JP4795920B2 - Control map optimization device - Google Patents

Description

  The present invention relates to a control map optimization device, and more particularly to a control map optimization device that optimizes a control map related to control of a control target.

  In many cases, the control parameters used for control of the controlled object are individually determined, but may be calculated by a control map having one or more arguments as input. Then, by optimizing the control map, it is possible to automatically determine the optimum value of the control parameter based on it.

  Conventionally, the optimization of the control map is generally performed by a developer or a worker through manual tuning based on experience and intuition through experiments and simulations. However, manual tuning does not necessarily guarantee that an optimal control map can be obtained, and there is also a problem that man-hours for tuning are enormous and it takes time and effort.

  As an apparatus or method for solving such a problem, for example, in the engine control of a vehicle, the ignition timing at which the maximum torque of the engine can be extracted is in the vicinity of the ignition timing at which knocking occurs. The change state of the heat release rate is calculated from the combustion physical quantity that changes with the combustion of the fuel, the state just before knocking is detected based on the change state of the heat release rate, and the operating condition at that time is used as a parameter of the combustion control map A method for creating a combustion control map to be used is disclosed (see Patent Document 1).

In addition, when the engine is held at a constant rotational speed, even if the throttle opening is the same, the torque output differs depending on whether the throttle valve is operated in the opening direction or the closing direction. There has been proposed a method in which torque curves in the opening direction and the closing direction are respectively determined by the number of rotations, and a control map is created based on each torque curve (see Patent Document 2).
Japanese Patent Laid-Open No. 2-22162 JP 2000-321175 A

  However, these control map creation devices and creation methods are devices and methods that depend on the characteristics of the controlled object such as the combustion state of the engine and the throttle valve, respectively, and cannot be applied to other controlled objects. It is not necessarily a general-purpose method.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a general-purpose control map optimization device capable of optimizing a control map without depending on the characteristics of a control target. .

In order to solve the above problem, the first invention provides:
The control map optimization device
For the input argument, the control parameter set at the corresponding grid point, or the control parameter set at the grid point near the map position corresponding to the argument when there is no grid point corresponding to the argument A control map for outputting control parameters calculated by interpolation from
An optimization means for optimizing the value of the control parameter set at the grid point of the control map based on the information about the value to be taken by the control parameter and the control parameter output from the control map;
The information is information on an ideal state of a control parameter input to the optimization unit or calculated by the optimization unit,
The optimization unit creates an error function that is a unimodal function based on the information and the output control parameter, and corresponds to the argument in the control map based on the error function by a steepest descent method. The control parameter values set at the grid points in the vicinity of the map position are updated, and this updating process is repeated.

  According to a second aspect of the present invention, in the control map optimization device according to the first aspect, the error function that is the unimodal function is a numerical value determined based on the information and a control parameter output from the control map. It is a quadratic function.

According to a third aspect of the present invention, in the control map optimizing device according to the first or second aspect of the invention, the optimization means determines a difference between a numerical value determined based on the information and a control parameter output from the control map. Accordingly, the control parameter values set at the grid points of the control map are updated.

According to a fourth aspect of the present invention, in the control map optimizing device according to the third aspect, the optimization means includes a numerical value with which the update amount of the control parameter set at the lattice point of the control map is determined based on the information. The value of the control parameter set at the lattice point of the control map is updated so as to be an amount proportional to the difference from the control parameter output from the control map.

According to a fifth invention, in the control map optimizing device according to any one of the first to fourth inventions, the information is determined based on a result of a control trial performed in accordance with a control parameter output from the control map. It is characterized by being a numerical value.

  A sixth invention is the control map optimization device according to any one of the first to fourth inventions, wherein the information is a control parameter set at a lattice point in the vicinity of a map position corresponding to the argument in the control map. Is an index that should be increased or decreased, and the optimization means determines the predetermined numerical value based on the index and creates the error function.

According to a seventh aspect of the present invention, in the control map optimizing device according to any one of the first to sixth aspects of the present invention, when the number of arguments input to the control map is n, the optimization means The control parameters set in 2 ⁿ lattice points around the argument, which are lattice points in the vicinity of the map position corresponding to the argument, are updated.

