JP2010013990A - Vehicle control device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a vehicle control device using hybrid predictive control and suppressing an increase in the number of areas and an increase in a computational quantity as much as possible. <P>SOLUTION: This vehicle control device provided with at least an engine and a transmission includes a control system regarding a portion to a vehicle body from the engine described by a state equation composed of input (u) and a state (x) as a controlling object and performing feedback control so that output (y) corresponds to the input (u). In order to optimize a change in the state (x) according to a performance function, the revised value of the input (u) obtained by searching a controller group with the optimum solution group of the input (u) described by u=Kx composed of the product of the state (x) and the optimum gain K, previously generated off-line, is applied as a manipulated variable from the control system to the controlling object, and a controller with the physically impossible state (x) described in a state space area is eliminated. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a vehicle control device, and more particularly, to a device that uses hybrid predictive control and suppresses an increase in the number of regions and an increase in calculation amount as much as possible.

  Model predictive control is a method of determining an operation amount while predicting future controlled amounts and other state changes based on a model to be controlled. The prediction interval is not infinite, and a finite-length prediction interval called horizon is set, and the prediction and operation are repeated. However, a large amount of calculation is required to calculate the state and the operation amount at all sample times in the prediction interval although it is a finite length.

Therefore, in the technique described in Patent Document 1 below, it is proposed to reduce the amount of calculation by dividing the controlled amount into a fast change and a slow change and by thinning out the predicted time for the slow change. Has been.
Japanese Patent Laid-Open No. 10-3302

  However, the method described in Patent Document 1 cannot be applied unless there are a plurality of controlled amounts and there is a large difference in their change speeds. For example, there is no significant difference in one input / output system and the change speed. It cannot be applied to things.

  In addition to the method described in Patent Document 1, as a method for calculating the operation amount of model predictive control, a controller is previously obtained offline, and when calculating the operation amount online, to which state space region from each current state A method called multi-parametric programming has been proposed, in which it is determined whether it belongs, and an operation amount is obtained using a controller corresponding to the area. In this method, since it is not necessary to perform the prediction calculation of the controlled amount or other states online, the calculation time can be greatly reduced.

  On the other hand, in this method, it is known that the number of areas for switching the controller increases rapidly as the prediction interval (horizon) is increased or the prediction problem becomes more complicated by setting more constraints. There is a problem that a storage area for holding the gain is enlarged. In addition, as a result of the increase in the number of regions, there is another problem that the search time for selecting a region increases.

  SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a vehicle control apparatus that eliminates the above-described disadvantages, uses hybrid predictive control, and suppresses an increase in the number of regions and an increase in calculation amount as much as possible.

  In order to achieve the above object, according to claim 1, in a control device for a vehicle including at least an engine, at least the engine is modeled and described by a state equation including an input u and a state x. The change in the state x is optimized according to a control system that performs feedback control so that the output y matches the input u, and an evaluation function that evaluates at least the change in the state x from the current time to the n-step future. As described above, a controller group that is generated offline in advance and stores a group of controllers in which u = Kx, which is a product of the state x and the optimal gain K, is stored in a state space region including the state x. And a correction value of the input u obtained by searching the optimum solution group of the input u from the state x is applied as an operation amount from the control system to the control target. With and an operation amount applying means for, and as configured to remove the controller describing the state x that can not occur physically in the state space region.

  According to claim 2, in a control device for a vehicle including at least an engine, a control system that is modeled and described by a state equation consisting of an input u and a state x, and at least the engine is controlled, An optimal solution group of the input u, which has been generated off-line in advance so as to optimize the change of the state x in accordance with an evaluation function that evaluates at least the change of the state x from the current time to n-step future, And storage means for storing a controller group described by u = Kx, which is a product of the optimal gain K, in the state space region consisting of the state x, and the optimal solution group of the input u obtained by searching the state x. An operation amount applying means for applying the correction value of the input u as an operation amount to the control target from the control system, and between adjacent regions in the state space region When the ratio of optimum gain K1, K2 is equal to or less than a predetermined value, and as configured as one area by combining the two regions.

  According to claim 3, in a control device for a vehicle including at least an engine, a control system which is modeled and described by a state equation consisting of an input u and a state x and which controls at least the engine, An optimal solution group of the input u, which has been generated off-line in advance so as to optimize the change of the state x in accordance with an evaluation function that evaluates at least the change of the state x from the current time to n-step future, And storage means for storing a controller group described by u = Kx, which is a product of the optimal gain K, in the state space region consisting of the state x, and the optimal solution group of the input u obtained by searching the state x. Operation amount applying means for applying the corrected value of the input u as an operation amount to the control target from the control system, and the product of the state x and the optimum gain K is a predetermined value or less. When, and as it configured to zero the optimum gain K.

