JP7421719B2 - Control devices, control systems, and control programs - Google Patents

Description

The present invention relates to a control device, a control system, and a control program for systems that involve complex and dynamic state changes, such as the movement and control of industrial, medical, and household robots and mobile systems, and manufacturing plants.

Robots and industrial machines are rapidly becoming more complex and sophisticated in industries such as industry, commerce, and agriculture, in the medical field such as surgery, nursing, and nursing care, and even in the home, such as cleaning. The components of these devices such as robots and industrial machines are not uniform, and the work objects and operating environments are not necessarily constant.

Under such circumstances, Patent Document 1 has been proposed as a bipedal walking robot that generates movement patterns for each individual by repeatedly performing learning on a device to which a neuron network is applied. As described in claim 1, Patent Document 1 includes a neuron network, and as described in FIG. 3 and [0004], it is intended to improve design accuracy and shorten design time through repeated learning.

JP2006-88331A

As described in FIG. 11 and paragraphs [0007] to [0008], the neuron network used in Patent Document 1 generates an output signal by applying a unique weighting coefficient W_k_* to multiple inputs. There is. This weighting coefficient W_k_* is optimized through repeated learning. Therefore, although the configuration of the individual device is fixed and optimization is possible under conditions where the environment is constant, it is not possible to follow unexpected changes in the environment due to disturbances or the like.

The present invention has been made in view of the above circumstances, and enables the controlled object to react autonomously and stably perform desired operations even when unexpected changes in the external environment occur. The purpose is to provide a control device, a control system, and a control program.

According to the present invention, the control device is configured to be able to control the controlled object by supplying a drive signal to the controlled object, and includes a spike signal train generation section and a drive signal generation section, The spike signal train generation unit is configured to be able to generate a spike signal train at a timing defined by a basic control signal for controlling the controlled object, an internal state including disturbance, and dynamics regarding the internal state, and A control device is provided in which the generation unit is configured to be able to generate the drive signal that changes continuously in time series based on the spike signal train.

In the control device according to the present invention, the basic control signal is once converted into a spike signal train by the spike signal train generation section, and then the control target is controlled using the drive signal generated by the drive signal generation section. . At this time, even if an unexpected change in the external environment occurs, the control device side reacts autonomously, and the advantageous effect is that the entire control system can perform the desired operation.

FIG. 1 is a functional block diagram of a control system including a control device and a controlled object according to an embodiment of the present invention. The optimization control flow diagram in the control device. FIG. 2 is a configuration diagram for controlling particle positions on a horizontal axis as an example of control using a spike signal train. FIG. 2 is a phase diagram showing triple/double/single well potentials for particles moving on a horizontal axis. Simulation result diagram for entropy reduction and pattern formation functions among order emergence functions. A diagram showing simulation results regarding the function of enlarging the region of attraction to the target state among the order emergence functions. A simulation result diagram regarding the binding function to natural frequencies among the order emergence functions. FIG. 1 is a configuration diagram of a musculoskeletal robot control system, which is an example of a control system. Diagram of movement speed simulation results for a musculoskeletal robot control system in a low-friction environment. Coordination ability simulation result diagram in a low-friction environment for a musculoskeletal robot control system. FIG. 2 is a functional block diagram of a control system in which a spike signal train generation section and a drive signal generation section are external control devices.

Embodiments of the present invention will be described below with reference to the drawings. Various features shown in the embodiments described below can be combined with each other. In particular, in this specification, the term "unit" may include, for example, a combination of hardware resources implemented by circuits in a broad sense and software information processing that can be specifically implemented by these hardware resources. . In addition, various types of information are handled in this embodiment, and these information include digital signal information expressed by the high and low signal values as a binary bit collection consisting of 0 or 1, and continuous voltage/current information. Communication and calculations can be performed on circuits in a broad sense using analog signal information that changes over time and spike signal information that instantaneously generates voltage and current on the time axis.