According to an eighth aspect of the present invention, in the control map optimization device according to any one of the first to seventh aspects, the optimization unit performs a control instead of performing the update process every time the value of the control parameter is updated. During the operation, the control parameter value is not updated, the update amount of the control parameter value is stored, and the control parameter value is updated collectively after the control operation is completed.

According to a ninth invention, in the control map optimizing device according to any one of the first to seventh inventions, the optimization unit updates the value of the control parameter at the initial stage of the iteration. In the later stage of the repetition, the control parameter value is not updated during the control operation, and the update amount of the control parameter value is stored. It is characterized by updating.

According to the first invention, the optimization means creates an error function that is a unimodal function based on the information about the ideal state of the control parameter and the control parameter actually output from the control map, and the steepest descent method Thus, based on the error function, the control parameter values set at the lattice points near the map position corresponding to the argument in the control map are sequentially updated, and the control map is optimized by repeating this updating process.

  In this way, by making the error function a unimodal function and applying the steepest descent method to it, it is possible to accurately calculate the optimal solution without falling into the local solution, and to optimize the control map accurately. With this method, it is possible to obtain a control map that accurately controls the control content of any control target, and the control map optimization device can be used as a control target. Therefore, it is possible to provide a general-purpose device capable of optimizing the control map without depending on the characteristics.

  According to the second invention, in addition to the effect of the first invention, the unimodal function used as the error function is a quadratic function, so that the error function can be configured very easily. The error function can be easily calculated, and the processing speed can be improved.

According to the third invention, in addition to the effects of the respective inventions, the update of the value of the control parameter set at the lattice point of the control map is controlled based on the numerical value and control determined based on the information on the ideal state of the control parameter. By performing according to the difference from the control parameter output from the map, it becomes easy to calculate the update amount of the control parameter, and the optimization means reliably recognizes the direction of increase / decrease of the control parameter based on the difference Therefore, the control map can be optimized more accurately.

According to the fourth invention, in the third invention, the update of the value of the control parameter is proportional to the difference between the numerical value determined based on the information regarding the ideal state of the control parameter and the control parameter output from the control map. By doing so, the effect of the third aspect of the invention is more accurately exhibited.

According to the fifth invention, in addition to the effects of the respective inventions, the information related to the ideal state of the control parameter is a numerical value determined based on the result of the control trial performed according to the control parameter output from the control map; By doing so, the result of the trial is reliably fed back to the optimization of the control parameter, and the optimization of the control map can be performed more accurately.

According to the sixth invention, in addition to the effects of the respective inventions, the information regarding the ideal state of the control parameter is used as an index that the control parameter of the control map should be increased or decreased, so that the optimization means can increase or decrease the control parameter. Since the control parameter is surely increased / decreased by simply instructing correctly, it is not necessary to precisely determine the numerical value determined based on the information , and the control map can be optimized more easily. The configuration of the control map optimization device can be further simplified.

According to the seventh invention, in addition to the effects of the above inventions, in order to update the control parameters set in 2 ⁿ grid points around the map position corresponding to the n-dimensional argument input to the control map The parameters set for the grid points around the map position can be updated reliably, and only the control parameters set for the grid points around the map position are updated in one update process, reducing the time required for the update process. As a result, the processing time of the entire apparatus can be shortened.

According to the eighth aspect of the present invention, the update process in the control map optimizing means can be configured to be performed sequentially every time the control parameter value is updated, and the control parameter value is updated during the control operation. It is also possible to store the update amount of the control parameter value without updating it and update the control parameter value collectively after the completion of the control operation.

  In addition to the effects of the above inventions, it is possible to further improve the convergence to the optimal solution in the optimization when the control parameter values are updated sequentially, and when updating the control parameter values all at once. Even in the case where a so-called noise is included in an argument or the like, it is difficult to be influenced by the noise, and it is possible to improve the noise resistance of the optimization in the control map optimizing device.