  According to a fourth aspect of the present invention, in a control device for a vehicle including at least an engine, at least the engine, which is modeled and described by a state equation composed of an input u and a state x, is controlled, and an output y is the input u. So that the change in the state x is optimized according to a first control system that performs feedback control every control cycle kT so as to match with the evaluation function that evaluates at least the change in the state x from the current time to the n-step future. Storage means for storing, in a state space region consisting of the state x, a controller group which is generated in advance and describes the optimal solution group of the input u by u = Kx, which is the product of the state x and the optimal gain K Correction value calculation means for calculating a correction value of the input u obtained by searching the optimal solution group of the input u from the state x, the control system and the control pair And a second control system that performs feedback control every control cycle T that is 1 / k times the control cycle kT so that the output y matches the correction value of the input u. It comprised so that it might be equipped with.

  In the vehicle control device according to claim 1, a control system that performs feedback control so that at least an engine described by a state equation consisting of an input u and a state x is controlled and an output y matches the input u is provided. In order to optimize the change of the state x according to the evaluation function, the optimal solution group of the input u generated in advance offline according to the multiparametric programming method is described by u = Kx consisting of the product of the state x and the optimal gain K The correction value of the input u obtained by searching the controller group from the state space region is applied to the control object as the manipulated variable, and describes a state that cannot physically occur in the state space region. Since the controller corresponding to is deleted, an increase in the number of areas and an increase in the calculation amount can be suppressed as much as possible.

  In other words, the hybrid predictive controller has a format for switching the optimum controller in accordance with the combination of states, but since the combination of states is generated exhaustively, an area that does not physically exist is also included. By deleting the controller in such an area, it is possible to reduce the storage area for holding the gain and reduce the amount of calculation.

  In addition, in developing a vehicle control device, it was difficult to perform adaptive control according to changes in the situation, in addition to requiring a huge amount of man-hours in order to respond to changes in driving characteristics and driving environments. Such inconvenience can also be eliminated.

  In the vehicle control device according to claim 2, a control system that performs feedback control so that at least the engine described by the state equation composed of the input u and the state x is controlled and the output y matches the input u is provided. In order to optimize the change of the state x according to the evaluation function, the optimal solution group of the input u generated in advance offline according to the multiparametric programming method is described by u = Kx consisting of the product of the state x and the optimal gain K A correction value of the input u obtained by searching a controller group to be searched from the state space region is applied as an operation amount to the control object, and the ratio of the optimum gains K1 and K2 between adjacent regions in the state space region is equal to or less than a predetermined value. At this time, since the two regions are combined to form one region, similarly, an increase in the number of regions and an increase in the amount of calculation can be suppressed as much as possible.

  In other words, the hybrid predictive controller can be obtained as a switching controller as described above. However, in order to satisfy mathematical optimality, a controller with a slightly different value is generated even if it can be regarded as substantially the same controller. There are many things to do. Therefore, when the ratio with the optimum gain of the adjacent area is sufficiently small, the storage area holding the optimum gain can be reduced and the amount of calculation can be reduced by assuming that the ratio is the same.

  In the vehicle control device according to claim 3, a control system that performs feedback control so that at least an engine described by a state equation consisting of an input u and a state x is controlled and an output y matches the input u is provided. In order to optimize the change of the state x according to the evaluation function, the optimal solution group of the input u generated in advance offline according to the multiparametric programming method is described by u = Kx consisting of the product of the state x and the optimal gain K The correction value of the input u obtained by searching the controller group from the state space region is applied to the controlled object as an operation amount, and when the product of the state x and the optimum gain K is equal to or less than a predetermined value, the optimum gain K is set to zero. Since it is configured as described above, similarly, the increase in the number of regions and the increase in the amount of calculation can be suppressed as much as possible.

  That is, since the hybrid predictive controller is given an operation amount by the product sum of a plurality of optimum gains and states, if the product of the optimum gain corresponding to a certain state is small, it is assumed that the control effect has no effect and is treated as 0. Thus, it is possible to reduce the storage area for holding the optimum gain and reduce the amount of calculation.

  In the vehicle control device according to claim 4, at least an engine that is modeled and described by a state equation including an input u and a state x is a control target, and a control cycle is performed so that an output y matches the input u. A first control system that performs feedback control for each kT, and an optimal solution group of the input u that has been generated off-line in advance according to the multiparametric programming method so as to optimize the change of the state x according to the evaluation function is referred to as the state x. A correction value calculation means for calculating a correction value of the input u obtained by searching the controller group described by u = Kx consisting of the product of the optimal gain K from the state space region, and an interpolated between the control system and the control target. And a second control system that performs feedback control every control cycle T obtained by multiplying the control cycle kT by 1 / k so that the output y matches the corrected value of the input u. Since form was, it is possible to suppress as much as possible similar increase increase and calculation of number of areas.

  That is, since the input (control command value) of the vehicle control device is generally updated in the millisecond order, for example, when it is desired to predict the second order, a horizon of 100 times or more is required. However, when such a long horizon is set, it becomes a very complicated prediction problem, and there is a disadvantage that it becomes a controller having an extremely large number of regions, and it is often impossible to solve the prediction problem numerically.

  Therefore, instead of the configuration in which the output of the hybrid prediction controller is directly input to the control target (operation amount), the original target value (input) given to the hybrid prediction controller is corrected in view of the evaluation function J and the like. A lower-layer controller (local controller, second control system) that performs feedback with the operation amount as a target value is arranged, and the lower-layer controller performs control in the original sampling time.

  As a result, it is possible to achieve both the reduction in the number of regions and the reduction in the amount of calculation due to the reduction in the prediction frequency of the hybrid prediction control, and the update frequency of the operation amount. In particular, in the case of vehicle control, it is difficult to secure a sufficient prediction interval for a relatively high-response control target. However, as described above, by increasing the sampling time for prediction control, the number of regions can be increased. By making the local controller determine the final operation amount while suppressing the increase in the calculation amount and securing the prediction interval, it is possible to avoid a decrease in control performance due to a longer control cycle.

  The best mode for carrying out a vehicle control apparatus according to the present invention will be described below with reference to the accompanying drawings.

  FIG. 1 is a schematic diagram showing a vehicle control apparatus according to a first embodiment of the present invention as a whole by taking a driving force control apparatus as an example.

  In FIG. 1, the code | symbol 10 shows an engine (internal combustion engine). The engine 10 is a spark-ignition water-cooled gasoline engine having a plurality of cylinders such as four cylinders, and is mounted on a vehicle body (not shown) of a vehicle (partially indicated by drive wheels W).

  A throttle valve (not shown) arranged in the intake system of the engine 10 is mechanically disconnected from an accelerator pedal (not shown) arranged in the vehicle driver's seat, and is a DBW (actuator) such as an electric motor. Drive By Wire) mechanism 16 is connected and driven.

  The intake air metered by the throttle valve flows through an intake manifold (not shown) and mixes with fuel injected from an injector (fuel injection valve) 20 near the intake port of each cylinder to form an air-fuel mixture, When an intake valve (not shown) is opened, it flows into a combustion chamber (not shown) of the cylinder. In the combustion chamber, the air-fuel mixture is ignited and combusted, and after driving a piston (not shown) to rotate the crankshaft 22, exhaust gas is discharged outside the engine 10.

  The rotation of the crankshaft 22 of the engine 10 is input to the transmission 26 via the torque converter 24. The transmission 26 changes the output of the engine 10.

  In other words, the crankshaft 22 is connected to the pump impeller 24a of the torque converter 24, while the turbine runner 24b disposed opposite thereto and receiving the fluid (hydraulic oil) is connected to the main shaft (mission input shaft) MS. The torque converter 24 includes a lockup clutch (not shown).

  The transmission 26 is composed of CVT (Continuous Variable Transmission), a drive pulley 26a disposed on the main shaft MS, a driven pulley 26b disposed on a countershaft CS parallel to the main shaft MS, and a metal hung between them. And a hydraulic mechanism 26d for supplying hydraulic oil to a piston chamber of a movable pulley half (described later).

  The drive pulley 26a includes a fixed pulley half 26a1 disposed on the main shaft MS and a movable pulley half 26a2 that can move relative to the fixed pulley half 26a1 in the axial direction. The driven pulley 26b includes a fixed pulley half 26b1 fixed to the countershaft CS and a movable pulley half 26b2 that can move relative to the fixed pulley half 26b1 in the axial direction.

  The CVT 26 is connected to the forward / reverse switching device 30. The forward / reverse switching device 30 includes a forward (forward) clutch 30a, a reverse brake 30b, and a planetary gear mechanism 30c disposed therebetween.

  In the planetary gear mechanism 30c, the sun gear 30c1 is fixed to the main shaft MS, and the ring gear 30c2 is fixed to the fixed pulley half 26a1 of the drive pulley 26a via the forward clutch 30a.

  A pinion 30c3 is disposed between the sun gear 30c1 and the ring gear 30c2. Pinion 30c3 is coupled to sun gear 30c1 by carrier 30c4. The carrier 30c4 is fixed (locked) when the reverse brake 30b is operated.

  The rotation of the counter shaft CS is transmitted to the secondary shaft SS via the reduction gears 34 and 36, and the rotation of the secondary shaft SS is transmitted to the left and right drive wheels (tires, only shown on the right side) W via the gear 40 and the differential D. It is done. A disc brake 42 is disposed in the vicinity of the drive wheel W.

  Switching between the forward clutch 30a and the reverse brake 30b is performed by the driver operating a shift lever 44 provided at a vehicle driver's seat, for example, having positions P, R, N, D, S, and L. That is, when any position of the shift lever 44 is selected by the driver, the selection operation is transmitted to a manual valve (not shown) of a hydraulic mechanism (not shown).

  For example, when the D, S, and L positions are selected, the spool of the manual valve moves accordingly and hydraulic oil (hydraulic pressure) is discharged from the piston chamber of the reverse brake 30b, while the hydraulic pressure is applied to the piston chamber of the forward clutch 30a. Is supplied and the forward clutch 30a is engaged. When the forward clutch 30a is engaged, all gears rotate together with the main shaft MS, and the drive pulley 26a is driven in the same direction (forward direction) as the main shaft MS.

  On the other hand, when the R position is selected, hydraulic oil is discharged from the piston chamber of the forward clutch 30a, while hydraulic pressure is supplied to the piston chamber of the reverse brake 30b to operate the reverse brake 30b. As a result, the carrier 30c4 is fixed, the ring gear 30c2 is driven in the opposite direction to the sun gear 30c1, and the drive pulley 26a is driven in the opposite direction (reverse direction) to the main shaft MS.