In addition, a circuit in a broad sense includes at least a digital circuit, an analog circuit, an optical circuit, a circuit, a processor, a memory, etc. This is a circuit realized by combining them. That is, digital circuits include application specific integrated circuits (ASICs), programmable logic devices (for example, simple programmable logic devices (SPLDs)), and complex programmable logic devices (COMPLDs). ex Programmable Logic Device: CPLD), field programmable gate array (FPGA), and the like. Analog circuits include passive components such as resistors, capacitors, and inductors, and discrete semiconductors such as diodes, transistors, and thyristors. tor), and comparator It includes analog integrated circuits such as (Comparators) and the like. Further, a circuit configuration using a D/A converter (Digital-to-Analog Converter) or an A/D converter (Analog-to-Digital Converter) at the boundary between the digital circuit and the analog circuit is also possible. Furthermore, optical circuits include light emitting elements such as light emitting diodes and semiconductor lasers, photodetectors such as photodiodes, and optical fibers. Al Fiber) and other light guides It includes an optical waveguide, an optical integrated circuit, and the like.

1. Overall Configuration In Section 1, the overall configuration of a control system 1 including a control device according to the present invention will be explained using the drawings. FIG. 1 is a diagram showing an outline of the configuration of a control system 1 according to the present embodiment. The control system 1 is a system that includes a control device 2 and a controlled object 3, and these are electrically connected. Controlled objects 3 include robots such as bipedal walking (described later), mobile objects, pacemakers, electric circuit systems, chemical reaction systems, communication networks, socio-economic management systems, financial systems, biological networks, and the surrounding environment in terms of movement and state. It has the characteristic that its electrical, mechanical, or chemical internal state fluctuates due to changes, and requires control in order to perform the desired operation. The internal state is not limited to any item as long as it is related to the function or operation of the control system 1 and can be detected.

1.1 Control device 2
As shown in FIG. 1, the control device 2 includes a communication section 21, a storage section 22, and a control section 23, and these components are electrically connected through a communication bus 20 inside the control device 2. Each component will be further explained below.

<Communication Department 21>
The communication unit 21 is for exchanging information with the controlled object 3. Although wired communication means such as USB, IEEE1394, Thunderbolt, wired LAN network communication, etc. are preferable, wireless LAN network communication, mobile communication such as 5G/LTE/3G, Bluetooth (registered trademark) communication, etc. may be included as necessary. Good too. These are just examples, and a dedicated communication standard may be adopted. That is, it is more preferable to implement it as a set of these plurality of communication means.

Although FIG. 1 shows that the communication unit 21 is connected to the state detection unit 31 and drive unit 30 in the controlled object 3 separately, there is only one physical connection. It is also possible to have a configuration in which the data is logically distributed.

<Storage unit 22>
The storage unit 22 is a volatile or nonvolatile storage medium that stores various information. This can be used as a storage device such as a solid state drive (SSD), or as a random access memory that stores temporarily necessary information (arguments, arrays, etc.) related to program operations. It can be implemented as a memory such as RAM). Alternatively, a combination of these may be used.

In particular, the storage unit 22 stores various parameters related to control execution contents, individual characteristic information such as shape, size, material, and weight regarding the controlled object 3, and past setting information during continuous control including during optimization.

Furthermore, the storage unit 22 stores various programs related to the control device 2 that are executed by the control unit 23 . Specifically, for example, the operation procedure regarding the controlled object 3 having multiple driving elements such as multiple muscles/tendons and joints like a biped walking robot, the basic control signal generation unit 231 forming the control unit 23, and the spike signal These are the initial values and updating procedures of the parameter group used by the column generation unit 232 and the drive signal generation unit 233.

<Control unit 23>
The control unit 23 processes and controls the overall operation related to the control device 2. The control unit 23 is, for example, a central processing unit (CPU) not shown. The control unit 23 implements various functions related to the control device 2 by reading predetermined programs stored in the storage unit 22. Specifically, based on information given in advance for each controlled object 3 and state information received from the controlled object 3 internal state detection section 31 via the communication section 21, the basic control signal generation section 231 and the spike signal train generation section This corresponds to the function of generating the drive signal AS to the controlled object 3 through the drive signal generating section 232 and the drive signal generating section 233 and performing control.

That is, information processing by software (stored in the storage unit 22) is concretely realized by the hardware (control unit 23), so that the basic control signal generation unit 231, the spike signal train generation unit 232, and the drive It can be executed as the signal generation unit 233. Note that although FIG. 1 shows a single control section 23, the actual configuration is not limited to this, and may be implemented so as to have a plurality of control sections 23 for each function. Alternatively, a combination thereof may be used. The basic control signal generation section 231, spike signal train generation section 232, and drive signal generation section 233 will be described in further detail below.

