CN115782606A - Small wheel speed self-regulation and control method and system based on pulse neural network - Google Patents

Small wheel speed self-regulation and control method and system based on pulse neural network Download PDF

Info

Publication number
CN115782606A
CN115782606A CN202111062796.XA CN202111062796A CN115782606A CN 115782606 A CN115782606 A CN 115782606A CN 202111062796 A CN202111062796 A CN 202111062796A CN 115782606 A CN115782606 A CN 115782606A
Authority
CN
China
Prior art keywords
neuron
value
pwm
trolley
pwm value
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.)
Pending
Application number
CN202111062796.XA
Other languages
Chinese (zh)
Inventor
高钢
李骁健
王成
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.)
Shenzhen Institute of Advanced Technology of CAS
Original Assignee
Shenzhen Institute of Advanced Technology of CAS
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 Shenzhen Institute of Advanced Technology of CAS filed Critical Shenzhen Institute of Advanced Technology of CAS
Priority to CN202111062796.XA priority Critical patent/CN115782606A/en
Priority to PCT/CN2021/137998 priority patent/WO2023035462A1/en
Publication of CN115782606A publication Critical patent/CN115782606A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60L—PROPULSION OF ELECTRICALLY-PROPELLED VEHICLES; SUPPLYING ELECTRIC POWER FOR AUXILIARY EQUIPMENT OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRODYNAMIC BRAKE SYSTEMS FOR VEHICLES IN GENERAL; MAGNETIC SUSPENSION OR LEVITATION FOR VEHICLES; MONITORING OPERATING VARIABLES OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRIC SAFETY DEVICES FOR ELECTRICALLY-PROPELLED VEHICLES
    • B60L15/00—Methods, circuits, or devices for controlling the traction-motor speed of electrically-propelled vehicles
    • B60L15/20—Methods, circuits, or devices for controlling the traction-motor speed of electrically-propelled vehicles for control of the vehicle or its driving motor to achieve a desired performance, e.g. speed, torque, programmed variation of speed
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B62—LAND VEHICLES FOR TRAVELLING OTHERWISE THAN ON RAILS
    • B62D—MOTOR VEHICLES; TRAILERS
    • B62D11/00—Steering non-deflectable wheels; Steering endless tracks or the like
    • B62D11/02—Steering non-deflectable wheels; Steering endless tracks or the like by differentially driving ground-engaging elements on opposite vehicle sides
    • B62D11/04—Steering non-deflectable wheels; Steering endless tracks or the like by differentially driving ground-engaging elements on opposite vehicle sides by means of separate power sources
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D3/00—Control of position or direction
    • G05D3/12—Control of position or direction using feedback
    • G05D3/20—Control of position or direction using feedback using a digital comparing device
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/09—Supervised learning
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/60—Other road transportation technologies with climate change mitigation effect
    • Y02T10/72—Electric energy management in electromobility

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Mechanical Engineering (AREA)
  • Transportation (AREA)
  • Combustion & Propulsion (AREA)
  • Chemical & Material Sciences (AREA)
  • Automation & Control Theory (AREA)
  • Power Engineering (AREA)
  • Feedback Control In General (AREA)
  • Electric Propulsion And Braking For Vehicles (AREA)

Abstract

The invention provides a small wheel speed self-regulating and controlling method and a self-regulating and controlling system based on a pulse neural network, wherein the method comprises the following steps: obtaining a PWM value of a left wheel motor of the trolley; the PWM value is subjected to numerical value adjustment through a control state processor; acquiring the offset of the trolley movement angle recorded by an inertia measurement unit; adjusting the offset value through an angle processor; through the supervised learning algorithm of the pulse neural network and in combination with the angle signal, the PWM control signal is adjusted and applied to a trolley, and the wheel speed of the trolley is adjusted; and circulating the steps to realize real-time adjustment of the wheel speed of the trolley. The invention utilizes the characteristics of low power consumption and high dynamic property of the pulse neural network, combines the pulse neural network and the IMU to carry out trolley straight walking control, enables the trolley to have wheel speed self-adaption capability, and solves the problem of straight walking caused by factors such as different motor characteristics and external environment interference of the trolley.