  When the P or N position is selected, the hydraulic oil is discharged from both piston chambers, the forward clutch 30a and the reverse brake 30b are both released, and the power transmission via the forward / reverse switching device 30 is cut off. Power transmission between the engine 10 and the drive pulley 26a of the CVT 26 is interrupted.

  In the CVT 26, when hydraulic oil is supplied from the hydraulic mechanism 26d to the piston chambers of the movable pulley halves 26a2 and 26b2, and pulley side pressure (belt transmission torque) that moves the movable pulley halves 26a2 and 26b2 in the axial direction is generated, The pulley widths of the drive pulley 26a and the driven pulley 26b change, and the winding radius of the belt 26c changes.

  Thus, the gear ratio for transmitting the output of the engine 10 to the drive wheels W can be changed steplessly by adjusting the pulley side pressure, in other words, by changing the belt transmission torque command value.

  A crank angle sensor 48 is provided at an appropriate position such as near the cam shaft (not shown) of the engine 10 and outputs a signal indicating the engine speed NE for each predetermined crank angle position of the piston. In the intake system, an absolute pressure sensor 50 is provided at an appropriate position downstream of the throttle valve, and outputs a signal proportional to the intake pipe absolute pressure (engine load) PBA.

  The actuator of the DBW mechanism 16 is provided with a throttle opening sensor 52 and outputs a signal proportional to the throttle opening THL through the amount of rotation of the actuator, and an accelerator opening sensor 54 is provided in the vicinity of the accelerator pedal. A signal proportional to the accelerator pedal opening AP corresponding to the accelerator pedal operation amount (accelerator pedal depression amount) is output.

  Further, a water temperature sensor 56 is provided in the vicinity of a cooling water passage (not shown) of the engine 10 to generate an output corresponding to the engine cooling water temperature TW, in other words, the temperature of the engine 10, and the intake air temperature in the intake system. A sensor 58 is provided and generates an output corresponding to the intake air temperature (outside air temperature) taken into the engine 10.

  The output of the crank angle sensor 48 and the like described above is sent to the engine controller 60. The engine controller 60 includes a microcomputer having a CPU, a ROM, a RAM, etc., determines the fuel injection amount based on the sensor output, drives the injector 20, and opens the throttle valve according to the accelerator pedal opening AP. Control THL.

  The main shaft MS is provided with an NT sensor (rotational speed sensor) 62, which indicates the rotational speed of the turbine runner 24b, specifically the rotational speed of the main shaft MS, more specifically, the input shaft rotational speed of the forward clutch 30a. Outputs a pulse signal.

  An NDR sensor (rotational speed sensor) 64 is provided at an appropriate position in the vicinity of the drive pulley 26a of the CVT 26 to output a pulse signal corresponding to the rotational speed of the drive pulley 26a, in other words, the output shaft rotational speed of the forward clutch 30a. At the same time, an NDN sensor (rotational speed sensor) 66 is provided at an appropriate position near the driven pulley 26b to output a pulse signal indicating the rotational speed of the driven pulley 26b.

  A VEL sensor (rotational speed sensor) 68 is provided in the vicinity of the gear 36 of the secondary shaft SS, and outputs a pulse signal indicating the output rotational speed of the CVT 26 or the vehicle speed VEL through the rotational speed of the gear 36. A shift lever position sensor 70 is provided in the vicinity of the shift lever 44 described above, and outputs a POS signal corresponding to a position such as R, N, or D selected by the driver.

  A brake opening sensor 72 is provided near the brake pedal, and outputs a signal proportional to the brake pedal opening BRK corresponding to the driver's brake pedal operation amount (brake pedal depression amount).

  The output of the NT sensor 62 and the like (and the output of the accelerator opening sensor 54 described above) is sent to the shift controller 74 including the outputs of other sensors (not shown). The shift controller 74 also includes a microcomputer having a CPU, ROM, RAM, and the like, and is configured to be communicable with the engine controller 60.

  Among the input sensor outputs, the outputs of the NT sensor 62, the NDR sensor 64, and the VEL sensor 68 are input to the waveform shaping circuit in the shift controller 74. The shift controller 74 counts the output of the waveform shaping circuit and detects the rotational speed and the vehicle speed V.

  Based on these detected values, the shift controller 74 controls the operation of the torque converter 24 and the CVT 26 by energizing / de-energizing an electromagnetic solenoid (not shown) of the hydraulic mechanism, and determines the engagement of the forward clutch 30a. To decide.

  Regarding the control of the operation of the CVT 26, the shift controller 74 determines a target gear ratio (ratio), drives the movable pulley halves 26a2 and 26b2 so as to obtain the determined target gear ratio, and controls the gear ratio.

  Hereinafter, the design of the shift controller 74 and the engine controller 60 will be described.

  In order to design the shift controller 74 as a controller based on the control theory, first, physical modeling of the controlled object is performed to derive a state equation.

  As will be described later, since one of the purposes is to reduce the driver's operation by appropriately controlling the gear ratio, the vehicle including the accelerator pedal opening AP and the brake pedal opening BRK is shown in FIG. 14 is modeled as a control target.

  The engine 10 and the torque converter 24 have different characteristics depending on whether the fuel cut is OFF (fuel cut (fuel supply stop) is not executed) or ON (fuel cut is executed). I do.

  Hereinafter, modeling of the engine controller 60, the engine 10, the torque converter 24, the transmission (CVT) 26, and the vehicle body will be described.

  Modeling of the engine controller 60 will be described.

  As described above, the engine controller 60 controls the throttle valve opening THL according to the DBW map based on the accelerator pedal opening AP operated by the driver as shown in FIG. Although the illustration of the characteristics of the DBW map is omitted, since the relationship between AP and THL can be considered to be almost linear, as shown in Equation 1, the accelerator pedal opening AP is multiplied by a coefficient ka and treated as a linear approximation.

  Modeling of the engine 10 will be described.

  Since the engine 10 has different characteristics depending on whether the fuel cut is OFF or ON, modeling is performed for each case. First, when the fuel cut is OFF, as shown in FIG. 4, the engine torque TENG is output as a model depending on the throttle opening THL and the engine speed ωENG. As shown in Equation 2, the engine torque when the fuel cut is OFF is denoted as NormalTrq.

  When the fuel cut is ON, as shown in FIG. 5, it is modeled as one that outputs the friction torque FrictionTrq depending on the engine speed ωENG.

  The modeling of the torque converter 24 will be described.

  As shown in FIG. 6, the torque converter 24 receives the engine torque TENG and the output shaft rotational speed ωTC of the torque converter, and outputs the torque TTC output from the torque converter 24 and the engine rotational speed ωENG (= input shaft rotational speed of the torque converter 24). ) As an output model.

  The input shaft rotational speed of the torque converter 24, that is, the engine rotational speed ωENG is obtained as follows.

  That is, the input torque to the pump impeller 24a of the torque converter 24 is the engine torque, and the output torque is the pump torque Tpump and the lockup clutch torque TLC. Accordingly, the engine rotational acceleration ωENGdot is obtained by dividing the difference between the input and output torques by the inertia IENG_pump of the engine 10 and the pump impeller 24a. Therefore, the engine rotational speed ωENG is obtained by integrating it.

  Further, since the pump torque Tpump and the turbine torque Tturbine have different characteristics when the fuel cut is OFF and when the fuel cut is ON, modeling is performed for each case.

  First, when the fuel cut is OFF, the pump torque Tpump is given by the product of the square of the engine speed ωENG and the capacity coefficient τTC of the torque converter 24. The turbine torque Tturbine is obtained by multiplying the pump torque Tpump by the torque ratio KTC. Since the capacity coefficient τTC and the torque ratio KTC depend on the speed ratio e of the torque converter 24, the capacity coefficient τTC is obtained by multiplying the speed e.

  Here, the speed ratio e is a value obtained by dividing the output shaft rotational speed ωTC of the torque converter by the input shaft rotational speed ωENG, as shown in the following equation.

  Next, when the fuel cut is ON, the relationship between the pump impeller 24a and the turbine runner 24b is reversed. That is, the turbine torque Tturbine is given by the product of the square of the output shaft speed ωTC, the torque converter capacity coefficient τTC, and the speed ratio e minus one, and the pump torque Tpump is the turbine torque Tturbine and the torque ratio KTC. It is obtained by multiplying by the speed ratio e.

  Further, the lockup clutch torque is modeled as being given by multiplying the maximum transmission torque TLCMAX of the lockup clutch by the difference between the input / output shaft rotational speeds ωENG and ωTC of the torque converter 24 and multiplying by the coefficient λLC.

  The output torque TTC of the torque converter 24 is given by the sum of the turbine torque Tturbine and the lockup clutch torque TLC.

  Modeling of the transmission (CVT) 26 will be described.

  Here, the transmission 26 includes a forward clutch 30a (of the forward / reverse switching device 30), pulleys (drive / driven pulleys) 26a and 26b, secondary gears (reduction gears 34 and 36), and final gear (gear 40). Shall be. As shown in FIG. 7, the transmission 26 is modeled as a model that receives the output torque of the torque converter 24 and the rotation speed of the tire (drive wheel) W and outputs the drive shaft torque TDS and the torque converter rotation speed ωTC.

  The output shaft rotational speed ωTC of the torque converter 24 is obtained as follows.

  First, it is assumed that the turbine speed is equal to the forward clutch speed. The input torque to the forward clutch 30a is obtained by subtracting the friction loss torque Tpumploss (= Oilpumploss) due to the oil pump from the output torque TTC of the torque converter 24 and multiplying by the forward gear efficiency ηFWDRVS. It is assumed that the friction loss torque Oilpumploss by the oil pump is given depending on the output shaft rotational speed ωTC of the torque converter 24.

  On the other hand, the output torque is the output torque TFWDRVS of the forward clutch 30a. Accordingly, the output shaft rotational acceleration ωTCdot of the torque converter 24 can be obtained by dividing the difference between the input and output torques by the inertia IFWDRVS of the forward clutch 30a. Therefore, by integrating this, the output shaft rotational speed ωTC of the torque converter 24 is obtained.