[Basic control signal generation unit 231]
The basic control signal generation unit 231 is a unit in which information processing by software (stored in the storage unit 22) is concretely realized by hardware (control unit 23). The basic control signal generation unit 231 generates a basic control signal in the form of a non-spike signal based on state information obtained from the state detection unit 31 of the controlled object 3 via the communication unit 21 and parameters given in advance for each controlled object 3. It generates a control signal CS. The control algorithm is not limited, and various algorithms such as feedback control, feedforward control, model predictive control, and control using deep learning can be used.

Note that in order to make maximum use of the order generation function described below, it is desirable to set a control algorithm and parameters that do not have a strong natural frequency/natural frequency peak as the frequency characteristic of the basic control signal generation section 231 alone.

[Spike signal train generation unit 232]
In the spike signal train generation section 232, information processing by software (stored in the storage section 22) is specifically realized by hardware (control section 23), and the hardware includes the digital circuit and analog circuit described above. Consists of a combination of circuits.

The spike signal train generation unit 232 receives the basic control signal CS generated by the basic control signal generation unit 231 as input, and includes neurons (not shown in FIG. 1) which are elements that generate the spike signal train ST. . The spike signal train generation unit 232 operates in the same manner as a neuron network that stochastically generates impulse-like action potentials in a living body, that is, a stochastically spiking neuron network (sSNN). Neurons in the spike signal train generation unit 232 include LIF (leaky integrate-and-fire) neurons, Poisson spike models, Hodgkin-Huxley models, models capable of burst firing, etc. A model that generates a spike signal train at a timing defined by a basic control signal, an internal state including disturbances, and dynamics related to the internal state is applicable.

When a plurality of neurons exist in the spike signal train generation unit 232 (sSNN), it is also possible to have arbitrary synaptic connections between these neurons. At that time, all neurons are designed so that they do not fire synchronously. Specifically, for example, each neuron receives independent noise, or sets a different firing threshold (described later) or reset potential (described later) for each neuron.

Here, the case of the LIF neuron will be explained in more detail using mathematical formulas. The j-th LIF neuron in the i-th spike signal train generation unit 232 (sSNN) can be expressed as [Equation 1], [Equation 2], and [Equation 3] described below.



Here, v_ij is the electric potential, dots on variables are time differentials, γ is the attenuation coefficient, b_ij is the bias input (current), I is the input signal, D_i is the noise intensity in the i-th spike signal train generation section 232, and ξ_ij is Gaussian (normal distribution) noise of unit strength, v^θ is a firing threshold, v_ij^R is a reset potential, τ_ref is a disabled period, and k is a spike signal generation order number.

It fires when the internal potential v_ij of the target LIF neuron reaches v^θ, generates a spike signal that is mathematically described by Dirac's delta function δ, and resets the internal potential to v_ij^R. and enters the disabled period of τ_ref. In the LIF neuron shown in [Equation 1], disturbance in the form of Gaussian noise is added to prevent a large number of LIFs from firing synchronously. As is clear from [Equation 1], synchronous firing can be prevented by individually setting the firing threshold v^θ and the reset potential v_ij^R in addition to applying the noise component. The spike signal train σ_ij in each LIF neuron can be formally described by [Equation 4].

At this time, if a plurality of LIF neurons exist in the i-th spike signal train generating section 232, all the spike signal trains output from each LIF neuron are utilized. In this case, a method using the average of the spike signal train or a method of linearly weighting each LIF neuron can be applied. The new spike signal train thus obtained is defined as the spike signal train ST, which is the output of the spike signal train generating section 232.

That is, the spike signal train generating section 232 has a plurality of neurons forming a neuron network, and is configured to be able to generate a spike signal train ST based on signals output by each of the plurality of neurons.

[Drive signal generation unit 233]
In the drive signal generation section 233, information processing by software (stored in the storage section 22) is concretely realized by hardware (control section 23), and the hardware includes the digital circuit and analog circuit described above. Consists of a combination of

The drive signal generation section 233 has a function of converting the spike signal train ST generated by the spike signal train generation section 232 into a drive signal to be supplied to the drive section 30 of the controlled object 3. This drive signal is generated based on a postsynaptic potential (PSP) in a biological neuron network.