Description

Small wheel speed self-regulation and control method and system based on pulse neural network
Technical Field
The invention relates to the field of intelligent trolley control, in particular to a trolley wheel speed self-regulation and control method and system based on a pulse neural network.
Background
In hazardous environments, people must take rigorous protective measures, and robots can enter or pass through these hazardous areas for maintenance and detection work without being protected as people do. The intelligent trolley is particularly suitable for working in the environment with severe conditions, has the characteristics of convenience in control, simplicity in algorithm deployment and test and the like, and can be widely applied to the fields of unmanned motor vehicles, unmanned production lines, warehouses, service robots and the like. In the field of education, the intelligent algorithm can be actually taught through the intelligent trolley.
In the current intelligent car, differential driving is still the mainstream. The differential driving is that the left wheel and the right wheel are respectively driven by a motor, and the advance and the steering of the trolley are realized by changing the rotating speed of the two motors. However, control using this method results in the vehicle not being able to "walk straight" steadily. The reason is as follows: firstly, due to the limitation of the preparation process, the driving characteristics of the motors for controlling the two wheels cannot be completely the same, the shapes and the sizes of the two motors cannot be completely consistent, and the precision in assembly is different; secondly, when the trolley runs, the tires of the trolley can slip or encounter small obstacles when rolling, and the speed of the left wheel and the speed of the right wheel are different, so that the trolley cannot run straight.
Open loop control without a feedback mechanism is unable to eliminate the left and right wheel speed error because the disturbances are random. When the trolley is disturbed, the left wheel and the right wheel can be fed back in time, the speed deviation of the two wheels is corrected, and therefore a straight line can be drawn.
At present, in the field of intelligent trolleys, the combination of a PID algorithm and a wheel speed detection device is a common method for solving the problem that trolleys are not straight, but the combination is not suitable for complex environments and has weak self-adaptive capacity, and the wheel speed detection device is not necessarily adaptive to the existing trolley motor scheme.
Therefore, a solution which can be applied to a complex environment and has a strong adaptive capability to make the trolley go straight is needed, and the above problems can be solved.
Disclosure of Invention
Based on the problems in the prior art, the invention provides a small wheel speed self-regulating and controlling method and a self-regulating and controlling system based on a pulse neural network. The specific scheme is as follows:
a small wheel speed self-regulation and control method based on a pulse neural network comprises the following steps:
acquiring a first PWM value of a left wheel motor of the trolley and a second PWM value of a right wheel motor of the trolley;
performing first value adjustment on the first PWM value and the second PWM value through a preset control state processor to obtain a PWM control signal capable of adapting to a preset pulse neural network;
acquiring the offset of the trolley movement angle recorded by a preset inertia measurement unit;
performing second numerical value adjustment on the offset through a preset angle processor to obtain an angle signal capable of adapting to the impulse neural network;
adjusting the PWM control signal by combining the angle signal through a supervised learning algorithm of the pulse neural network to obtain a left PWM value and a right PWM value;
respectively applying the left PWM value and the right PWM value to a left wheel motor and a right wheel motor of the trolley, and adjusting the speed of the trolley;
and circulating the steps to realize real-time adjustment of the wheel speed of the trolley.
In one embodiment, the "adjusting the wheel speed value by the supervised learning algorithm of the impulse neural network in combination with the angle signal" specifically includes:
on the basis of a preset neuron model, taking the PWM control signal output by the control state processor as a training signal, and combining a nonlinear nerve activation function of the neuron model to obtain abstract neuron activity;
obtaining a neuron decoder based on the neuron activity and the PWM control signal;
the angle signal output by the angle processor is taken as a learning signal, a supervised learning algorithm of the impulse neural network is taken as a learning algorithm, and the neuron decoder and the neuron activity are combined to obtain the connection weight between neurons;
synthesizing the neuron activity, the neuron decoder, and the connection weights to obtain the modified left PWM value and the modified right PWM value.
In a specific embodiment, the first numerical adjustment comprises scaling the first PWM value and the second PWM value to be linearly mapped to a [0,1] interval;
the second numerical adjustment includes scaling the offset, linearly mapping to the [ -1,1] interval.
In a specific embodiment, the expression of the first numerical adjustment is:
Figure BDA0003257012070000031
wherein x' represents the PWM adjustment signal, x represents the first PWM value or the second PWM value, min (x) represents a minimum value of the first PWM value or a minimum value of the second PWM value, and max (x) represents a maximum value of the first PWM value or a maximum value of the second PWM value;
the expression of the second numerical adjustment is:
Figure BDA0003257012070000032
where θ' represents the angle signal, θ represents the offset amount, max (θ) represents the maximum value of the offset amount, min (θ) represents the minimum value of the offset amount, and mean (θ) represents the average value of the offset amounts.
In one embodiment, the abstract expression of neuron activity is:
a＝G[αe·x]
wherein a represents the neuron activity, α is a scale factor associated with the neuron, e represents an encoder of the neuron, x represents information to be encoded, and G [. Cndot. ] represents a nonlinear neural activation function determined by a preset neuron model;
the decoder expression includes:
d＝r -1 T
T ij ＝∫a i a j dx
r＝∫a j xdx
wherein d denotes a decoder, a i Representing the activity of neuron i, a j Representing the activity of neuron j, T and r are intermediate variables, and x represents the input data.
In an embodiment, the supervised learning algorithm is a pre-written Error Sensitivity supervised learning algorithm, and the obtained weight correction value Δ ω is obtained ij The following equation can be used:
Figure BDA0003257012070000041
wherein, Δ ω ij Representing weight correction values, E representing an error vector to be minimized, k representing a learning rate, d representing a decoder, n being a dimension of the decoder, ω representing a connection weight, α representing a scale factor associated with the neuron, E representing an encoder of the neuron, i, j representing the neuron, a i Representing the activity of neuron i.
In a specific embodiment, the neuron activity, the neuron decoder, and the connection weight are integrated to obtain a modified PWM value, and the specific expression is as follows:
Figure BDA0003257012070000042
where a represents neuron activity, ω represents connection weights between neurons, d represents a neuron decoder, and P represents a modified PWM value.
A small wheel speed self-regulating system based on a pulse neural network comprises the following components:
a wheel speed acquisition unit: the system comprises a first PWM (pulse width modulation) value acquisition module, a second PWM value acquisition module and a third PWM value acquisition module, wherein the first PWM value acquisition module is used for acquiring a first PWM value of a left wheel motor of the trolley and a second PWM value of a right wheel motor of the trolley;
a wheel speed adjusting unit: the built-in control state processor is used for carrying out first numerical value adjustment on the first PWM value and the second PWM value through the control state processor to obtain a PWM control signal capable of adapting to a preset pulse neural network;
an offset acquisition unit: the device is used for acquiring the offset of the trolley movement angle recorded by the inertial measurement unit;
an offset adjustment unit: the built-in angle processor is used for carrying out second numerical value adjustment through the angle processor to obtain an angle signal which can be adapted to the pulse neural network;
an adjustment control unit: the built-in pulse neural network is used for adjusting the PWM control signal by combining the angle signal through a supervised learning algorithm of the pulse neural network to obtain a left PWM value and a right PWM value;
a shift speed application unit: the left PWM value and the right PWM value are respectively applied to a left wheel motor and a right wheel motor of the trolley, and the speed of the trolley is adjusted;
a real-time adjusting unit: the system is used for circulating the wheel speed acquisition unit, the wheel speed adjustment unit, the deviation acquisition unit, the deviation adjustment unit, the adjustment control unit and the speed shifting application unit in sequence to realize real-time adjustment of the wheel speed of the trolley.
In a specific embodiment, the adjusting and controlling unit specifically includes:
a neuron activity module: the device is used for obtaining abstract neuron activity by taking the PWM control signal output by the control state processor as a training signal based on a preset neuron model and combining a nonlinear nerve activation function of the neuron model;
a decoder module: for deriving a neuron decoder based on the neuron activity and the PWM control signal;
a connection weight module: the device is used for obtaining connection weight between the neurons by taking the angle signal output by the angle processor as a learning signal, taking a supervised learning algorithm of the impulse neural network as a learning algorithm and combining the neuron decoder and the neuron activity;
an output module: for synthesizing the neuron activity, the neuron decoder, and the connection weights to obtain the modified left PWM value and the modified right PWM value.
In one embodiment, the wheel speed adjustment unit comprises scaling the first PWM value and the second PWM value to be linearly mapped to a [0,1] interval;
the offset adjustment unit comprises scaling the offset amount to be linearly mapped to the [ -1,1] interval.
The invention has the following beneficial effects:
the invention provides a small wheel speed self-regulation and control method and system based on a pulse neural network. The method for automatically controlling the wheel speed of the trolley utilizes the characteristics of low power consumption and high dynamic property of the pulse neural network, combines the pulse neural network and the IMU to carry out the direct walking control of the trolley, ensures that the trolley has the self-adaptive capability of the wheel speed, and solves the problem of direct walking caused by factors such as different motor characteristics and interference of the external environment of the trolley. Compared with the scheme of combining the existing PID with the wheel speed detection device, the scheme provided by the invention can realize the straight walking of the trolley by adjusting the wheel speed in real time even in a complex environment, has high adaptability and anti-interference capability, and can be adapted to more extensive types of intelligent trolleys.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and those skilled in the art can also obtain other related drawings based on the drawings without inventive efforts.
FIG. 1 is a schematic flow chart of a method for automatically controlling wheel speed of a cart according to an embodiment of the present invention;
FIG. 2 is a schematic diagram illustrating a principle of a method for automatically controlling a wheel speed of a cart according to an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating a principle formula of a method for automatically controlling a wheel speed of a cart according to an embodiment of the present invention;
FIG. 4 is a flowchart illustrating step 105 according to an embodiment of the present invention;
FIG. 5 is a graph of experimental results for an embodiment of the present invention;
FIG. 6 is a block diagram of a wheel speed self-regulating system according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of an adjustment control unit according to an embodiment of the present invention.
Reference numerals: 1-a cyclic wheel speed acquisition unit; 2-wheel speed adjusting unit; 3-an offset acquisition unit; 4-an offset adjustment unit; 5-adjusting the control unit; 6-a shift speed application unit; 7-a real-time adjustment unit; 51-neuron activity module; 52-a decoder module; 53-connection weight module; 54-output module.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
The embodiment provides a small car wheel speed self-regulating method based on a pulse neural network, which realizes the straight-walking control of a small car by combining the pulse neural network and an IMU (inertial measurement Unit), and solves the problem of straight-walking caused by factors such as different motor characteristics and external environment interference of the small car. The flow diagram of the self-regulation method is shown in the attached figure 1 in the specification. The specific scheme is as follows:
a small wheel speed self-regulation and control method based on a pulse neural network comprises the following steps:
101. acquiring a first PWM value of a left wheel motor of the trolley and a second PWM value of a right wheel motor of the trolley;
102. performing first value adjustment on the first PWM value and the second PWM value through a preset control state processor to obtain a PWM control signal capable of adapting to a preset pulse neural network;
103. acquiring the offset of the trolley movement angle recorded by a preset inertia measurement unit;
104. performing second numerical value adjustment on the offset through a preset angle processor to obtain an angle signal capable of adapting to the impulse neural network;
105. adjusting the PWM control signal by combining an angle signal through a supervised learning algorithm of a pulse neural network to obtain a left PWM value and a right PWM value;
106. respectively applying the left PWM value and the right PWM value to a left wheel motor and a right wheel motor of the trolley, and adjusting the speed of the trolley;
107. and (4) circulating the steps 101-106 to realize the real-time adjustment of the wheel speed of the trolley.
In the present embodiment, the wheel speed control of the wheel employs Pulse Width Modulation (PWM). Pulse width modulation is a very effective technique for controlling an analog circuit using a digital output of a microprocessor, and equivalently obtains a required waveform by modulating the width of a pulse, thereby realizing the control of the wheel speed of a vehicle.
Fig. 2 in the description is a schematic diagram of the embodiment, and fig. 3 in the description is a schematic diagram of formula expression of each module in fig. 2 in the description. The self-adaptation of the wheel speed of the trolley is realized through closed-loop control.
Specifically, 101, a first PWM value of a trolley left wheel motor and a second PWM value of a trolley right wheel motor are obtained. In this embodiment, the left and right wheels of the cart are each driven by a motor, with the left wheel motor driving the left wheel of the cart and the right wheel motor driving the right wheel of the cart.
The PWM value of the trolley is changed along with the speed of the trolley. Due to the limitation of the preparation process, the driving characteristics of the motors for controlling the two wheels cannot be completely the same, the sizes of the two motors cannot be completely consistent, and the precision in assembly is different; in addition, when the trolley runs, the tires may slip or encounter fine obstacles when rolling, and the speed of the left wheel and the speed of the right wheel are different, so that the trolley cannot run straight. The embodiment collects and controls the PWM values of the left wheel motor and the right wheel motor of the trolley in real time so as to realize the stable straight movement of the trolley.
Specifically, 102, a first value adjustment is performed on the first PWM value and the second PWM value through a preset control state processor, so as to obtain a PWM control signal capable of adapting to a preset pulse neural network. Since the magnitude of the PWM value of the cart is variable, its value needs to be scaled in order to enable the PWM value to fit the neuron model of the impulse neural network.
It should be noted that the scaling manner adopted by the first numerical adjustment includes, but is not limited to, any known numerical scaling manner. This embodiment is illustrated in a rescale example.
Illustratively, the first and second PWM values are first numerically adjusted by the control state processor to be linearly mapped to an interval of [0,1] to fit a neuron model of the spiking neural network. The expression for the first numerical adjustment is:
Figure BDA0003257012070000091
wherein x' represents the adjusted output signal, which is the PWM adjustment signal in this embodiment. x represents the input signal before adjustment, i.e. the first PWM value or the second PWM value in this embodiment. min (x) represents the minimum value of the input signal, which is the minimum value of the first PWM value or the minimum value of the second PWM value in this embodiment. max (x) represents the maximum value of the input signal, i.e. the maximum value of the first PWM value or the maximum value of the second PWM value in this embodiment.
Specifically, 103, acquiring the offset of the trolley movement angle recorded by the preset inertia measurement unit. In the embodiment, the offset of the trolley during movement is measured by the inertial measurement unit.
An Inertial Measurement Unit (IMU) is a device for measuring the three-axis attitude angle (or angular velocity) and acceleration of an object. Gyroscopes and accelerometers are the main components of the IMU, the accuracy of which directly affects the accuracy of the inertial system. In actual work, errors of the gyroscope and the accelerometer are generated due to various unavoidable interference factors, and navigation errors of the gyroscope and the accelerometer grow along with time from initial alignment, particularly position errors, which are main defects of the inertial navigation system. Therefore, the method needs to be assisted by external information to realize combined navigation, so that the problem of error accumulation over time is effectively reduced. To increase reliability, more sensors may be provided for each axis. Generally, the IMU is mounted at the center of gravity of the object being tested. Inertia measurement unit compares in traditional fast detection device of wheel, and is wider to the adaptation scope of dolly model, installs on the dolly more easily additional.
Specifically, 104, a preset angle processor performs second numerical adjustment on the offset to obtain an angle signal capable of adapting to the impulse neural network. The offset is processed similarly to the PWM values to adapt the neural model of the impulse neural network.
It should be noted that the scaling manner used for the second numerical adjustment includes, but is not limited to, any known numerical scaling manner. This embodiment is illustrated in a rescale example.
Illustratively, the offset is second numerically adjusted by the angle processor to fit a neuron model of the spiking neural network. In contrast to the first numerical adjustment, the second numerical adjustment is a linear mapping of the offset to the range of [ -1,1] to obtain an angle signal, and the positive and negative values of the angle signal are used to represent the offset direction of the vehicle. The expression for the second numerical adjustment is:
Figure BDA0003257012070000101
where θ' represents the adjusted output signal, which in this embodiment is an angle signal. θ represents the input signal before adjustment, and is an offset in the present embodiment. max (θ) represents the maximum value of the input signal, in this embodiment, the maximum value of the offset amount, min (θ) represents the minimum value of the input signal, in this embodiment, the minimum value of the offset amount, and mean (θ) represents the average value of the offset amounts.
Specifically, 105, the PWM control signal is adjusted by a supervised learning algorithm of the pulse neural network in combination with the angle signal, so as to obtain a left PWM value and a right PWM value. The step is realized by a pulse neural network, the pulse neural network takes a PWM control signal output by a control state processor as a training signal, takes an angle signal output by an angle processor as a learning signal, and uses a corresponding pulse neural network supervised learning algorithm as a learning algorithm to adjust the PWM control signal.
Step 105 is as shown in fig. 4 of the specification, and specifically includes:
10501. based on a preset neuron model, taking a PWM control signal output by a control state processor as a training signal, and combining a nonlinear nerve activation function of the neuron model to obtain abstract neuron activity;
10502. obtaining a neuron decoder based on the neuron activity and the PWM control signal;
10503. the angle signal output by the angle processor is taken as a learning signal, a supervised learning algorithm of a pulse neural network is taken as a learning algorithm, and a neuron decoder and neuron activities are combined to obtain the connection weight between neurons;
10504. and synthesizing the neuron activity, the neuron decoder and the connection weight to obtain a corrected left PWM value and a corrected right PWM value.
The Spiking Neural Network (SNN) is considered to be a third generation neural network, which is a neural network model with more biological significance and the most similar operation mechanism to the brain. The impulse neural network utilizes different neuron models to transmit and calculate information, the different neuron models have different characteristics, and a more appropriate model can be found by replacing neurons.
In step 10501, the abstractable representation of neuron activity in the spiking neural network is:
a＝G[αe·x]
where a represents neuron activity, α is a scale factor associated with the neuron, e represents the encoder of the neuron, and x represents the information to be encoded, i.e., the PWM control signal. G [. Cndot. ] represents a nonlinear neural activation function, determined by a specific neuron model;
in step 10502, the decoder d can be calculated according to the following formula:
d＝r -1 T
T ij ＝∫a i a j dx
r＝∫a j xdx
wherein d denotes a decoder, a i Representing the activity of neuron i, a j Representing the activity of neuron j, T and r being intermediate variables, and x representing the input data, i.e. the PWM control signal.
In step 10503, the supervised learning algorithm includes a plurality of algorithms that can be replaced based on actual demand. In this embodiment, the supervised learning algorithm is a pre-written Error Sensitivity supervised learning algorithm. Weight correction value delta omega obtained by using Prescribed Error Sensitivity supervised learning algorithm ij The following equation can be used:
Figure BDA0003257012070000111
wherein, Δ ω ij Indicating the weight correction value and E the error vector, i.e. the angular offset in this embodiment, that is desired to be minimized. K denotes the learning rate, d denotes the decoder, n is the decoder dimension, ω denotes the connection weight, α denotes the scale factor associated with the neuron, e denotes the encoder for the neuron, i, j denote the neuron, a i Representing the activity of neuron i.
And synthesizing the neuron activity, the neuron decoder and the connection weight to obtain a corrected PWM value, wherein the specific expression is as follows:
Figure BDA0003257012070000112
where a represents neuron activity, ω represents connection weights between neurons, d represents a neuron decoder, and P represents a modified PWM value.
And inputting the PWM control signal into the pulse neural network to obtain the corrected PWM values, namely a left PWM value and a right PWM value. The pulse neural network has the characteristics of low power consumption and good dynamic property, and a highly adaptive and relatively low-power-consumption solution can be provided for the problem of 'walking straightness' of the trolley by combining the pulse neural network with an Inertial Measurement Unit (IMU).
And specifically, 106, applying the left PWM value and the right PWM value to a left wheel motor and a right wheel motor of the trolley respectively, and adjusting the speed of the trolley. And step 105, obtaining the corrected values of the left wheel and the right wheel, and applying the left PWM value and the right PWM value to the trolley to adjust the moving speed of the left wheel and the right wheel of the trolley.
Specifically, 107, the steps 101-106 are repeated to realize the real-time adjustment of the wheel speed of the trolley. The current wheel speed and offset can be acquired when the wheel speed of the trolley is adjusted once, and the wheel speed self-adaption is realized. Especially under the complex environment, the dolly receives the interference of environmental factor constantly, can let the dolly possess higher adaptability through real-time adjustment wheel speed.
Description of the drawings fig. 5 is a graph showing experimental results of this embodiment, and results of cart straight-moving control using a combination of a pulse neural network and an IMU. Where the x-axis is time and the y-axis is angular offset. The experimental result shows that after 6s self-adaptive adjustment, the trolley can walk linearly and basically has no angular deviation.
The embodiment provides a small car wheel speed self-regulation and control method based on a pulse neural network, which utilizes the characteristics of low power consumption and high dynamic property of the pulse neural network to combine the pulse neural network and an IMU (inertial measurement Unit) to carry out straight car walking control, so that the small car has wheel speed self-adaption capability, and the problem of 'straight car walking' caused by factors such as different motor characteristics of the small car and external environment interference is solved. Compare in the scheme of current PID combination wheel speed detection device, the scheme that this embodiment provided also can realize the dolly through real-time adjustment wheel speed and walk directly under the complex environment, possesses high adaptability and interference killing feature to the type of the intelligent vehicle that can adapt is also more extensive.
Example 2
The embodiment provides a small wheel speed self-regulating system based on a pulse neural network, which adopts the small wheel speed self-regulating method based on the pulse neural network provided by the embodiment 1, and the module schematic diagram of the small wheel speed self-regulating system is shown as the attached figure 6 in the specification. The specific scheme is as follows:
a small wheel speed self-regulating system based on a pulse neural network comprises a circulating wheel speed acquisition unit 1, a wheel speed regulation unit 2, an offset acquisition unit 3, an offset regulation unit 4, a regulation control unit 5, a moving speed application unit 6 and a real-time regulation unit 7 as follows:
wheel speed acquisition unit 1: the system comprises a control module, a first PWM (pulse width modulation) module, a second PWM module and a first PWM module, wherein the control module is used for acquiring a first PWM value of a left wheel motor of the trolley and a second PWM value of a right wheel motor of the trolley;
wheel speed adjusting means 2: the built-in control state processor is used for carrying out first numerical value adjustment on the first PWM value and the second PWM value through the control state processor to obtain a PWM control signal which can be adapted to the preset pulse neural network. The wheel speed adjustment unit 2 includes scaling the first PWM value and the second PWM value to be linearly mapped to the [0,1] interval.
Offset acquisition unit 3: the system is used for acquiring the offset of the trolley movement angle recorded by the inertia measurement unit;
offset adjustment unit 4: and an angle processor is arranged in the device and used for carrying out second numerical value adjustment on the offset through the angle processor to obtain an angle signal which can be adapted to the impulse neural network. The offset adjustment unit 4 includes scaling the offset amount to be linearly mapped to the [ -1,1] interval.
The adjustment control unit 5: the built-in pulse neural network is used for adjusting the PWM control signal through a supervised learning algorithm of the pulse neural network and combining the angle signal to obtain a left PWM value and a right PWM value;
the shift speed application unit 6: the left PWM value and the right PWM value are respectively applied to a left wheel motor and a right wheel motor of the trolley, and the speed of the trolley wheel is adjusted;
the real-time adjusting unit 7: the device is used for sequentially circulating the wheel speed acquisition unit 1, the wheel speed adjustment unit 2, the offset acquisition unit 3, the offset adjustment unit 4, the adjustment control unit 5 and the shifting speed application unit 6 to realize real-time adjustment of the wheel speed of the trolley.
The tuning control unit 5 includes a neuron activity module 51, a decoder module 52, a connection weight module 53, and an output module 54, as shown in fig. 7.
The neuron activity module 51: and the method is used for obtaining abstract neuron activity by taking the PWM control signal output by the control state processor as a training signal based on a preset neuron model and combining a nonlinear nerve activation function of the neuron model.
An abstractable representation of neuron activity in a spiking neural network is:
a＝G[αe·x]
where a denotes the neuron activity, α is a scale factor associated with the neuron, e denotes the encoder of the neuron, and x denotes the information to be encoded, i.e. the PWM control signal. G · represents a nonlinear neural activation function, determined by a specific neuron model.
The decoder module 52: for deriving a neuron decoder based on the neuron activity and the PWM control signal.
The decoder d can be calculated according to the following formula:
d＝r -1 T
T ij ＝∫a i a j dx
r＝∫a j xdx
wherein d denotes a decoder, a i Representing the activity of neuron i, a j Representing the activity of neuron j, T and r are intermediate variables, and x represents input data, i.e., a PWM control signal.
Connection weighting module 53: the method is used for obtaining the connection weight between the neurons by taking the angle signal output by the angle processor as a learning signal, taking the supervised learning algorithm of the impulse neural network as a learning algorithm and combining the neuron decoder and the neuron activity.
In this embodiment, the supervised learning algorithm is the Prescribed Error Sensitivity supervised learning algorithm. Weight correction value delta omega obtained by using Presscribed Error Sensitivity supervised learning algorithm ij The following equation can be used:
Figure BDA0003257012070000141
wherein, Δ ω ij Indicating the weight correction value and E the error vector that is desired to be minimized, in this embodiment the angle offset. K denotes the learning rate, d denotes the decoder, n is the decoder dimension, ω denotes the connection weight, α denotes the scale factor associated with the neuron, e denotes the encoder for the neuron, i, j denote the neuron, a i Representing the activity of neuron i.
The output module 54: for synthesizing the neuron activity, the neuron decoder and the connection weights to obtain the corrected left and right PWM values.
And synthesizing the neuron activity, the neuron decoder and the connection weight to obtain a corrected PWM value, wherein the specific expression is as follows:
Figure BDA0003257012070000151
where a represents neuron activity, ω represents connection weights between neurons, d represents a neuron decoder, and P represents a modified PWM value.
And inputting the PWM control signal into the pulse neural network to obtain the corrected PWM values, namely a left PWM value and a right PWM value. The pulse neural network has the characteristics of low power consumption and good dynamic property, and a solution with high self-adaption and relatively low power consumption can be provided for the problem of 'non-straight walking' of the trolley by combining the pulse neural network with an Inertial Measurement Unit (IMU).
The embodiment provides a small wheel speed self-regulating system based on a pulse neural network, and the small wheel speed self-regulating system based on the pulse neural network in the embodiment 1 is systematized, so that the small wheel speed self-regulating system has higher practicability.
The invention provides a small wheel speed self-regulating and controlling method and system based on a pulse neural network. The method for automatically controlling the wheel speed of the trolley utilizes the characteristics of low power consumption and high dynamic property of the pulse neural network, combines the pulse neural network and the IMU to carry out the direct walking control of the trolley, ensures that the trolley has the self-adaptive capability of the wheel speed, and solves the problem of direct walking caused by factors such as different motor characteristics and interference of the external environment of the trolley. Compared with the scheme of combining the existing PID with the wheel speed detection device, the scheme provided by the invention can realize the straight walking of the trolley by adjusting the wheel speed in real time even in a complex environment, has high adaptability and anti-interference capability, and can be adapted to more extensive types of intelligent trolleys.
It will be understood by those skilled in the art that the modules or steps of the invention described above may be implemented by a general purpose computing device, they may be centralized on a single computing device or distributed across a network of computing devices, and optionally they may be implemented by program code executable by a computing device, such that it may be stored in a memory device and executed by a computing device, or it may be separately fabricated into various integrated circuit modules, or it may be fabricated by fabricating a plurality of modules or steps thereof into a single integrated circuit module. Thus, the present invention is not limited to any specific combination of hardware and software.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. Those skilled in the art will appreciate that the present invention is not limited to the particular embodiments described herein, and that various obvious changes, rearrangements and substitutions will now be apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in some detail by the above embodiments, the invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the invention, and the scope of the invention is determined by the scope of the appended claims.
The above disclosure is only a few specific implementation scenarios of the present invention, however, the present invention is not limited thereto, and any variations that can be made by those skilled in the art are intended to fall within the scope of the present invention.

Claims (10)

1. A small wheel speed self-regulation and control method based on a pulse neural network is characterized by comprising the following steps:
acquiring a first PWM value of a left wheel motor of the trolley and a second PWM value of a right wheel motor of the trolley;
performing first value adjustment on the first PWM value and the second PWM value through a preset control state processor to obtain a PWM control signal capable of adapting to a preset pulse neural network;
acquiring the offset of the trolley movement angle recorded by a preset inertia measurement unit;
performing second numerical value adjustment on the offset through a preset angle processor to obtain an angle signal capable of adapting to the impulse neural network;
adjusting the PWM control signal by combining the angle signal through a supervised learning algorithm of the pulse neural network to obtain a left PWM value and a right PWM value;
respectively applying the left PWM value and the right PWM value to a left wheel motor and a right wheel motor of the trolley, and adjusting the speed of the trolley;
and circulating the steps to realize the real-time adjustment of the wheel speed of the trolley.
2. The method for self-regulating wheel speed of a small vehicle as claimed in claim 1, wherein the step of regulating the wheel speed value by combining the angle signal through a supervised learning algorithm of the impulse neural network specifically comprises the steps of:
on the basis of a preset neuron model, taking the PWM control signal output by the control state processor as a training signal, and combining a nonlinear nerve activation function of the neuron model to obtain abstracted neuron activity;
obtaining a neuron decoder based on the neuron activity and the PWM control signal;
the angle signal output by the angle processor is taken as a learning signal, a supervised learning algorithm of the pulse neural network is taken as a learning algorithm, and the neuron decoder and the neuron activity are combined to obtain the connection weight between neurons;
synthesizing the neuron activity, the neuron decoder, and the connection weights to obtain the modified left PWM value and the modified right PWM value.
3. The method as claimed in claim 1 or 2, wherein the first value adjustment comprises scaling the first PWM value and the second PWM value to be linearly mapped to a [0,1] interval;
the second numerical adjustment includes scaling the offset, linearly mapping to [ -1,1] interval.
4. A method of self-regulating wheel speed as claimed in claim 3, wherein the first value adjustment is expressed by the following expression:
Figure FDA0003257012060000021
wherein x' represents the PWM adjustment signal, x represents the first PWM value or the second PWM value, min (x) represents a minimum value of the first PWM value or a minimum value of the second PWM value, and max (x) represents a maximum value of the first PWM value or a maximum value of the second PWM value;
the expression of the second numerical adjustment is:
Figure FDA0003257012060000022
where θ' represents the angle signal, θ represents the offset amount, max (θ) represents the maximum value of the offset amount, min (θ) represents the minimum value of the offset amount, and mean (θ) represents the average value of the offset amounts.
5. The method as claimed in claim 2, wherein the abstract expression of neuron activity is:
a＝G[αe·x]
wherein a represents the neuron activity, α represents a scale factor associated with the neuron, e represents an encoder of the neuron, x represents information to be encoded, and G [. Cndot. ] represents a nonlinear neural activation function determined by a preset neuron model;
the decoder expression includes:
d＝r -1 T
T ij ＝∫a i a j dx
r＝∫a j xdx
wherein d denotes a decoder, a i Representing the activity of neuron i, a j Representing the activity of neuron j, T and r are intermediate variables, and x represents the input data.
6. The method as claimed in claim 5, wherein the supervised learning algorithm is a Prescripted Error Sensitivity supervised learning algorithm, and the obtained weight correction value Δ ω is obtained ij The following equation can be used:
Figure FDA0003257012060000031
wherein, Δ ω ij Representing weight modification values, E representing error vectors desired to be minimized, k representing a learning rate, d representing a decoder, n being a dimension of the decoder, ω representing a connection weight, α representing a scale factor associated with the neuron, E representing an encoder of the neuron, i, j representing the neuron, a i Representing the activity of neuron i.
7. The method for self-regulating wheel speed of a small vehicle as claimed in claim 6, wherein the specific expression for obtaining the modified PWM value by integrating the neuron activity, the neuron decoder and the connection weight is as follows:
Figure FDA0003257012060000032
where a represents neuron activity, ω represents connection weights between neurons, d represents a neuron decoder, and P represents a modified PWM value.
8. The utility model provides a dolly wheel speed self-regulating system based on pulse neural network which characterized in that includes as follows:
a wheel speed acquisition unit: the system comprises a first PWM (pulse width modulation) value acquisition module, a second PWM value acquisition module and a third PWM value acquisition module, wherein the first PWM value acquisition module is used for acquiring a first PWM value of a left wheel motor of the trolley and a second PWM value of a right wheel motor of the trolley;
a wheel speed adjusting unit: the built-in control state processor is used for carrying out first value adjustment on the first PWM value and the second PWM value through the control state processor to obtain a PWM control signal which can be adapted to a preset pulse neural network;
an offset acquisition unit: the system is used for acquiring the offset of the trolley movement angle recorded by the preset inertia measurement unit;
an offset adjustment unit: the built-in angle processor is used for carrying out second numerical value adjustment through the angle processor to obtain an angle signal which can be adapted to the pulse neural network;
an adjustment control unit: the built-in pulse neural network is used for adjusting the PWM control signal by combining the angle signal through a supervised learning algorithm of the pulse neural network to obtain a left PWM value and a right PWM value;
a shift speed application unit: the left PWM value and the right PWM value are respectively applied to a left wheel motor and a right wheel motor of the trolley, and the speed of the trolley is adjusted;
a real-time adjusting unit: the device is used for sequentially circulating the wheel speed acquisition unit, the wheel speed adjustment unit, the offset acquisition unit, the offset adjustment unit, the adjustment control unit and the moving speed application unit to realize real-time adjustment of the wheel speed of the trolley.
9. The wheel speed self-regulating system of a small vehicle as claimed in claim 8, wherein the regulating control unit specifically comprises:
a neuron activity module: the PWM control signal output by the control state processor is used as a training signal based on a preset neuron model, and abstract neuron activity is obtained by combining a nonlinear nerve activation function of the neuron model;
a decoder module: for deriving a neuron decoder based on the neuron activity and the PWM control signal;
a connection weight module: the device is used for obtaining connection weight between the neurons by taking the angle signal output by the angle processor as a learning signal, taking a supervised learning algorithm of the impulse neural network as a learning algorithm and combining the neuron decoder and the neuron activity;
an output module: for synthesizing the neuron activity, the neuron decoder, and the connection weights to obtain the modified left PWM value and the modified right PWM value.
10. The small wheel speed self-regulating system as claimed in claim 8 or 9, wherein the wheel speed regulating unit comprises scaling the first PWM value and the second PWM value to be linearly mapped to [0,1] interval;
the offset adjustment unit comprises scaling the offset amount to be linearly mapped to the [ -1,1] interval.
CN202111062796.XA 2021-09-10 2021-09-10 Small wheel speed self-regulation and control method and system based on pulse neural network Pending CN115782606A (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202111062796.XA CN115782606A (en) 2021-09-10 2021-09-10 Small wheel speed self-regulation and control method and system based on pulse neural network
PCT/CN2021/137998 WO2023035462A1 (en) 2021-09-10 2021-12-14 Pulse neural network-based vehicle wheel speed self-regulating method and self-regulating system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111062796.XA CN115782606A (en) 2021-09-10 2021-09-10 Small wheel speed self-regulation and control method and system based on pulse neural network