  The forward clutch torque TFWDRVS is modeled as being given by multiplying the maximum transmission torque TFWDMAX of the forward clutch 30a by the difference between the input / output shaft rotational speeds ωTC and ωFWDRVS of the forward clutch 30a and multiplying by the coefficient λFWD.

  Next, transmission of the rotation speed will be described.

  The rotational speed ωDN of the driven pulley 26b is obtained by multiplying the tire rotational speed ωtire by the final gear ratio ρF and the secondary gear ratio ρS. The rotational speed ωDR of the drive pulley 26a is obtained by multiplying the driven pulley rotational speed ωDN by the pulley speed ratio ρpulley. The rotational speed ωFWDRVS of the forward clutch 30a is obtained by multiplying the drive pulley rotational speed ωDR by the forward gear ratio ρFWD.

  In summary, the forward clutch rotational speed ωFWDRVS is obtained by multiplying the tire rotational speed ωtire by the forward gear ratio ρFWD, the pulley gear ratio ρpulley, the secondary gear ratio ρS, and the final gear ratio ρF.

The pulley speed ratio ρpulley, more specifically, the speed ratio ρpulley, follows the target pulley speed ratio, more specifically, the target speed ratio ρpulleytgt with a first-order lag.

  Next, torque transmission will be described.

  The drive pulley torque TDR is obtained by multiplying the forward clutch torque TFWDRVS by the forward gear ratio ρFWD. The driven pulley torque TDN is obtained by multiplying the drive pulley torque TDR by the pulley ratio ρpulley and the CVT efficiency ηCVT. The drive shaft torque TDS is obtained by multiplying the driven pulley torque TDN by the secondary gear ratio ρS and the final gear ratio ρF.

  In summary, the drive shaft torque TDR is obtained by multiplying the forward clutch torque TFWDRVS by the secondary gear ratio ρS, final gear ratio ρF, CVT efficiency ηCVT, pulley speed ratio ρpulley, and forward gear ratio ρFWD.

  Car body modeling will be described.

  As shown in FIG. 8, the vehicle body is modeled as a model having the drive shaft torque TDS and the braking force FBRK as inputs and the tire rotational speed ωtire as an output. Here, only the front-rear direction is handled. The vehicle body acceleration Vdot is obtained by subtracting each running resistance and the braking force FBRK from the driving force and dividing it by the vehicle weight mbody. The driving force is obtained by dividing the drive shaft torque TDS by the tire radius Rtire. The vehicle speed V can be obtained by integrating the vehicle body acceleration Vdot.

  Here, regarding the running resistance, a gradient resistance Fslope, a rolling resistance Froad, and an air resistance Fair are considered. The gradient resistance Fslope is given by multiplying the vehicle weight mbody by the gravitational acceleration g and multiplying by the sine of the gradient θslope. The rolling resistance Froad is obtained by multiplying the vehicle weight mbody by the rolling resistance coefficient μroad and the gravitational acceleration g and multiplying by the cosine of the gradient. The air resistance Fair is obtained by multiplying the square value of the vehicle speed V by the air resistance coefficient μair.

  The braking force FBRK can be obtained linearly with respect to the brake pedal opening by multiplying the brake pedal opening BRK by a coefficient Fbrakemax for the sake of simplicity.

  The tire rotational speed ωtire is obtained by dividing the vehicle speed V by the tire radius Rtire.

  Since the mathematical model of the engine 10, the torque converter 24, the transmission (CVT) 26, and the vehicle body has been described above, a state equation that combines them is derived here.

  In the case of fuel cut OFF, the model is derived with the input u as [AP, ρpulleytgt] and the state x as [Vdot, ωENG, ωTC, ρpulley].

  Similarly, in the case of fuel cut ON, the model is derived with the input u as [BRK, ρpulleytgt] and the state x as [Vdot, ωENG, ωTC, ρpulley]. Thus, the input u of the state equation is the target gear ratio ρpulley and the accelerator pedal opening AP or the target gear ratio ρpulley and the brake pedal opening BRK, and the state x is the vehicle body acceleration Vdot, the engine speed ωENG, and the torque converter output. The shaft speed ωTC and the gear ratio ρpulley are assumed.

  As is apparent from the above, this controlled object is characterized by being switched by fuel cut OFF / ON. Therefore, this control target is regarded as a hybrid system in which the model is switched by fuel cut OFF / ON. In other words, this controlled object is modeled separately when the fuel cut is executed and when it is not executed.

  Next, the design of the gear ratio control system (shift controller 74) will be described.

  Here, the gear ratio control system is configured as shown in FIG. First, as shown in the table at the bottom of the figure, a representative target speed ratio ρpulley-ref is set based on the vehicle speed V, more precisely, the reference vehicle speed Vref. In the subsequent stage, a configuration in which a constraint condition and an evaluation function described later are considered with respect to the value, that is, a corrected target gear ratio is given to the control target.

  Here, as the reference vehicle speed for determining the target gear ratio, the current vehicle speed itself or the future target vehicle speed in consideration of the accelerator pedal opening or the brake pedal opening and the current gear ratio may be given.

  Further, as will be described later, in order to improve fuel efficiency by appropriately controlling the speed ratio, the engine speed obtained from the fuel efficiency optimum operating line is also given as a target value, and the target speed ratio is also considered in consideration of the target engine speed. Is generated.

  The required specifications for the gear ratio control system will be described.

  First, since the transmission ratio is fed back, an evaluation function Jρpulley for evaluating the difference between the representative target transmission ratio ρpulley-ref and the transmission ratio ρpulley is set.

  One of the requirements for speed ratio control is a reduction in the amount of driver operation. For example, when you want to accelerate greatly, if you want to accelerate with only the accelerator pedal without changing the gear ratio, you must depress the accelerator pedal greatly, but by temporarily changing the gear ratio to the low side, you can reduce the number of accelerator pedals. The same acceleration force can be obtained by stepping on. Thus, by appropriately changing the gear ratio, it is possible to reduce the amount of driver operation, that is, reduce the driving burden.

  In order to realize this, an evaluation function JABPRK for the driver operation amount is set as follows.

  Here, ΔAP (k + 1 | k) or ΔBRK (k + 1 | k) represents an i-step future predicted value when the current time is k. Therefore, the above equation means that the sum of squares of the predicted values of ΔAP or ΔBRK of each step from ΔAP or ΔBRK at the current time k to the future of Hu−1 steps is obtained. That is, it is a function that evaluates the sum of driver operation amounts from the current time to the Hu-1 step future. By controlling the gear ratio so as to minimize this value, the driver operation amount, that is, the driving burden is reduced. Reduction is possible.

  Another requirement is an improvement in fuel consumption. In order to achieve good fuel efficiency, it is known that the engine may be operated along a fuel efficiency optimum operating line as shown in FIG. 10 by appropriately controlling the speed ratio.

  One method for this is to obtain the optimal engine speed for the target fuel economy from the characteristics of the optimal fuel efficiency line according to the current engine torque, and to control the gear ratio so that the actual engine speed matches it. Can be considered. However, always trying to follow the fuel efficiency optimum operating line has a disadvantage that the ride comfort is uncomfortable, and another problem arises such that a sufficient acceleration force cannot be obtained even if acceleration is requested.

  Therefore, instead of always trying to follow the optimum fuel consumption operating line, control is performed so that the difference between the optimum fuel consumption operating line and the actual operating point of the engine is as small as possible as a sum in a certain time range. In this way, for example, when there is an acceleration request, the driver is temporarily operated at a driving point away from the optimum fuel consumption operating line, and immediately after the acceleration is completed, the driver is quickly returned to the vicinity of the optimum fuel consumption operating line. This makes it possible to improve fuel efficiency while satisfying these requirements. In order to realize this, the following evaluation function JENG is set.

  The above equation is a function JENG that evaluates the sum of the difference ωENG-ref between the fuel efficiency optimum operating line from the current time to the Hp-1 step future and the actual engine operating point, and the gear ratio is set to minimize this evaluation function. By controlling the above, it becomes possible to realize the previous request.

  Further, inadvertently changing the gear ratio is not preferable from the viewpoint of riding comfort. Therefore, the evaluation function Jρpulley is also set for the target gear ratio ρpulleytgt.

  As described above, the evaluation function for evaluating each of the gear ratio, the driver's operation amount, the fuel consumption (engine speed), and the target gear ratio is set. However, for example, when accelerating over a long period of time, to reduce the amount of driver operation, it is necessary to maintain the gear ratio on the low side to ensure driving force, but on the other hand, driving away from the optimal fuel consumption operating line If the operation is continued at a point, the fuel efficiency is greatly deteriorated, so the gear ratio cannot be kept low.

  Thus, there are cases where each evaluation function has a trade-off relationship. Therefore, instead of evaluating individual evaluation functions, a new evaluation function is set by weighting and adding each evaluation function.

  In this way, it is possible to simultaneously optimize the respective evaluation functions while considering which characteristics are more important by adjusting the weights Q and R.

  When the gear ratio is changed, the engine speed changes for the same vehicle speed. For example, if the gear ratio is changed to the low side when the vehicle speed is very high, the engine speed becomes very high. However, the engine speed has a rev limit speed, and operation exceeding this speed must be avoided to protect the engine.

  Conversely, when the gear ratio is changed to the high side when the vehicle speed is very low, if the engine to the tire remains directly connected, the engine speed will be very low, which may cause engine stall. This must also be avoided.

  Also, the speed ratio itself has a limited range in which the speed can be changed, and it is necessary to shift within the range.

  Furthermore, it is not preferable to change the gear ratio too rapidly from the viewpoint of ride comfort and transmission protection. In order to realize the above, constraint conditions are provided for each of the engine speed ωENG, the speed ratio ρpulley, and the target speed ratio change speed Δρpulleytgt.

  That is, the target speed ratio ρpulley is limited by a limit value (target speed change speed) Δρpulleytgt given depending on the current speed ratio and the engine speed. By doing so, it is possible to perform gear ratio control in consideration of the above problems.