A synapse-like system can be used to generate the drive signal AS, which is equivalent to a postsynaptic potential, from the spike signal train ST. Specifically, these include a low-pass filter, a classical α-function synapse model, a square wave synapse model, and a dynamic synapse model.

When the signal processing method in the drive signal generation unit 233 is a low-pass filter, the postsynaptic potential (PSP) y_i, which is the basis of the output, and the activation level A_i of the controlled object 3 drive unit 30 are each expressed as [Equation 5] [ 6].


Here, τ_s is a synaptic time constant, N is the number of LIF neurons in the target i-th spike signal train generating section 232, g_i^A is an amplification gain, and A_i^0 is an offset.

Note that in FIG. 1, the drive signal AS is transmitted through a direct connection line between the drive signal generation section 233 and the drive section 30, but after A/D (Analog-to-Digital) conversion, the drive signal AS is transmitted as a digital signal to the communication section 21. A configuration in which the signal is transmitted to the drive unit 30 via the transmission line is also possible.

1.2 Controlled object 3
Specifically, the controlled object 3 is, for example, a robot or a movement system that performs a mechanical task. Note that the control device 2 of the present invention is not limited to the technical field of the task, but can be applied to electrical circuit systems or chemical reactions as long as it acts dynamically and can detect changes in operating state due to environmental changes. The system can also be the controlled object 3.

[Drive section 30]
The drive unit 30 operates the controlled object 3 based on an external control signal when executing a task. Specifically, for example, when the controlled object 3 is a robot, it may be a motor, a pneumatic/hydraulic actuator, etc., but is not limited to these.

[State detection unit 31]
The state detection unit 31 detects the internal state of the controlled object 3, including when there is an environmental change due to a disturbance or the like when the controlled object 3 is in operation. The internal state includes mechanical information such as the position, velocity, acceleration, rotation, angular velocity, angular acceleration, force and moment of the point of interest in the controlled object 3, electrical information such as voltage, current and resistance, sound and These include, but are not limited to, physical information of light, fluid dynamics information such as temperature, pressure, and flow rate, and chemical information such as concentration, pH, and molecular weight. The detected internal state is configured to be able to be transmitted to the communication unit 21 as state information. In other words, the state information is information indicating an internal state that changes due to the behavior of the controlled object and environmental fluctuations at a point of interest in the controlled object.

2. Optimization Method for Control System 1 In Section 2, a parameter optimization method for the control device 2 in the control system 1 will be described. Here, for example, if the controlled object 3 is a bipedal walking robot, a task such as moving to a specified point is executed, but the optimization method of the present invention depends on the controlled object 3 and the type of its work. It is not limited.

FIG. 2 shows an optimization flow when optimizing the parameters of the control device 2 by repeatedly executing a basic task a plurality of times in an environment with few disturbances.

[Start optimization]
(Step S1)
Each parameter group in the basic control signal generation section 231, spike signal train generation section 232, and drive signal generation section 233 is initialized. Information stored in the storage unit 22 can be used as the parameter values used for initialization. The information stored in the storage unit 22 includes not only information that is continuously stored in a non-volatile manner, but also information input from the outside by the user at the start of work, taking into account the individual characteristics of the controlled object 3 and the external environment situation. Also included.

The basic control signal generation unit 231 initializes parameters used in the adopted algorithm. Specifically, for example, in PID (Proportional-Integral-Differential) control, which is a type of feedback control, there are proportional gain K_P, integral gain K_I, differential gain K_D, etc.

The spike signal train generation unit 232 initializes all parameters for the stochastic spiking neuron network (sSNN). For example, when the spike signal train generation unit 232 is configured with LIF neurons, the number of neurons N and the variables included in [Equation 1] and [Equation 2], specifically, the attenuation coefficient γ, bias input b_ij, noise intensity D_i, firing These include the threshold value v^θ, the reset potential v_ij^R, and the disabled period τ_ref.

The drive signal generation unit 233 initializes parameters related to the method adopted to generate the drive signal AS from the spike signal train ST. For example, when a low-pass filter method is adopted, the time constant τ, passband gain, etc. are set as filter characteristic values. When all or part of the drive signal generation section 233 is configured as an electrical analog circuit, it is also possible to set the resistance value of a resistor, the capacitance value of a capacitor, etc. in advance in a fixed or semi-fixed manner. .