  According to the ninth aspect of the invention, the control parameter value is sequentially updated in the initial stage of the update process repetition, and is configured to be updated collectively after the end of the series of control operations in the later stage of the repetition. In addition to the effects of the inventions described above, it is possible to improve the anti-noise property of optimization by batch updating at the later stage of iteration while improving convergence to the optimal solution in optimization by successive updating at the initial stage of iteration.

  Embodiments of a control map optimization apparatus according to the present invention will be described below with reference to the drawings.

[First Embodiment]
FIG. 1 is a block diagram illustrating the configuration of a control map optimization apparatus according to the first embodiment of the present invention. The control map optimizing device 1 is composed of a general-purpose computer in which a CPU, ROM, RAM, input / output interface and the like (not shown) are connected to a bus. The control map optimization apparatus 1 includes a memory 2 in which a control map M to be optimized is stored, and an optimization unit 3 that optimizes the control map M.

  The control map is normally stored in a memory or the like of a control computer (not shown) on the controlled object S side. In the control map optimizing device 1 of FIG. 1, the same control map M as the control map of the control target S is stored in the memory 2, and after the control map M is optimized, it is stored in the memory or the like of the control computer of the control target S. However, the control map itself stored in the memory of the control computer of the control target S may be the optimization target.

One or more arguments are input to the control map, and one or more control parameters are output. In the present embodiment, the case control map M is two arguments x, so-called three-dimensional control map for outputting an input to one of the control parameters P _E and y. The number of arguments, the number of output control parameters, and the number of dimensions of the control map are not limited to this.

In the present embodiment, since there are two arguments as described above, for example, when the argument x is taken on the vertical axis and the argument y is taken on the horizontal axis, the control map M is expressed in a plane as shown in FIG. . Further, one control parameter is set at each position where the arguments x and y on the control map M take discrete specific values x _i and y _j , respectively.

Hereinafter, these positions on the control map M are referred to as lattice points, and are represented as lattice points (x _i , y _j ). 2 and FIG. 3 to be described later, the specific values x _i and y _{j of} the arguments x and y are set at equal intervals, and the lattice points (x _i , y _j ) are arranged in a grid pattern. However, the setting of the specific value of the argument, that is, the setting of the grid point is not limited to this.

If the control map M is expressed including the control parameters P _{i, j} set at the respective lattice points (x _i , y _j ), it can be expressed three-dimensionally as shown in FIG.

When an argument is input to the control map M, if a lattice point (x _E , y _E ) corresponding to the input argument x, y exists on the map, the corresponding lattice point (x _E , y _E ) The control parameter P _E set in (1) is output.

Further, when there are no grid points corresponding to the input arguments x _E and y _E on the map, the control map M has map positions (x _E , four lattice points surrounding the _{y E) (x a, y} a), (x B, y B), (x C, y C), (x D, each control set to y _D) parameters P _a , P _B , P _C , and P _D , the control parameter _PE is calculated and output by linear interpolation.

Since in this embodiment a number of arguments 2, when the grid points corresponding to the input argument does not exist on the map are four surrounding map position corresponding to the argument (= 2 ²⁾ The control parameter _PE is calculated by linear interpolation from the control parameters set at the grid points. If the number of arguments is n, the control parameters P _E are set to 2 ⁿ grid points around the map position corresponding to the arguments. A control parameter is calculated by linear interpolation from each control parameter.

As shown in FIG. 1, the optimization unit 3 is _supplied with information Q relating to a value to be taken by the control parameter together with the control parameter PE output from the control map M.

When the output control parameter _PE and information Q are input, the optimization unit 3 creates an error function that is a unimodal function based on the control parameter _PE and the information Q, and performs the control based on the error function by a steepest descent method. The value of the control parameter set at the lattice point near the map position (x _E , y _E ) corresponding to the arguments x and y in the map M is updated. Then, by repeating this update process, the control parameters of each grid point of the control map M are optimized.

The information Q regarding the value to be taken by the control parameter can take various forms, but in the present embodiment, the control parameter P _E output from the control map M when an argument is input to the control map M. An ideal value to be taken is input as information Q.

  Further, in the present embodiment, the optimization unit 3 performs the update process of the control parameter of the control map M every time the arguments x and y are input, and the series of control operations is completed. However, if the control map M is not optimized, the control map M is optimized by repeating the control operation and the update of the control parameters in the meantime. The actual control of the controlled object S is not necessarily performed, and the control map M can be optimized only by simulation inside the control map optimizing apparatus 1.

  Hereinafter, based on the flowchart of the update process shown in FIG. 5, the update process of the control parameter in the optimization means 3 is demonstrated concretely.

First, when arguments x and y are input to the control map optimization device 1, the control map M in the memory 2 is referred to, and the control parameter _PE is calculated and output from the control map M according to the above calculation method. (Step S1). In the present embodiment, the output control parameter P _E is input to the control object S, and control is performed according to the control parameter P _E.

Moreover, the control parameter P _E output from the control map M is transmitted to the optimization unit 3 at the same time. In the optimization means 3, the control parameter P _E transmitted from the control map M and the information Q regarding the value to be taken by the control parameter, that is, the ideal of the control parameter P _E to be output from the control map M in this embodiment. Based on the value Q, the update amount of the control parameter set at each grid point of the control map M is calculated (step S2).

Optimization means 3, the 2 ⁿ pieces of grid points surrounding a grid point near the map location corresponding to the argument in the control map M when the number of arguments to be input to the control map M is n arguments The set control parameter is updated. In this embodiment, since the number n of arguments is 2, as shown in FIG. 4, the map position corresponding to the input arguments x and y is shown. Each of the four grid points (x _A , y _A ), (x _B , y _B ), (x _C , y _C ), (x _D , y _D ) around (x _E , y _E ) The control parameters P _A , P _B , P _C and P _D are updated.

Optimization means 3, in the update amount calculation of the control parameters, first, a single-peak function on the basis of the output control parameters P _E from the ideal value Q and the control map M of the control parameter P _E as information Q Create an error function Error. In this embodiment, a quadratic function is used as a unimodal function, and an error function Error is created as shown in the following equation (1).
_{Error = (Q-P E)} 2/2 ... (1)

  Here, the reason why the error function is a unimodal function is that when the steepest descent method is applied to the error function as a multimodal function having a number of local maxima and minima, it converges to a local solution that is not necessarily the optimal solution. Although the possibility remains, such a problem does not occur if the error function is a unimodal function.

Subsequently, the optimization means 3 applies the steepest descent method to the error function expressed by the above equation (1), and, as shown in FIG. 6, each grid point (x _A , y _A ), (x _B , Y _B ), (x _C , y _C ), (x _D , y _D ), the update amounts ΔP _A , ΔP _B , ΔP _{C of} the control parameters P _A , P _B , P _C , P _D set in (X _D , y _D ), ΔP _D is calculated according to the following equations (2) to (5).

Wherein (2) - (5) as indicated in formula, in the present embodiment, the optimization means 3, output from the ideal value Q and the control map M of numerical and which control parameter P _E representing the information Q The grid points (x _A , y _A ), (x _B , y _B ), (x _C , y _C ), (x _D , y _D ) of the control map M are set according to the difference from the control parameter P _E. The control parameter values P _A , P _B , P _C , and P _D are updated. In particular, the update amounts ΔP _A , ΔP _B , ΔP _C , and ΔP _D are proportional to the difference between the ideal value Q and the control parameter P _E. Updated to be a quantity.

Subsequently, the optimization unit 3 uses the calculated update amounts ΔP _A , ΔP _B , ΔP _C , ΔP _D to use the lattice points (x _A , y _A ), ( x _B , y _B ), (x _C , y _C ), and control parameter values P _A , P _B , P _C , and P _D set in (x _D , y _D ) are set to P _A + ΔP _A , P _B + ΔP _B , P _C + ΔP _C , P _D + ΔP _D are updated, and these values are newly overwritten and set as control parameter values P _A , P _B , P _C , P _D. .

  Subsequently, in the present embodiment, the optimization unit 3 determines whether or not a series of control operations has been completed based on the information from the control target S (step S4 in FIG. 5), and the control operations must be completed. If so (step S4: NO), the control parameter updating process of the control map M is repeated during the control operation (steps S1 to S3).

  Further, when the optimization unit 3 determines that a series of control operations has been completed (step S4: YES), the optimization unit 3 determines whether or not an end condition is satisfied (step S5).

The termination condition is appropriately determined from the viewpoint of the accuracy of each control parameter required for the control map M, the time required for optimization, and the like. For example,
1) The absolute value of the update amount ΔP _{i, j} of the control parameter P _{i, j} set at each grid point (x _i , y _j ) of the control map M does not exceed a predetermined threshold value set in advance. 2) The error between the control result of the control object S based on the control map M and the ideal control result is within a predetermined range. 3) The number of updates of the control parameter P _{i, j} of the control map M is set in advance. 4) An end condition can be set such that the time required for updating the control parameter P _{i, j} of the control map M has reached a preset time.

  When the optimization unit 3 determines that the end condition is not satisfied (step S5: NO), the optimization unit 3 requests a new series of control operations, and performs a process of updating the control parameter of the control map M until the end condition is satisfied. Repeat (Steps S1 to S4).

When the optimization unit 3 determines that the termination condition is satisfied (step S5: YES), the optimization unit 3 outputs a control map M in which the value of the control parameter P _{i, j} is updated and optimized, and controls the control target S. The control map stored in the memory or the like of the computer is overwritten and transferred.

  The operation of the control map optimizing apparatus 1 according to the present embodiment is as described in the above configuration, and a description thereof is omitted.

As described above, according to the control map optimization apparatus 1 according to the present embodiment, optimization means 3, the information Q (the embodiment for the value to be taken by the control parameter control parameter P _E is the ideal value should take and Q), to create an error function error unimodal function based on the control parameter P _E that is actually output from the control map M, the argument x in the control map M based on the error function error by the steepest descent method , map location corresponding to y (x _E, y _E) grid point near the _{(x a, y a),} (x B, y B), (x C, y C), (x D, y D) The control map values P _A , P _B , P _C , and P _{D set in the above} are sequentially updated, and the control map M is optimized by repeating this update process.

  In this way, by making the error function a unimodal function and applying the steepest descent method to it, it becomes possible to accurately calculate an optimal solution without falling into a local solution, and to optimize the control map M With this method, it is possible to obtain the control map M that accurately controls the control content of any control object S, and the control map optimization device 1 Can be a general-purpose device capable of optimizing the control map M without depending on the characteristics of the controlled object S.

In the present embodiment, the information Q about the values should take control parameter for the case of the ideal value Q to be taken by the control parameter P _E argument is output from the control map M in input to the control map M Although described, it is not limited to this. The information Q may be, for example, an indicator that the control parameter set at the grid point near the map position corresponding to the argument in the control map M should be increased or decreased.

Specifically, when the arguments x and y are input to the control map M and the control parameter P _E is output, the control unit P _E should be increased or decreased as the information Q with respect to the optimization unit 3. indicators "+" or when it is determined "-" configured to is input, the optimization means 3, adding or subtracting to a predetermined value, for example, based on the index to the control parameter P _E wherein ( It can be configured to have the ideal value Q in the equations (1) to (5). Index "+", "-" may be configured such that predetermined multiple of the control parameter P _E according to.

Even with this configuration, it is possible to appropriately optimize the control parameter P _{i, j} of the control map M, and the same effect as the control map optimization apparatus 1 according to the present embodiment can be obtained. .

Further, in the present embodiment, the case where the optimization unit 3 sequentially performs the update process of the control parameter P _{i, j} of the control map M every time the arguments x and y are input has been described. For example, during a series of control operations, the control parameter value update amount ΔP _{i, j} is accumulated without updating the control parameter value P _{i, j} , and the control parameter value P _i is collectively obtained after the control operation is completed. _{, j} can be configured to be updated.

