112258929A (en) * | 2020-11-19 | 2021-01-22 | 浙江工业大学 | Remote interactive teaching aid for children |
| CN114589720B (en) * | 2020-12-04 | 2023-10-10 | 国网智能科技股份有限公司 | Intelligent paw capable of automatically identifying and positioning files, method and robot |
| CN115476338B (en) * | 2022-07-08 | 2024-01-26 | 深圳市越疆科技股份有限公司 | Gesture adjustment method and device, robot mechanism, electronic device and storage medium |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102011079117B4 (en) * | 2011-07-14 | 2022-09-29 | Kuka Deutschland Gmbh | Method of programming a robot |
| CN102608351B (en) * | 2012-02-14 | 2014-12-17 | 三一重工股份有限公司 | Detection method and system of three-dimensional gesture of mechanical arm and system controlling mechanical arm to operate |
| WO2013164470A1 (en) * | 2012-05-04 | 2013-11-07 | Leoni Cia Cable Systems Sas | Imitation learning method for a multi-axis manipulator |
| JP6150386B2 (en) * | 2013-04-24 | 2017-06-21 | 国立大学法人横浜国立大学 | Robot teaching method |
| DE102013218823A1 (en) * | 2013-09-19 | 2015-04-02 | Kuka Laboratories Gmbh | Method for manually adjusting the pose of a manipulator arm of an industrial robot and associated industrial robots |
| CN104552230A (en) * | 2013-10-22 | 2015-04-29 | 西华大学 | Book fetching robot for library |
| CN204054037U (en) * | 2014-03-13 | 2014-12-31 | 湖南信息职业技术学院 | A kind of service robot with vision system |
| CN103878772B (en) * | 2014-03-31 | 2017-04-26 | 北京工业大学 | Biomorphic wheeled robot system with simulation learning mechanism and method |
-
2015
- 2015-05-13 CN CN201510244111.1A patent/CN104924313B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN104924313A (en) | 2015-09-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104924313B (en) | There is teach-by-doing teaching mechanical arm system and the method for learning by imitation mechanism | |
| CN103878772B (en) | Biomorphic wheeled robot system with simulation learning mechanism and method | |
| CN111618847B (en) | Mechanical arm autonomous grabbing method based on deep reinforcement learning and dynamic motion elements | |
| CN110977992B (en) | Non-kinematic model trajectory tracking method for mechanical arm and mechanical arm system | |
| CN100573389C (en) | All fours type bionic robot control device | |
| CN107363813A (en) | A kind of desktop industrial robot teaching system and method based on wearable device | |
| Fang et al. | Skill learning for human-robot interaction using wearable device | |
| CN101916071A (en) | CPG feedback control method of biomimetic robot fish movement | |
| CN106945059B (en) | A kind of gesture tracking method based on population random disorder multi-objective genetic algorithm | |
| CN107121256A (en) | A kind of six degree of freedom captive trajectory testing method of continuous in-orbit motion | |
| CN105404744B (en) | A kind of space manipulator total state dynamics semi-physical system | |
| CN106041912A (en) | Master-slave mode snake-like robot system and position control method thereof | |
| CN108284436A (en) | Remote mechanical dual arm system and method with learning by imitation mechanism | |
| CN105259888A (en) | Teaching control system and method for excavator, and excavator | |
| CN107685343A (en) | A kind of Mechanical transmission test parameter calibration configuration optimization method | |
| CN110653801A (en) | Guide control system of robot operating arm and flexible control and teaching learning method thereof | |
| CN110154024A (en) | A kind of assembly control method based on shot and long term Memory Neural Networks incremental model | |
| Li et al. | Neural learning and kalman filtering enhanced teaching by demonstration for a baxter robot | |
| CN104914867A (en) | Hexapod robot autonomous navigation closed-loop controller with fuzzy neural network | |
| WO2021184505A1 (en) | Soft joint gripper based on 4d printing, and consistency control method therefor | |
| Abdulghani et al. | Design and evaluation of a MIMO ANFIS using MATLAB and V-REP | |
| CN112959330A (en) | Robot double-arm motion man-machine corresponding device and method based on master-slave dynamic motion elements | |
| CN107479537A (en) | Control performance assessment system and its process for using for robot controller design | |
| CN201124413Y (en) | Control device of four-footed bionic robot | |
| CN206578829U (en) | A kind of bionical body-sensing mechanical arm of seven freedom |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |