JP2008282159A - Multi-input/multi-output-system control device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To excellently handle a problem of a multi-input/multi-output control system by use of a comparatively simple linear model with respect to a multi-input/multi-output-system control device. <P>SOLUTION: The linear model 50 with a current Ib applied to a secondary battery 28 and a cooling temperature Tc for cooling the secondary battery 28 as model input information and with a battery temperature Tb and battery power Pb of the secondary battery 28 as model output information is constructed inside an ECU (Electronic Control Unit) 30. A model parameter a11 or the like inside a linear expression (refer to an expression (1)) expressing the linear model 50 is acquired by sequential learning using the linear expression. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a multi-input multi-output control apparatus suitable for handling a multi-input multi-output control system.

  Conventionally, for example, Patent Document 1 discloses an autonomous mobile robot travel control device including a neural network model that learns the characteristics of a control system with multiple inputs and multiple outputs. In this conventional technique, the speed and angular velocity are controlled using a neural network model so that the target value of the speed and angular velocity of the autonomous mobile robot can be realized.

JP-A-6-337713 WO2003 / 103164 JP-A-7-59269 JP 2006-105821 A JP-A-8-147038 Japanese Patent Laid-Open No. 7-83949 JP 2005-233346 A

  By the way, the neural network model includes nonlinear elements. For this reason, according to the control method using the neural network model, although good control accuracy can be obtained, a long learning time is required and the neural network model is a resource (ROM or RAM) of the computer (ECU). Will occupy a lot.

  The neural network model has the above problems. For this reason, it is only necessary to describe the control target of a multi-input multi-output control system using a simple linear model, but the conventional linear model is an effective model for a multi-input multi-output control system. It was difficult to construct and control.

  The present invention has been made in order to solve the above-described problems, and uses a relatively simple linear model to make it possible to satisfactorily handle the problems of a multi-input multi-output control system. An object of the present invention is to provide a control device.

A first invention is a control device for a multi-input multi-output system including a linear model that handles a plurality of inputs and outputs,
The linear model includes a current value of a plurality of model input information whose coefficients are model parameters to be learned, a past value of a plurality of model input information whose coefficients are other model parameters to be learned, and a learning target. The sum of the current value of a plurality of model output information using the other model parameter as a coefficient and the past value of the plurality of model output information using the other model parameter as a coefficient as a coefficient, Includes line formats representing multiple model output information,
The control device of the multi-input multi-output system is
Model input / output information acquisition means for acquiring the plurality of model input information and the plurality of model output information;
Parameter calculation means for calculating the model parameter based on the linear format for a predetermined sampling period into which the plurality of model input information and the plurality of model output information are substituted,
It is characterized by providing.

In a second aspect based on the first aspect, the multi-input / multi-output system control device comprises:
The target model output value is obtained based on a control expression derived from an equation representing an evaluation function that determines the responsiveness of the system until the target model output value is reached after a predetermined time has elapsed and the linear form. Control amount calculation means for calculating the model input information for
Control means for controlling the model input information so as to be the model input information calculated by the control amount calculation means;
Is further provided.

According to a third aspect of the present invention, in the first or second aspect, the multi-input multi-output system control device controls a state of a secondary battery,
The model input information is a parameter for determining the state of the secondary battery,
The model output information is a parameter indicating a state of the secondary battery.

In a fourth aspect based on the third aspect, the parameters for determining the state of the secondary battery are the charging current of the secondary battery, the discharging current of the secondary battery, and the cooling of the secondary battery. Temperature,
The parameters indicating the state of the secondary battery are the power of the secondary battery, the temperature of the secondary battery, and the charge amount of the secondary battery.

According to a fifth aspect of the present invention, in the first or second aspect, the multi-input multi-output system control device is a control device for controlling an electric motor,
The model input information is a parameter for determining the operating state of the electric motor,
The model output information is a parameter indicating an operating state of the electric motor.

In a sixth aspect based on the first or second aspect, the multi-input multi-output system control device is a control device for controlling an autonomous mobile robot,
The model input information is a parameter for determining the driving state and course of the autonomous mobile robot,
The model output information is a parameter indicating a driving state or a course of the autonomous mobile robot.

  According to the first invention, the linear model can be adapted to state changes such as changes in the system over time and environmental changes surrounding the system by sequential learning of model parameters. For this reason, according to the present invention, it is possible to satisfactorily handle the problem of a multi-input multi-output control system using a relatively simple linear model.

  According to the second aspect of the present invention, in order to satisfy the target model output value after a predetermined time has elapsed from the current time, based on the control expression in which the sequentially learned model parameters are reflected (that is, the system state is reflected). A plurality of model input information required at the present time is calculated. And a predetermined actuator is controlled so that it may become the model input information concerned. Thus, according to the present invention, smooth and accurate control can be realized for the target value by adaptive control using a linear model.

  According to the third aspect of the present invention, in the control system for a secondary battery with multiple inputs and multiple outputs, smooth and accurate control can be realized with respect to the target value by adaptive control using a linear model.

  According to the fourth invention, the secondary battery charging current, discharging current, and cooling temperature are used as model input information, and the secondary battery power, temperature, and charge amount are used as model output information. The battery can be controlled smoothly and accurately with respect to the target value.

  According to the fifth aspect of the present invention, smooth and accurate control can be realized with respect to the target value by adaptive control using a linear model in a control system for a multi-input multi-output motor.

  According to the sixth invention, in the control system of a multi-input multi-output autonomous mobile robot, it is possible to realize smooth and accurate control over the target value by adaptive control using a linear model.

Embodiment 1 FIG.
[System configuration]
FIG. 1 is a diagram for explaining a hybrid vehicle drive system to which a multi-input multi-output control device according to Embodiment 1 of the present invention is applied. The drive system 10 includes an internal combustion engine 12 and a vehicle drive motor (hereinafter simply “motor”) 14 as a power source of the vehicle. The drive system 10 also includes a generator 16 that receives power and generates electric power.

  The internal combustion engine 12, the motor 14, and the generator 16 are connected to each other via a power split mechanism 18. A reduction gear 20 is connected to the rotating shaft of the motor 14 connected to the power split mechanism 18. The speed reducer 20 connects the rotation shaft of the motor 14 and the drive shaft 24 connected to the drive wheels 22. The power split mechanism 18 is a device that splits the driving force of the internal combustion engine 12 into the generator 16 side and the speed reducer 20 side.

  The motor 14 and the generator 16 are connected to a secondary battery (battery) 28 via an inverter 26, respectively. The electric power generated by the generator 16 can be supplied to the motor 14 via the inverter 26, or the secondary battery 28 can be charged via the inverter 26. Further, the electric power charged in the secondary battery 28 can be supplied to the motor 14 via the inverter 26.

  The drive system 10 shown in FIG. 1 is controlled by an ECU (Electronic Control Unit) 30. The ECU 30 comprehensively controls the entire drive system 10 including the internal combustion engine 12, the motor 14, the generator 16, the power split mechanism 18, the inverter 26, the secondary battery 28, and the like.

  The secondary battery 28 is configured by connecting a plurality of modules in series. Each module is configured by connecting a plurality of cells in series. For each module, the secondary battery 28 is provided with a voltage sensor 32 that measures the voltage of each module and a battery temperature sensor 34 that measures the temperature of each module. These sensors 32 and 34 are connected to the ECU 30. The ECU 30 is also connected to a current sensor 36 that measures the current flowing through the secondary battery 28.

  Further, the drive system 10 shown in FIG. 1 includes a cooling fan 38 for cooling the secondary battery 28 that generates heat along with charge and discharge. The ECU 30 controls the temperature of the secondary battery 28 by controlling the operation of the cooling fan 38 based on the temperature information of the secondary battery 28. The ECU 30 is connected to a temperature sensor 40 that measures the temperature of air blown from the cooling fan 38 to the secondary battery 28.

[Linear model of Embodiment 1]
FIG. 2 is a diagram showing the linear model 50 used in the first embodiment of the present invention. A linear model 50 shown in FIG. 2 is a mathematical model virtually constructed in the ECU 30. More specifically, the linear model 50 uses the current Ib applied to the secondary battery 28 and the cooling temperature (cooling air temperature by the cooling fan 38) Tc for cooling the secondary battery 28 as model input information. This is a model for calculating (outputting) battery power Pb and battery temperature Tb (model output information) of the secondary battery 28.

ただし、上記（１ａ）〜（１ｃ）式において、ｔは制御開始時点（ｔ＝０）からのサンプリング時間（サンプリング回数）であり、ｄは所定の遅れ時間であり、ａ１１、ａ１２、ｂ１１、ｂ１２などは何れも逐次学習されるモデルパラメータである。 That is, the linear model 50 according to the present embodiment inputs the parameters (actuator) for determining the battery state (battery voltage, battery charge state, battery temperature) with the secondary battery 28 as a control target. It is a linear model of a two-input two-output system that outputs the indicated parameters. In this embodiment, the linear model 50 is expressed as the following equation (1a).

However, in the above formulas (1a) to (1c), t is a sampling time (sampling number) from the control start time (t = 0), d is a predetermined delay time, and a11, a12, b11, b12 Are model parameters that are sequentially learned.

  When the determinant of the above (1a) is expanded, it can be expressed as the above (1b) and (1c). According to the equation (1b), the battery power Pb (t + d) at the time point delayed by the time d is the current battery power Pb (t) with the parameter a11 as a coefficient, and the previous time with the parameter a21 as a coefficient. Battery power Pb (t−1), current battery current Ib (t) with parameter b11 as a coefficient, current cooling temperature Tc (t) with parameter b12 as a coefficient, and previous time with parameter b21 as a coefficient Battery current Ib (t−1) and the previous cooling temperature Tc (t−1) with the parameter b22 as a coefficient can be expressed in a linear form.

  Further, according to the above equation (1c), similarly, the battery temperature Tb (t + d) at the time point delayed by time d is the current battery temperature Tb (t) with the parameter a12 as the coefficient, and the parameter a22 is the coefficient. The previous battery temperature Tb (t−1), the current battery current Ib (t) with the parameter b13 as a coefficient, the current cooling temperature Tc (t) with the parameter b14 as a coefficient, and the parameter b23 as a coefficient As the sum of the previous battery current Ib (t−1) and the previous cooling temperature Tc (t−1) with the parameter b24 as a coefficient.

The model parameters a11, b11 and the like included in the linear model 50 as described above can be calculated by a method described with reference to the following equation (2). Here, the battery power Pb will be described as an example, but the same applies to the battery temperature Tb.

  The above equation (1b) for each predetermined sampling period is the control value of the battery current Ib (t) and the cooling temperature Tc (t) during the system operation of the present embodiment, as shown in order as the above equation (2). Alternatively, by using the measured values of the battery power Pb (t) and the battery temperature Tb (t), it can be expressed as a linear equation having six model parameters a11 and the like as variables. Therefore, as shown as the above equation (2), if a linear equation corresponding to six samplings is acquired, six model parameters a11 and the like can be calculated.

  In the method shown in the above equation (2), for example, after calculating the model parameter a11 and the like using six linear equations whose left side is Pb (d) to Pd (5 + d), then the left side is Pb (1 + d) to The linear equations used to calculate the model parameters a11 and the like are sequentially shifted along the time series for each sampling period, such as calculating the model parameters a11 and the like using six linear equations of Pb (6 + d). According to such a calculation method, the model parameter a11 and the like can be sequentially calculated for each sampling period.

  In the method shown in the above equation (2), the model parameter a11 and the like sequentially calculated as described above are identified by the weighted least square method. According to such a method, it is possible to stably calculate the model parameter a11 and the like while suppressing the memory consumption of the ECU 30. In the present embodiment, such sequential learning of the model parameter a11 and the like is continued during system operation. As a result, the model parameter a11 and the like are sequentially identified (rewritten) by the above-described weighted least square method so as to adapt to changes in the state such as changes in the system over time and environmental changes surrounding the system. As a result, the change in the characteristics of the secondary battery 28 can be learned by the model parameter a11 or the like.

[Construction of control system using linear model of embodiment 1]
In the present embodiment, the battery current Ib (t (t) that can satisfy the target battery power Pb_target and the target battery temperature Tb_target by the following method using the identification result of the model parameter a11 and the like of the linear model 50 described above. ) And the cooling temperature Tc (t).

In the present embodiment, first, an evaluation function E expressed as the following equation (3) is introduced. The first term on the right side of the evaluation function E is the difference between the output values Pb (t) and Tb (t) of the linear model 50 and the target values Pb_target and Tb_target. The second term on the right side is the difference between the current value and the previous value for the input values Ib (t) and Tc (t) of the linear model 50. The difference between the input values is multiplied by a constant γ.

このような（４）式によれば、現時刻ｔからｄ秒後に電池電力Ｐｂや電池温度Ｔｂの目標値Ｐｂ_target、Ｔｂ_targetに到達するまでのシステムの応答性を決めることができる。より具体的には、上記（４）式中の定数γの値をゼロに近づけるようにした場合には、電池電力Ｐｂや電池温度Ｔｂがそれらの目標値Ｐｂ_target、Ｔｂ_targetに比較的急激に到達するような応答速度をシステムに与えることができるようになる。また、定数γの値を正方向に大きくした場合には、電池電力Ｐｂや電池温度Ｔｂがそれらの目標値Ｐｂ_target、Ｔｂ_targetに比較的緩やかに到達するような応答速度をシステムに与えることができるようになる。このように、定数γは、制御の応答性の味付けを決定するパラメータである。 Next, when the evaluation function E of the above equation (3) is set to zero, the following equation (4) can be obtained.

According to the equation (4), it is possible to determine the responsiveness of the system until reaching the target values Pb_target and Tb_target of the battery power Pb and the battery temperature Tb after d seconds from the current time t. More specifically, when the value of the constant γ in the above equation (4) is made close to zero, the battery power Pb and the battery temperature Tb reach their target values Pb_target and Tb_target relatively rapidly. Such a response speed can be given to the system. Further, when the value of the constant γ is increased in the positive direction, it is possible to give the system a response speed such that the battery power Pb and the battery temperature Tb reach the target values Pb_target and Tb_target relatively slowly. become. Thus, the constant γ is a parameter that determines the seasoning of control responsiveness.

In the present embodiment, by substituting the above equation (4) into the linear model 50 of the above equation (1a) and arranging it, a control equation such as the following equation (5) is obtained. In the present embodiment, the target value Pb_target, Tb_target, the current value (such as Ib (t)) and the past value (such as Ib (t-1)) and the past value (such as Ib (t-1)) of the model input / output information are updated in the equation (5) by sequential learning. By reflecting the latest model parameter a11 and the like, the actuator (cooling fan 38) required to obtain the target value Pb_target and the like after d seconds while the latest state of the system is reflected. Etc.) can be acquired.

  For example, when there is a request to extract the maximum power of the secondary battery 28, the target battery temperature Tb_target may be set to the maximum allowable temperature of the secondary battery 28. Further, in particular, when there is no request to draw the maximum power from the secondary battery 28, the target battery temperature Tb_target is the cooling capacity (cooling temperature Tc) by the cooling fan 38 and the power loss when realizing the cooling capacity. Calculate from the above.

[Specific Processing in First Embodiment]
FIG. 3 is a flowchart of a routine executed in the system according to the first embodiment. Note that this routine is periodically executed every predetermined control cycle (sampling cycle). In the routine shown in FIG. 3, first, battery current Ib (t), cooling temperature Tc (t), battery power Pb (t), and battery temperature Tb (t) are acquired (step 100).

  Next, the model parameter a11 and the like of the linear model 50 are sequentially identified by the weighted least square method by the method described with reference to the above equation (2) (step 102). Next, the battery current Ib (t) and the cooling temperature Tc required to control the target values Pb_target and Tb_target according to the above equation (5) into which the latest model parameter a11 and the like identified in step 102 are substituted. (T) is calculated (step 104).

  Next, control of the battery current Ib and the cooling fan 38 is executed so that the battery current Ib (t) and the cooling temperature Tc (t) acquired in step 104 are obtained (step 106).

  According to the processing of the routine shown in FIG. 3 described above, the linear model 50 can be adapted to state changes such as changes in the system over time and environmental changes surrounding the system by sequential learning of the model parameter a11 and the like. . Further, according to the processing of the above routine, the battery power Pb after d seconds from the current time t according to the above equation (5) in which the sequentially learned model parameters a11 and the like are substituted (that is, the system state is reflected) It is possible to accurately obtain the battery current Ib (t) and the cooling temperature Tc (t) that are currently required to satisfy the target values Pb_target and Tb_target of the battery temperature Tb at the same time. As a result, the battery current Ib and the battery temperature (cooling air temperature) can be controlled so that the target power can be output within a range not exceeding the upper limit of the allowable temperature of the secondary battery 28.

  As described above, according to the system of the present embodiment, it can be configured relatively easily, and the linear model 50 in which the model parameter a11 and the like are sequentially learned so that the problem of the control system with two inputs and two outputs can be handled well. Become. Further, the adaptive control using such a linear model 50 can realize smooth and accurate control with respect to the target value.

In the first embodiment described above, the ECU 30 executes the process of step 100, so that the “model input / output information acquisition means” in the first invention executes the process of step 102. The “parameter calculation means” in the first aspect of the present invention is realized.
Further, when the ECU 30 executes the process of step 104, the “control amount calculation means” in the second invention causes the “control means” in the second invention to execute the process of step 106. , Each has been realized.

Embodiment 2. FIG.
Next, a second embodiment of the present invention will be described with reference to FIG.
[Linear Model of Embodiment 2]
FIG. 4 is a diagram showing a linear model 60 used in the second embodiment of the present invention. In the linear model 60 shown in FIG. 4, the battery current Ib (t) serving as input information to the model is the battery hysteresis (I-V characteristics) during charging and discharging, compared to the linear model 50 of the first embodiment described above. ) In consideration of charging current Ii (t) and discharging current Io (t). Further, as the output information of the model, a battery charge amount (charge state SOC) Sc (t) is added.

また、本実施形態の線形モデル６０においても、上述した実施の形態１と同様の手法で、モデルパラメータａ１１等が逐次学習によって同定されるようになっている。 That is, the linear model 60 of the present embodiment is constructed as a three-input three-output linear model. In this embodiment, the linear model 60 is represented as the following equation (6) based on the same concept as the above-described linear model 50 of the two-input two-output system.

Also in the linear model 60 of the present embodiment, the model parameter a11 and the like are identified by sequential learning in the same manner as in the first embodiment.

[Construction of control system using linear model of embodiment 2]
In the present embodiment, the target battery charge amount Sc_target, the target battery power Pb_target, and the target battery temperature Tb_target can be satisfied by the following method using the identification result of the model parameter a11 and the like of the linear model 60 described above. Such charging current Ii (t), discharging current Io (t), and cooling temperature Tc (t) are obtained.

Therefore, in the present embodiment, an evaluation function E expressed as the following expression (7) is introduced based on the same idea as in the first embodiment.

本実施形態においても、上記図３に示すルーチンと同様の処理を実行するようにすれば、線形モデル６０を、モデルパラメータａ１１等の逐次学習によって、システムの経時変化などの状態変化やシステムを取り巻く環境変化に適応させることが可能となる。また、逐次学習されたモデルパラメータａ１１等が反映された上記（８）式に従って、現時刻ｔからｄ秒後における電池充電量Ｓｃ、電池電力Ｐｂ、および電池温度Ｔｂの目標値Ｓｃ_target、Ｐｂ_target、Ｔｂ_targetを同時に満足させるために現時点で必要とされる充電電流Ｉｉ（ｔ）、放電電流Ｉｏ（ｔ）、および冷却温度Ｔｃ（ｔ）を正確に求めることが可能となる。 Also in the present embodiment, using the same method as in the first embodiment, the formula after setting the evaluation function E in the formula (7) to zero is substituted into the formula (6) and rearranged. Thus, the following control expression (8) is obtained.

Also in the present embodiment, if the same processing as the routine shown in FIG. 3 is executed, the linear model 60 is subjected to sequential learning of the model parameter a11 and the like, so as to surround state changes such as system aging and the system. It becomes possible to adapt to environmental changes. Further, according to the above equation (8) that reflects the sequentially learned model parameter a11 and the like, the target values Sc_target, Pb_target, Tb_target of the battery charge amount Sc, the battery power Pb, and the battery temperature Tb after d seconds from the current time t It is possible to accurately obtain the charging current Ii (t), the discharge current Io (t), and the cooling temperature Tc (t) that are currently required to satisfy the above.

  More strictly, the battery power Pb (t) is restricted not only by the battery temperature Tb (t) but also by the state of charge SOC of the battery. For this reason, according to the linear model 60 of this embodiment that also includes the battery charge amount Sc (t) as model output information, the battery power Pb (t) can be calculated more accurately. In addition, as described above, the input value to the linear model 60 is divided into two current terms in consideration of the IV characteristics of the secondary battery 28. Therefore, also in this respect, the accuracy is high. Model construction has been realized.

  As described above, according to the system of the present embodiment, the configuration of the control system with three inputs and three outputs can be satisfactorily handled by the linear model 60 in which the model parameter a11 and the like can be sequentially learned. Become. Further, by the adaptive control using such a linear model 60, it is possible to realize smooth and accurate control with respect to the target value as compared with the first embodiment described above. In addition, as described above, improvement in model accuracy can be ensured by a simple method of increasing the number of inputs and outputs of the linear model as compared with the first embodiment described above.

Embodiment 3 FIG.
Next, a third embodiment of the present invention will be described with reference to FIG.
[Linear Model of Embodiment 3]
FIG. 5 shows a linear model 70 used in the third embodiment of the present invention. In this embodiment, the vehicle drive motor 14 shown in FIG. Therefore, in the present embodiment, as shown in FIG. 5, the linear model 70 is obtained using the d-axis current Id and the q-axis current Iq applied to the motor 14 as parameters for determining the operating state of the motor as model input information. The motor torque Tq and the motor speed (rotational speed) θm, which are parameters indicating the operation state of the motor, are constructed as models for calculating (outputting).

本実施形態の線形モデル７０においても、上述した実施の形態１と同様の手法で、モデルパラメータａ１１等が逐次学習によって同定されるようになっている。 That is, the linear model 70 of the present embodiment is a two-input two-output linear model. Therefore, in the present embodiment, such a linear model 70 is represented as the following equation (9) based on the same idea as in the first embodiment.

Also in the linear model 70 of the present embodiment, the model parameter a11 and the like are identified by sequential learning in the same manner as in the first embodiment.

[Construction of control system using linear model of Embodiment 3]
In the present embodiment, the identification result of the model parameter a11 and the like of the linear model 70 described above is used, and the d-axis current Id () that satisfies the target motor torque Tq_target and the target motor speed θm_target by the following method. t) and q-axis current Iq (t) are obtained.

Therefore, in the present embodiment, an evaluation function E expressed as the following equation (10) is introduced based on the same concept as in the first embodiment.

本実施形態においても、上記図３に示すルーチンと同様の処理を実行するようにすれば、線形モデル７０を、モデルパラメータａ１１等の逐次学習によって、システムの経時変化などの状態変化やシステムを取り巻く環境変化に適応させることが可能となる。また、逐次学習されたモデルパラメータａ１１等が反映された上記（１１）式に従って、現時刻ｔからｄ秒後におけるモータトルクＴｑおよびモータ速度θｍの目標値Ｔｑ_targetおよびθｍ_targetを同時に満足させるために現時点で必要とされるｄ軸電流Ｉｄ（ｔ）およびｑ軸電流Ｉｑ（ｔ）を正確に求めることが可能となる。 Also in the present embodiment, using the same method as in the first embodiment, the formula after setting the evaluation function E in the formula (10) to zero is substituted into the formula (9) and rearranged. Thus, the following control expression (11) is obtained.

Also in the present embodiment, if the same processing as the routine shown in FIG. 3 is executed, the linear model 70 is subjected to sequential learning of the model parameter a11 and the like so as to surround state changes such as system aging and the system. It becomes possible to adapt to environmental changes. Further, at the present time, in order to satisfy the target values Tq_target and θm_target of the motor torque Tq and the motor speed θm after d seconds from the current time t at the same time in accordance with the above equation (11) reflecting the sequentially learned model parameter a11 and the like. The required d-axis current Id (t) and q-axis current Iq (t) can be accurately obtained.

  According to the system of the present embodiment described above, whether the control target is the above-described secondary battery 28 or the motor 14 by the linear model 70 in which the model parameter a11 and the like can be sequentially learned can be configured relatively easily. Regardless of the case, the problem of the control system with two inputs and two outputs can be handled satisfactorily. In addition, the adaptive control using such a linear model 70 can realize smooth and accurate control with respect to the target value.

Embodiment 4 FIG.
Next, a fourth embodiment of the present invention will be described with reference to FIG. 6 and FIG.
[System configuration]
FIG. 6 is a diagram for explaining an autonomous guided vehicle system to which a multi-input / multi-output control apparatus according to Embodiment 4 of the present invention is applied. An autonomous transport vehicle (autonomous mobile robot) 80 shown in FIG. 6 is a vehicle that travels with a right wheel 82, a left wheel 84, and caster wheels 86. The autonomous transport vehicle 80 includes a right wheel driving motor 88 that drives the right wheel 82 and a left wheel driving motor 90 that drives the left wheel 84.

  The left and right wheel driving motors 88 and 90 are connected to the ECU 92, respectively. The left wheel torque τl and the right wheel torque τr are controlled by the control of the motors 88, 90 by the ECU 92. The ECU 92 is connected to a vehicle speed sensor 94 that detects the vehicle speed v of the autonomous guided vehicle 80 and an angular speed sensor 96 that detects the vehicle angular speed ω.

  According to the autonomous guided vehicle 80 configured as described above, the autonomous vehicle 80 can travel at an arbitrary speed v by controlling the left wheel torque τl and the right wheel torque τr by the ECU 92. . Further, by controlling the motor 88.90 by the ECU 92 so that the rotational speeds of the right wheel 82 and the left wheel 84 are different, the angular velocity ω in the yaw axis direction of the autonomous transport vehicle 80, that is, the autonomous transport vehicle 80. You will be able to control the course.

[Linear Model of Embodiment 4]
FIG. 7 is a diagram showing the linear model 100 used in the fourth embodiment of the present invention. In the present embodiment, the autonomous guided vehicle 80 shown in FIG. Therefore, in the present embodiment, as shown in FIG. 7, the linear model 100 is obtained by inputting the left wheel torque τl and the right wheel torque τr, which are parameters for determining the driving state and the running path (route) of the autonomous guided vehicle 80, as model input information. As a model for calculating (outputting) the vehicle speed v and the vehicle angular speed ω, which are parameters indicating the driving state and the running path of the autonomous guided vehicle 80.

本実施形態の線形モデル１００においても、上述した実施の形態１と同様の手法で、モデルパラメータａ１１等が逐次学習によって同定されるようになっている。 That is, the linear model 100 of this embodiment is a two-input two-output linear model. Therefore, in the present embodiment, such a linear model 100 is represented as the following expression (12) based on the same idea as in the first embodiment.

Also in the linear model 100 of the present embodiment, the model parameter a11 and the like are identified by sequential learning in the same manner as in the first embodiment.

[Construction of control system using linear model of embodiment 4]
In the present embodiment, the left wheel torque τl (t that satisfies the target vehicle speed v_target and the target vehicle angular speed ω_target by the following method using the identification result of the model parameter a11 and the like of the linear model 100 described above. ) And right wheel torque τr (t).

Therefore, in the present embodiment, an evaluation function E expressed as the following expression (13) is introduced based on the same concept as in the first embodiment.

本実施形態においても、上記図３に示すルーチンと同様の処理を実行するようにすれば、線形モデル１００を、モデルパラメータａ１１等の逐次学習によって、システムの経時変化などの状態変化やシステムを取り巻く環境変化に適応させることが可能となる。また、逐次学習されたモデルパラメータａ１１等が反映された上記（１４）式に従って、現時刻ｔからｄ秒後における車両速度ｖおよび車両角速度ωの目標値ｖ_targetおよびω_targetを同時に満足させるために現時点で必要とされる左輪トルクτｌ（ｔ）および右輪トルクτｒ（ｔ）を正確に求めることが可能となる。 Also in the present embodiment, using the same method as in the first embodiment, the formula after setting the evaluation function E in the above equation (13) to zero is substituted into the above equation (12) and rearranged. Thus, the following control expression (14) is obtained.

Also in the present embodiment, if the same processing as the routine shown in FIG. 3 is executed, the linear model 100 is subjected to sequential learning of the model parameter a11 and the like so as to surround state changes such as system aging and the system. It becomes possible to adapt to environmental changes. Further, in order to satisfy simultaneously the target values v_target and ω_target of the vehicle speed v and the vehicle angular speed ω after d seconds from the current time t according to the above equation (14) reflecting the sequentially learned model parameter a11 and the like. The required left wheel torque τl (t) and right wheel torque τr (t) can be accurately obtained.

  According to the system of the present embodiment described above, it can be configured relatively simply, and the controlled object is the above-described secondary battery 28 or the like by the linear model 70 in which the model parameters a11 and the like are sequentially learned. Regardless of whether or not, the problem of the control system with two inputs and two outputs can be handled well. In addition, the adaptive control using the linear model 100 can realize smooth and accurate control with respect to the target value.

  By the way, in the first to fourth embodiments described above, the previous values (t−1) of the model input information (Ib and the like) and the model output information (Pb and the like) are converted into a linear form or a control expression. Although included, the past value of the model input / output information included in the line format and the control expression derived therefrom in the present invention is not limited to the previous value (t−1). In other words, the line format and the control expression of the present invention may include a previous value (t-2) or the like as necessary.

It is a figure for demonstrating the drive system of the hybrid vehicle to which the control apparatus of the multiple input multiple output system in Embodiment 1 of this invention was applied. It is a figure showing the linear model used in Embodiment 1 of this invention. It is a flowchart of the routine performed in the system of this Embodiment 1. FIG. It is a figure showing the linear model used in Embodiment 2 of this invention. It is a figure showing the linear model used in Embodiment 3 of this invention. It is a figure for demonstrating the system of the autonomous guided vehicle to which the control apparatus of the multiple input multiple output system in Embodiment 4 of this invention was applied. It is a figure showing the linear model used in Embodiment 4 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Drive system 12 Internal combustion engine 14 Vehicle drive motor 16 Generator 26 Inverter 28 Secondary battery 30, 92 ECU (Electronic Control Unit)
32 Voltage sensor 34 Battery temperature sensor 36 Current sensor 38 Cooling fan 40 Temperature sensor 50, 60, 70, 100 Linear model 80 Autonomous transport vehicle 82 Right wheel 84 Left wheel 86 Caster wheel 88 Right wheel drive motor 90 Left wheel drive motor a11, b11, etc. Model parameter d Delay time E Evaluation function Ib Battery current Id d-axis current Ii charge current Io discharge current Iq q-axis current Pb battery power Sc battery charge amount Tb battery temperature Tc cooling temperature Tq motor torque v vehicle speed γ constant θm Motor speed τl Left wheel torque τr Right wheel torque ω Vehicle angular speed

Claims (6)

A multi-input multi-output control device having a linear model that handles multiple inputs and outputs,
The linear model includes a current value of a plurality of model input information whose coefficients are model parameters to be learned, a past value of a plurality of model input information whose coefficients are other model parameters to be learned, and a learning target. The sum of the current value of a plurality of model output information using the other model parameter as a coefficient and the past value of the plurality of model output information using the other model parameter as a coefficient as a coefficient, Includes line formats representing multiple model output information,
The control device of the multi-input multi-output system is
Model input / output information acquisition means for acquiring the plurality of model input information and the plurality of model output information;
Parameter calculation means for calculating the model parameter based on the linear format for a predetermined sampling period into which the plurality of model input information and the plurality of model output information are substituted,
A multi-input multi-output system control device. The control device of the multi-input multi-output system is
The target model output value is obtained based on a control expression derived from an equation representing an evaluation function that determines the responsiveness of the system until the target model output value is reached after a predetermined time has elapsed and the linear form. Control amount calculation means for calculating the model input information for
Control means for controlling the model input information so as to be the model input information calculated by the control amount calculation means;
The multi-input multi-output control apparatus according to claim 1, further comprising: The control device of the multi-input multi-output system is a control device that controls the state of the secondary battery,
The model input information is a parameter for determining the state of the secondary battery,
3. The control apparatus for a multi-input multi-output system according to claim 1, wherein the model output information is a parameter indicating a state of the secondary battery. The parameters for determining the state of the secondary battery are the charging current of the secondary battery, the discharging current of the secondary battery, and the cooling temperature of the secondary battery,
4. The multi-input multi-output system according to claim 3, wherein the parameters indicating the state of the secondary battery are the power of the secondary battery, the temperature of the secondary battery, and the charge amount of the secondary battery. Control device. The control device of the multi-input multi-output system is a control device that controls an electric motor,
The model input information is a parameter for determining the operating state of the electric motor,
The multi-input / multi-output control apparatus according to claim 1, wherein the model output information is a parameter indicating an operating state of the electric motor. The multi-input multi-output control device is a control device for controlling an autonomous mobile robot,
The model input information is a parameter for determining the driving state and course of the autonomous mobile robot,
The multi-input multi-output control apparatus according to claim 1, wherein the model output information is a parameter indicating a driving state or a course of the autonomous mobile robot.

Images