In this case, the flow of the update process is performed according to the flowchart shown in FIG. Specifically, as in the above embodiment, the control parameter P _E is output from the control map M in response to the input of the arguments x and y (step S6), and each grid point (x _A , y _{A of the} control map M is output. _{), (x B, y B} ), (x C, y C), (x D, the control parameter is set to _{y D) P a, P B} , P C, P D of the update amount ΔP _a, ΔP _B , ΔP _C , ΔP _D are calculated according to the equations (1) to (5) (step S7). At this time, the control parameter is not updated until a series of control operations is completed (step S8: NO), and the update amount is repeatedly calculated and stored and stored in the memory 2 each time (step S7).

When the optimization means 3 determines that a series of control operations has been completed based on the information from the control object S (step S8: YES), the optimization means 3 stored in the memory 2 for each control parameter P _{i, j} of the control map M The total update amount ΔP _{i, j} of the control parameter P _{i, j} is added and divided by the number of update processes to average the update amount ΔP _{i, j} (step S9), and the original control parameter value P _{i, j The} control parameter P _{i, j} is updated at once by adding the average value of the update amounts ΔP _{i, j to j} (step S10). Then, the optimization unit 3 requests a new series of control operations until the end condition is satisfied (step S11: NO), and repeats the update process of the control parameter of the control map M until the end condition is satisfied (step S11). S6-S10).

If the optimization unit 3 determines that the termination condition is satisfied (step S11: YES), the optimization unit 3 outputs a control map M in which the value of the control parameter P _{i, j} is updated and optimized, and controls the control target S. The control map stored in the computer memory or the like is overwritten and migrated.

Even with this configuration, it is possible to appropriately optimize the control parameter P _{i, j} of the control map M, and the same effect as the control map optimization apparatus 1 according to the present embodiment can be obtained. . In addition, by averaging the update amount, even if so-called noise is included in the argument etc., it becomes difficult to be affected by the noise, and it is possible to improve the noise resistance of the optimization in the control map optimization device Become.

  Furthermore, it is possible to sequentially update the control parameter values in the initial stage of the update process, and to update the values in a batch after the end of a series of control operations in the latter stage of the repetition. In this case, it is possible to improve the anti-noise property of the optimization by batch updating at the later stage of the iteration while improving the convergence to the optimal solution in the optimization by the successive updating at the initial stage of the iteration.

[Second Embodiment]
In the second embodiment according to the present invention, using the control map optimization device of the first embodiment, the control object S is a vehicle S such as an automobile, and the target acceleration and the current speed are used as arguments. A control map optimizing device that optimizes a control map that outputs a control parameter related to an accelerator operation amount will be described. The trial is performed by placing an actual vehicle on a drum type tester, for example.

  Of the members and means of the control map optimizing apparatus according to the present embodiment, the members and means having the same functions as those of the control map optimizing apparatus 1 according to the first embodiment are the same. This will be described with reference numerals.

  As shown in FIG. 8, the control map optimizing apparatus 10 according to the present embodiment, as in the first embodiment, optimizes the memory 2 in which the control map M is stored and the control map M. In addition to the means 3, a target acceleration calculating means 11, an estimated speed calculating means 12, and an ideal output calculating means 13 are provided. In addition, the vehicle S that is the control target S includes a vehicle speed sensor Sv.

  The target speed Vg is input to the control map optimization device 10. The target speed Vg is artificially created for optimizing the control map M. After the control map M optimized by the control map optimizing device 10 is transferred to and mounted on the vehicle S that is the control target S, the target speed Vg is determined by the accelerator pedal opening information and the brake pedal ON / OFF. An ECU (Electric Control Unit) of the vehicle S is created based on the OFF information and the like, converted into a target acceleration ag through a target acceleration calculation means 11 formed in the ECU, and input to the control map M. Is.

  In addition, the vehicle speed V of the vehicle S is input to the control map optimizing device 10 from a vehicle speed sensor Sv mounted on the vehicle S.

  A target speed Vg and a vehicle speed V are input to the target acceleration calculation means 11, and the target acceleration calculation means 11 calculates a target acceleration ag that the vehicle S should take from the target speed Vg and the vehicle speed V. And output it.

The control map M takes the target acceleration ag and the vehicle speed V transmitted from the vehicle speed sensor Sv as arguments x and y. When the arguments ag and V are input, an accelerator (not shown) of the vehicle S is displayed. and it outputs a brake accelerator control parameter P _E for performing opening control and oN / OFF control of the brake pedal of the pedal automatically.

The brake / accelerator control parameter _PE is input to the vehicle S that is the control target S and the optimization means 3.

  In addition, the vehicle speed V of the vehicle S transmitted from the vehicle speed sensor Sv is input to the estimated speed calculation means 12. The estimated speed calculation means 12 stores the vehicle speed V data transmitted from the vehicle speed sensor Sv in a time series from the present to a certain period before, and from the transition of these data to the next processing timing and the like from the present. The vehicle speed V of the vehicle S after a certain period is estimated, an estimated speed Vest, which is an estimated value, is calculated and transmitted to the ideal output calculating means 13.

  The ideal output calculation means 13 receives the estimated speed Vest and the target speed Vg described above. The ideal output calculation means 13 determines the first output based on the difference between the target speed Vg and the estimated speed Vest. Information Q regarding the value to be taken by the control parameter described in the embodiment is determined.

  In the present embodiment, the ideal output calculation means 13 outputs a numerical value obtained by multiplying the difference between the target speed Vg and the estimated speed Vest by a constant as information Qa. That is, if the target speed Vg is greater than the estimated speed Vest, the difference is a positive value and the information Qa is also a positive value. If the target speed Vg is smaller than the estimated speed Vest, the difference is a negative value and the information Qa is also a negative value.

Then, the optimization unit 3, so that a value obtained by adding the this information Qa and brake acceleration control parameter P _E described above is the information Q relating to the value to control parameter takes described in the first embodiment ing. Although the configuration and operation of the optimization means 3 are the same as those in the first embodiment and the description thereof is omitted, the information Q is thus Q = P _E + Qa (6)
Therefore, Qa is directly substituted for (Q−P _E ) on the right side of the expressions (2) to (5), and the calculation is performed.

  As described above, the control map optimizing apparatus 10 according to the present embodiment can achieve the same effects as those of the control map optimizing apparatus 1 according to the first embodiment. That is, by making the error function a unimodal function and applying the steepest descent method thereto, it is possible to accurately calculate an optimal solution without falling into a local solution, and to optimize the control map M accurately. With this method, it is possible to acquire a control map M that accurately controls the control content of any control target S, and the control map optimization apparatus 10 A general-purpose device capable of optimizing the control map M without depending on the characteristics of the controlled object S can be obtained.

  Here, as a result of optimizing the control map M by using the control map optimizing apparatus 10 according to the present embodiment and inputting a target speed Vg that varies with time as shown in FIG. 9, for example, FIG. As indicated by the solid line in the graph, the vehicle speed V of the vehicle S can be controlled so as to substantially match the target speed Vg.

In this case, the optimization of the control map M was performed by the above-described batch method (see FIG. 7). That is, attempts collectively every time the end to update the control parameter P _{i, j} using the stored control parameters P _{i, j} of the update amount [Delta] P _{i, j} during the 20 seconds of trial shown in FIG. 9 For example, optimization was performed by repeating trials a predetermined number of times.

  The result shown by the broken line in FIG. 10 is the result when the control map M is manually adjusted over the same time as the optimization by the control map optimizing device 10, that is, a predetermined number of times times 20 seconds. When the vehicle speed V is adjusted manually in this way, the achieved vehicle speed V is close to the target speed Vg, but the timing of acceleration and deceleration is deviated forward in time as a whole. On the other hand, it can be seen that the optimization by the control map optimizing device 10 according to the present embodiment provides better results at the same time.

  FIG. 11 is a time-series plot of the sum of the error function Error calculated according to the equation (1) during the 20-second trial in the optimization by the control map optimizing device 10 for each trial. Is. As described above, according to the control map optimizing apparatus 10 according to the present embodiment, it can be seen that the sum of the error functions Error rapidly decreases and converges rapidly toward the optimal solution as the number of trials is repeated.