(Step S2)
Each parameter in the basic control signal generation section 231, spike signal train generation section 232, and drive signal generation section 233 is updated. All parameters can be updated, but since the general directionality of the control section 23 as a whole depends largely on the basic control signal CS, it is desirable to optimize the basic control signal generation section 231 as the main target. Regarding the spike signal train generation section 232 and the drive signal generation section 233, a control method may be adopted in which, for example, the number of neurons N, the synaptic time constant τ_s, and the noise intensity D are updated, and other parameters related to the synaptic connection strength are not updated. It is possible. Each time it is determined that convergence has not occurred in step S4, which will be described later, each of the parameters is updated.

(Step S3)
The controlled object 3 executes basic tasks under control from the control device 2. The state detection unit 31 detects the various internal states described above in the controlled object 3 and transmits state information to the communication unit 21 within the control device 2 . The basic control signal generation unit 231 calculates an evaluation value for the controlled object 3 as a whole. Each time it is determined that convergence has not been achieved in step S4, which will be described later, the number of executions of the basic task increases, such as the first, second, and third times.

(Step S4)
The basic control signal generation unit 231 determines whether the entire system has converged. If it is determined that the process has not converged (NO), the process returns to step S2 and continues from the parameter update work. As a learning algorithm when updating parameters, an evolutionary strategy such as a genetic algorithm can be applied, but the learning algorithm is not limited thereto. If it is determined that the process has converged (YES), the work is finished.
[Optimization finished]

3. Order Emergence Function In Section 3, the order emergence function possessed by the control device 2 in the configuration of the present invention will be described in detail. This is done by first generating a spike signal sequence ST using a spike signal sequence generation unit 232 (sSNN) from the basic control signal CS generated by the basic control signal generation unit 231, and then driving it using the drive signal generation unit 233. This is a function unique to the control device 2 of the present invention, which has a configuration for generating the signal AS, and is not known in the past.

In this section, as an example showing the order emergence function, a case with two stochastic spiking neuron networks (sSNN) is shown in FIG. 3. Here, it is assumed that the position of particles existing on the horizontal axis in FIG. 3 is controlled. In FIG. 3, S_0 and S_1 are sSNNs, which include the spike signal train generation section 232 and drive signal generation section 233 of the present invention. The two sSNNs each receive I_0(t) and I_1(t) as input signals. Here, the input signal I_*(t) is defined as the amount of difference between the current position x(t) of the particle and the target positions x_0^g and x_1^g. As the spike signal train generation section 232, an LIF neuron (see Section 1) is used, and as the drive signal generation section 233, a low-pass filter (see Section 1) is used.

3.1 Entropy reduction/pattern formation function FIG. 4A shows a case where the position of a particle with mass in a triple well potential function is controlled by sSNN. Note that although the center (x=0) is the minimum value of the potential, it is not the minimum value, and there are locations on both sides of the center where the potential is minimum.

The simulation results of the particle position movement state under the environment of FIG. 4A are shown in FIGS. 5A and 5B. The horizontal axis shows time t, and the vertical axis shows particle position x. Further, FIG. 5A shows the case where the number of neurons is N=2, and FIG. 5B shows the case where the number of neurons is N=150. ApEn is the entropy (approximate entropy) calculated from the moving state. In Figure 5A, where the number of neurons is small, the absolute value of the amount of movement of the particle is larger than in Figure 5B, but this is because the particle moves periodically between two potential minimum positions, and its regularity. Due to the periodicity, the entropy ApEn has a small value. In this way, in a stochastic spiking neuron network (sSNN), the higher the spiking property, the more entropy reduction function and pattern formation function are expressed.

3.2 Target state attraction region expansion function Figure 4B shows a case where the position of a particle with mass in a double well potential function is controlled by sSNN. Here, the center (x = 0) is the maximum potential value, and like the mountain car task, it is not possible to directly reach the maximum potential value x = 0 from the bottom of the valley, and the reaction The problem should be set up in such a way that it requires the help of external forces.