Publications (1)

Publication Number Publication Date
CN115782606A true CN115782606A (en) 2023-03-14

Family

ID=85416788

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111062796.XA Pending CN115782606A (en) 2021-09-10 2021-09-10 Small wheel speed self-regulation and control method and system based on pulse neural network

Country Status (2)

Country Link
CN (1) CN115782606A (en)
WO (1) WO2023035462A1 (en)

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104950887B (en) * 2015-06-19 2017-07-21 重庆大学 Conveying arrangement based on robotic vision system and independent tracking system
JP6479627B2 (en) * 2015-10-29 2019-03-06 株式会社クボタ Electric work vehicle
CN105305895B (en) * 2015-11-17 2017-09-05 吉林大学 A kind of brushless motor control method based on torque feedback and switch compensation
CN108427415A (en) * 2018-04-02 2018-08-21 济南大学 Homing guidance avoidance robot car and its control method in a kind of modern logistics
CN108594829A (en) * 2018-06-08 2018-09-28 武汉蛋玩科技有限公司 A kind of Intelligent patrol robot and control method
CN109240292A (en) * 2018-09-10 2019-01-18 广州大学 A kind of self-balancing trolley and its attitude control method based on BP neural network
CN109085855A (en) * 2018-09-18 2018-12-25 北京图森未来科技有限公司 A kind of sensor position adjustment device, system, method of adjustment and vehicle
CN110109456A (en) * 2019-04-25 2019-08-09 杭州师范大学钱江学院 A kind of trolley automatic Pilot method
WO2021138700A1 (en) * 2020-01-05 2021-07-08 Eva, Llc Automated steering control mechanism and system for wheeled vehicles
CN111716359B (en) * 2020-06-30 2021-09-28 福州大学 Composite navigation type differential storage robot and working method thereof
CN112141210B (en) * 2020-08-21 2022-08-02 天津市天波科达科技有限公司 Course control method and device based on electronic differential chassis

Also Published As

Publication number Publication date
WO2023035462A1 (en) 2023-03-16

Similar Documents

Publication Publication Date Title
Yue et al. Neural network based terminal sliding mode control for WMRs affected by an augmented ground friction with slippage effect
Corradini et al. Robust tracking control of mobile robots in the presence of uncertainties in the dynamical model
CN110989597B (en) Adaptive path tracking method of integrated fuzzy neural network
CN104942809A (en) Mechanical arm dynamic fuzzy approximator based on visual servo system
Kuo Trajectory and heading tracking of a mecanum wheeled robot using fuzzy logic control
CN108037658A (en) A kind of method for correcting error of the robot kinematic error based on navigation system
CN111897347A (en) Course retainer of double-motor-propelled unmanned ship based on neural network PID control and retaining method
CN110109363B (en) Neural network self-adaptive control method for wheeled mobile robot formation
Prkačin et al. State and parameter estimation of suspended load using quadrotor onboard sensors
Zhang et al. EKF-based LQR tracking control of a quadrotor helicopter subject to uncertainties
US20230409035A1 (en) Method for building controller for robot, method, device for controlling motion of robot, and robot
CN115782606A (en) Small wheel speed self-regulation and control method and system based on pulse neural network
CN111258326B (en) Bionic robot fish roll stability control method and system based on reaction wheel
CN111679681B (en) Underwater robot motion path tracking method based on sliding mode robust control
Gavani et al. Path planning of three wheeled omni-directional robot using bezier curve tracing technique and pid control algorithm
Cheng et al. Trajectory tracking control of nonholonomic mobile robots by backstepping
Reis et al. Super-twisting control with quaternion feedback for a 3-DoF inertial stabilization platform
Díaz-Téllez et al. ROS-based Controller for a Two-Wheeled Self-Balancing Robot
CN114734437A (en) Robot joint control method and device
CN113848954A (en) Improved unmanned vehicle servo control system of multi-neuron neural network PID algorithm
CN116811592A (en) Trolley direction regulating and controlling method based on impulse neural network and application thereof
Sanca et al. Stability analysis of a multirotor UAV with robust backstepping controller
Fényes et al. Observer design with performance guarantees for vehicle control purposes via the integration of learning-based and LPV approaches
Ivoilov et al. The power consumption decreasing of the two-wheeled balancing robot
CN115840367B (en) Nonlinear robust control method for attitude and altitude system of disturbed four-rotor unmanned aerial vehicle

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