  In the present embodiment, the case has been described in which the information Q related to the value to be taken by the control parameter is determined as a numerical value based on the vehicle speed V as the result of the trial. Not. For example, an index that the control parameter set at the lattice point in the vicinity of the map position corresponding to the argument in the control map M should be increased or decreased is determined based on the result of the trial, and this index is determined based on the information Q. It is good.

It is a block diagram which shows the structure of the control map optimization apparatus which concerns on 1st Embodiment. It is a figure which shows an example of the control map expressed two-dimensionally. It is a figure which shows the control map of FIG. 2 expressed three-dimensionally. It is a figure explaining the linear interpolation by the control parameter set to the grid point around the map position corresponding to an argument. It is a flowchart which shows the procedure of the update process of the control parameter in an optimization means. It is a figure explaining the update amount of the control parameter set to the lattice point. It is a flowchart which shows the procedure of the update process in the case of updating a control parameter collectively in the optimization means. It is a block diagram which shows the structure of the control map optimization apparatus which concerns on 2nd Embodiment. It is a graph which shows an example of the target speed input into the control map optimization apparatus of FIG. It is a graph which shows the result of control by the control map optimized based on the target speed of FIG. It is the graph which plotted the sum of the error function in the case of optimization of Drawing 10 for every trial.

Explanation of symbols

1, 10 Control map optimization device 3 Optimization means
Error Error function M Control map n Number of arguments P _E Output control parameter P _{i, j} Control parameter ΔP _{i, j} Update amount Q Information on the value that the control parameter should take (ideal value)
x, y Argument (x _i , y _j ) Grid point (x _E , y _E ) Map position

Claims (9)

For the input argument, the control parameter set at the corresponding grid point, or the control parameter set at the grid point near the map position corresponding to the argument when there is no grid point corresponding to the argument A control map for outputting control parameters calculated by interpolation from
An optimization means for optimizing the value of the control parameter set at the grid point of the control map based on the information about the value to be taken by the control parameter and the control parameter output from the control map;
The information is information on an ideal state of a control parameter input to the optimization unit or calculated by the optimization unit,
The optimization unit creates an error function that is a unimodal function based on the information and the output control parameter, and corresponds to the argument in the control map based on the error function by a steepest descent method. An apparatus for optimizing a control map, comprising updating control parameter values set at lattice points in the vicinity of a map position and repeating the updating process.   The control map according to claim 1, wherein the error function that is a unimodal function is a quadratic function of a numerical value determined based on the information and a control parameter output from the control map. Optimization device. The optimization means updates the value of the control parameter set at the grid point of the control map according to the difference between the numerical value determined based on the information and the control parameter output from the control map. The control map optimizing device according to claim 1 or 2, characterized in that In the optimization means, the update amount of the control parameter set at the grid point of the control map is an amount proportional to the difference between the numerical value determined based on the information and the control parameter output from the control map. The control map optimization apparatus according to claim 3, wherein the control parameter values set at the grid points of the control map are updated as described above. The said information is a numerical value determined based on the result of the trial of the control performed according to the control parameter output from the said control map, The Claim 1 characterized by the above-mentioned. Control map optimization device. The information is an index to increase or decrease a control parameter set in a grid point near the map position corresponding to the argument in the control map,
The control map optimizing device according to any one of claims 1 to 4, wherein the optimization unit determines a predetermined numerical value based on the index and creates the error function. When the number of the arguments input to the control map is n, the optimization means has 2 ⁿ pieces around the argument that are lattice points in the vicinity of the map position corresponding to the argument in the control map. The control map optimization apparatus according to any one of claims 1 to 6, wherein the control parameter set to the lattice point is updated. The optimization means stores the update amount of the control parameter value without updating the control parameter value during the control operation, instead of performing the update process every time the control parameter value is updated. control map optimization apparatus according to any one of claims 1 to 7, characterized in that updating the value of said control parameter collectively leave after control operation ends. The optimization means performs the update process every time the control parameter value is updated in the initial stage of the iteration, and does not update the control parameter value during the control operation in the later stage of the iteration. 8. The control according to claim 1, wherein an update amount of the value of the control parameter is stored in the control parameter, and the value of the control parameter is updated in a lump after completion of the control operation. Map optimization device.

Images