A simulation was performed by varying the initial position x_0 and initial velocity v_0 of the particle and defining the case where the particle could stay within the range [-0.1, 0.1] near the center for a certain period of time as the attraction region. The results are shown in FIGS. 6A and 6B. 6A shows the case where the number of neurons is N=1, and FIG. 6B shows the case where the number of neurons is N=100. The white area in FIGS. 6A and 6B is the pull-in area. Further, FIG. 6C shows a simulation result of the draw-in area ratio (basin rate) when the number N of neurons is changed using the bias input b as a parameter. As is clear from FIGS. 6A, 6B, and 6C, the stochastic spiking neuron network (sSNN) has the ability to expand the attraction area, and the smaller the number of neurons included in the sSNN, the more the ability to expand the attraction area. It is often strongly expressed.

3.3 Binding function to natural frequency Figure 4C shows the case where the position of a particle with mass in a spring-mass system is controlled by sSNN. Here, the spring mass system is equivalent to a single double well potential function. A spring-mass system that does not perform any control has a natural frequency f_0 (natural frequency) determined by the spring constant k and the particle mass m. It is known that when normal feedback control is performed on the spring mass system, the natural frequency f_0 is modulated due to the influence of the gain of the feedback control.
FIG. 7 shows the results of simulating the SNR (signal-to-noise ratio) for the natural frequency f_0 by performing control using sSNN with the center position (x^g=0) as the target. In FIG. 7, the horizontal axis is the synaptic time constant τ_s, and the vertical axis is the amplification gain g^A, indicating that the whiter region has a higher SNR. A resonance phenomenon to the natural frequency f_0 can be confirmed in a very wide parameter range where the natural frequency f_0 is 1 to 10 Hz. Furthermore, since the white striped pattern extends vertically in many areas, it is clear that when driving a spring-mass system using a stochastic spiking neuron network (sSNN), it has almost no effect on the natural frequency f_0. It's obvious.

4. Robot Control System In Section 4, as an embodiment, simulation results of a biped walking robot control system using a robot, more specifically a musculoskeletal robot, as the controlled object 3 will be described.

Figure 8 shows a functional schematic diagram of the musculoskeletal robot control system used in the simulation. The left side of FIG. 8 shows the structure of the skeleton (links), joints, and muscles (muscle lines in the figure, some of which are omitted). The robot drive unit 30 includes eight muscles and multi-joint muscles for each leg. are connected. In addition, as a robot state detection unit 31 (Sensory Input in Figure 8), a 9-axis inertial measurement device ( It is configured to be able to measure 3-axis attitude, 3-axis acceleration, and 3-axis angular velocity obtained by Inertia Measurement Unit).

Each leg has multiple phases, such as rest and swinging, and each phase has different reflex activation rules. The reflex activation rule (Reflex System: basic control signal generation unit 231 in FIG. 8) is a combination of positive feedback control rules for generated force, feedback control rules for muscle length, and PD (proportional differential) control rules for joint angles or body posture. Constructed with.

FIGS. 9A, 9B, and 9C show simulation results of center-of-gravity movement speed in a slippery, low-friction environment. The horizontal axis of FIGS. 9A and 9B is time t, and the vertical axis is the center of gravity moving speed v^g. Normally (broken line in the figure) the friction coefficient μ is set to 10, but in a low friction environment (solid line in the figure) the friction coefficient μ is set as low as 0.04 at time t=[10,40]. From FIGS. 9A and 9B, stable bipedal walking motion is performed in both the normal environment and the low-friction environment. At that time, the overall speed is shifted to the lower speed side in the low friction environment than in the normal environment. FIG. 9C shows the relationship between gait frequency and amplitude, which shifts to the lower frequency side in a low friction environment.

Figure 10 shows the results of a simulation of coordination ability in a situation where a large slip occurred. In FIGS. 10A and 10C, the horizontal axis x indicates the position, where x=[4,16] is a low friction section and the friction coefficient μ=0.04, and the other areas are normal and the friction coefficient μ=10. Note that the sampling interval in FIG. 10A is 0.1 s, and the sampling interval in FIG. 10C is 0.25 s. The black band on the low-friction section (slippery band in the figure) indicates slipping of the right foot, and the gray band indicates slipping of the left foot. FIGS. 10B and 10D show the relationship between time t and center of gravity moving speed v^g corresponding to FIGS. 10A and 10C, respectively.

In FIGS. 10A and 10B, although a slip of about 0.5 m occurs for a period of 0.5 s or more, the patient is able to adapt and continue walking. At this time, it should be noted that the distances slid by the left and right feet are asymmetrical. In addition, in Figures 10C and 10D, v^g drops extremely near the end point of the low-friction section at x = 16 m, t = 13 s, and this is due to a slightly longer right foot slip just before the end of the low-friction section. This indicates that the person was on the verge of falling when entering the vehicle. Even in this situation, the patient is able to return to normal walking using the fact that his toes are in contact with the hard-to-slip ground in the normal section (μ=10) as a foothold. In this way, the robot control system using the control device 2 according to the present invention has extremely high adaptability, which has been difficult to achieve with conventionally known control using only reflection circuits.

For cooperative motion in the biped walking robot control system described in this section, it is necessary to reduce the entropy of the motion sequence. Furthermore, to avoid falling, it is also necessary to control ZMP (zero-moment point) within a certain range. As explained in Section 3, the order generation function (see 3.1 Entropy reduction/pattern formation function, 3.2 Target state attraction area expansion function) of the control device 2 of the present invention works effectively. Therefore, it can be said that an immediate fall avoidance function is realized.

5. Modifications The present embodiment may be further modified by the following aspects.

In Section 4, the embodiment of the bipedal walking robot control system was described. Generally speaking, a mobile system involves changes in the external environment as it moves. You can take advantage of it. Furthermore, the order generation function can be utilized either as an unmanned fully autonomous control system or as an auxiliary control system for a manned system. Specific examples of mobile systems include multilegged robots, wheeled and caterpillar robots, unmanned aerial vehicles (UAVs, drones), unmanned surface vehicles (USVs), and unmanned submersibles (unmanned submersibles). Underwater vehicles (UUV), automobiles including self-driving vehicles, aircraft, ships, etc., but are not limited to these.

Furthermore, the present invention can also be applied to industrial, medical, agricultural, and household robots in which work objects are frequently changed in terms of transportation and processing.

The order generation function of the control device 2 according to the present invention is not limited to any technical field as long as the control system 1 needs to adapt to changes in state. In other words, it can be applied not only to mechanical motion, but also to electrical fluctuations or chemical reaction fluctuations. Furthermore, it can be expected to be applied to control of financial systems, control of inflow, outflow, and propagation of information in communication networks such as the Internet, and air conditioning systems. Therefore, it is also possible to apply it, for example, to a cardiac pacemaker or a heart-lung machine that utilizes the natural frequency binding function (see 3.3) among the order emergence functions.

The spike signal train generation section 232 and the drive signal generation section 233 in the present invention can be configured as one control device 2 together with the basic control signal generation section 231, the communication section 21, the storage section 22, etc., as shown in FIG. However, it is also possible to use the spike signal train generation section 232 and the drive signal generation section 233 as external control devices.

FIG. 11 shows a functional block diagram. In FIG. 11, 2b is a basic control device that generates a basic control signal CS, and 3 is a controlled object. Conventional control is possible with just the combination of the basic control device 2b and the controlled object 3, but the external control device 2a is connected to supplement the conventional control. A spike signal train generation section 232 and a drive signal generation section 233 are provided in the external control device 2a. By adding the external control device 2a to an existing control system, it becomes possible to utilize the order generation function described in Section 3, and the functions and performance of the control system 1 can be improved.

Furthermore, spike signal trains are not limited to those generated stochastically due to disturbances, but can also be spike signal sequences generated deterministically using chaos etc. if they contain sufficient complexity and unpredictability. However, you can get the same functionality.

6. Conclusion As described above, according to the present embodiment, even if an unexpected change in the external environment occurs, the control device 2 reacts autonomously and enables the entire control system 1 to perform the desired operation. can be carried out.

The control device 2 is configured to be able to control the controlled object 3 by supplying a drive signal to the controlled object 3, and includes a spike signal train generation section 232 and a drive signal generation section 233, and includes a spike signal train generation section 232 and a drive signal generation section 233. The train generation unit 232 is configured to be able to generate a spike signal train ST at a timing defined by the basic control signal CS for controlling the controlled object 3 and an internal state including disturbance, and the drive signal generation unit The drive signal AS is configured to be able to generate the drive signal AS that continuously changes in time series based on the spike signal train ST.

Moreover, this makes it possible to implement the following control system 1.

Such a control system 1 includes a controlled object 3 and a control device 2 that controls the controlled object 3,
The controlled object 3 is at least one of a robot, a moving body, a pacemaker, an electric circuit system, and a chemical reaction system, and the control device 2 is the control device 2 described above.

Software for implementing the control device 2 or the control system 1 as hardware can also be implemented as a program. Such a program may be provided as a computer-readable non-temporary recording medium, may be provided for download from an external server, or may be provided by running the program on an external computer. So-called cloud computing, in which each function can be performed on a client terminal, may be implemented.

This control program is for controlling a controlled object and causes a computer to execute a spike signal train generation function and a drive signal generation function. According to the spike signal train generation function, the control object 3 is A spike signal train ST is generated at a timing defined by a basic control signal CS for control and an internal state including disturbance, and according to the drive signal generation function, the spike signal train ST is chronologically generated based on the spike signal train ST. The drive signal AS that continuously changes is generated.

Finally, although various embodiments according to the present invention have been described, these are presented as examples and are not intended to limit the scope of the invention. The new embodiment can be implemented in various other forms, and various omissions, substitutions, and changes can be made without departing from the gist of the invention. The embodiment and its modifications are included within the scope and gist of the invention, and are included within the scope of the invention described in the claims and its equivalents.

1: Control system 2: Control device 2a: External control device 2b: Basic control device 20: Communication bus 21: Communication section 22: Storage section 23: Control section 231: Basic control signal generation section 232: Spike signal train generation section 233 : Drive signal generation unit 3 : Control object 30 : Drive unit 31 : State detection unit CS : Control signal ST : Spike signal sequence AS : Drive signal I : Input signal ApEn : Entropy N : Number of neurons x_0 : Initial position v_0 : Initial velocity b: Bias input f_0: Natural frequency τ_s: Synaptic time constant g^A: Amplification gain v^g: Center of gravity movement speed μ: Friction coefficient

Claims (6)

A control device,
comprising a basic control signal generation section, a spike signal train generation section, and a drive signal generation section,
The basic control signal generation unit is configured to be able to generate a non-spike basic control signal for controlling the controlled object based on state information of the controlled object,
The state information is information indicating an internal state that changes due to the behavior of the controlled object and environmental changes at a point of interest in the controlled object,
The spike signal train generation unit has a plurality of neurons forming a neuron network, and is configured to be able to generate a spike signal train based on spike signals generated from each of the plurality of neurons,
Each of the neurons is configured to generate the spike signal based on a disturbance and a firing threshold of the neuron by inputting the basic control signal;
The drive signal generation unit is configured to be able to generate a drive signal that continuously changes in time series based on the spike signal train, wherein the drive signal corresponds to a postsynaptic potential in the neuron network, and the drive signal corresponds to a postsynaptic potential in the neuron network. configured to be able to control the control target by being supplied to the target,
At least the parameters included in the basic control signal generation unit are configured to update the parameters, cause the control target to execute a preset basic task based on the parameters, and execute the basic task. updated by repeatable optimization, including calculating an evaluation value based on the results ;
Control device. The control device according to claim 1,
Additionally equipped with a communications department,
The communication unit is configured to be able to receive the status information.
Control device. The control device according to claim 1 or 2,
The neuron is configured to be able to generate the spike signal using the generation of stochastic impulse-like action potentials in a living body as a model.
Control device. A control system,
comprising a controlled object and a control device that controls the controlled object,
The controlled object is at least one of a robot, a mobile object, a pacemaker, an electric circuit system, a chemical reaction system, a communication network, a socio-economic management system, a financial system, a biological network, and an animal or plant,
The control device is the control device according to any one of claims 1 to 3,
control system. The control system according to claim 4,
The controlled object is a musculoskeletal robot, and includes a robot drive unit and a robot state detection unit,
The robot driving unit includes a plurality of bones, a plurality of joints, a muscle that applies a pulling force between the adjacent bones, and/or a multi-joint muscle that applies a pulling force across the plurality of bones,
The robot state detection unit is configured to be able to detect at least one state of muscle generation force, muscle length, joint angle, body posture, center of gravity position, sole reaction force, 3-axis posture, 3-axis acceleration, and 3-axis angular velocity. be done,
control system. A control program,
causing the computer to function as each part of the control device according to any one of claims 1 to 3;
control program.