The design characteristics of the gear ratio control system in this embodiment are summarized as follows.
1. It will be possible to predict the future operation amount and condition.
2. Constraints are provided on the operation amount and the operating range of the state.
3. The control target is a hybrid system that is switched by fuel cut OFF / ON.

  1. One suitable means for this is model predictive control. As shown in FIG. 11, model predictive control is a technique for predicting future state trajectories sequentially and generating manipulated variables while optimizing the predicted trajectory online based on a certain evaluation index. be able to.

  However, since it is necessary to solve the optimal control problem online, it is often difficult to apply due to the problem of calculation time, and it is mainly limited to the use of operation of chemical plants. As an application to a vehicle, a system with relatively slow dynamics, such as the inter-vehicle control of the vehicle found in Patent Document 2, is a target.

  In recent years, attention has been focused on a method for solving an optimal control problem offline by using an optimization method called multiparametric programming as an alternative control method. Since the optimum solution can be obtained explicitly in a form depending on the initial state, the optimum value can be generated immediately by simply obtaining the state at every sampling time when mounting.

  In this embodiment, the shift controller 74 is designed in consideration of the constraint conditions and the evaluation function described above by using the multiparametric programming method.

  In addition, since it is possible to handle a hybrid system by the framework of the multiparametric programming method, the controller is designed in consideration of the model switching by the fuel cut OFF / ON.

  Hybrid predictive control based on multiparametric programming is a method of designing a controller for a linear discrete model. Since the previously derived equation of state includes nonlinear terms, it must be linearized. Since linearization is performed around the equilibrium point of the original nonlinear equation, it is necessary to first obtain the correspondence of each variable in the equilibrium state.

  Since each input and state do not change in the equilibrium state, the left side of the nonlinear equation is 0 because it is the differential value of each state. For example, regarding the gear ratio ρpulley, the correspondence between the actual gear ratio ρpulley and the target gear ratio ρpulleytgt is obtained as follows.

  The subscript e means a value in an equilibrium state. Similarly, when the correspondence relations are obtained and arranged for other formulas, the other states and inputs can be uniquely determined by providing an equilibrium point between the vehicle speed and the gear ratio.

  Next, how to give an equilibrium point between the vehicle speed and the gear ratio will be described. As already described, the representative speed ratio corresponding to the reference vehicle speed is obtained, and this is used to determine the corresponding speed ratio equilibrium point by giving the reference vehicle speed as the vehicle speed equilibrium point. As described above, only the equilibrium point of the vehicle speed is given and all other equilibrium points are uniquely determined.

The linear model derivation around the equilibrium point will be described. Taking the above transmission ratio as an example, δρpulley = ρpulley-ρpulley ^e is set, and a linear model for the change δρpulley from the equilibrium point ρpulley ^e is obtained. The same applies to other variables.

  From the above, the respective linear models of fuel cut OFF / ON are obtained. Also, when the fuel cut is OFF, the engine is on the rotating side, so ωENG ≧ ωTC. When the fuel cut is ON, the engine is on the rotating side, so ωENG <ωTC, so the magnitude of ωENG and ωTC From the relationship, the fuel cut OFF / ON switching is determined, and the following linear model is obtained.

_{Here, a 11a ~a 44a, b 21a} , b 42a, a 11b ~a 44b, b 11b, b 42b is given by linearization coefficients by linearization around the physical parameters and equilibrium. Furthermore, since the hybrid predictive control is a discrete design method, the linear model is discretized.

  Based on the above settings, a hybrid predictive controller based on multiparametric programming is obtained as the shift controller 74. The problem we are dealing with now is to design a servo controller that follows a time-varying target value, and it cannot be applied to multiparametric programming as it is, so the problem has been changed to a format that can be applied to multiparametric programming. To do.

  Δu (k) = u (k) −u (k−1) is set as a new input, and an expanded system having a state [δx (k); δu (k−1); δyref (k)] is configured.

  By setting the above-described evaluation function and constraint conditions for this extended system, the multiparametric programming problem is solved to obtain a vector of manipulated variables over the horizon interval. In the case of hybrid predictive control, only the operation amount Δu (k) at the initial step among the obtained operation amount vectors is added to the system. This can be obtained explicitly as follows.

  Here, Fi and Gi are controller gains, respectively. Since Δu (k) is the input difference from the previous step at the current time, the operation amount actually added to the system is as follows.

  When the fuel cut is OFF, the input to be controlled is [AP, ρpulleytgt], so when applied to the previous operation amount calculation formula, the following is obtained, and not only the target gear ratio but also the accelerator pedal opening is calculated. .

  Similarly, when the fuel cut is ON, since the input to be controlled is [BRK, ρpulleytgt], not only the target gear ratio but also the brake pedal opening is calculated.

  The reason for including the accelerator pedal opening and the brake pedal opening in the input is that the accelerator pedal opening and the brake pedal opening are treated as inputs, and this is incorporated into the evaluation function to minimize the amount of driver operation. This is to enable reduction gear ratio control. However, in this embodiment, since the accelerator pedal operation and the brake pedal opening are performed by the driver, the accelerator pedal opening and the brake pedal opening generated by the controller are not actually reflected in the controlled object.

  For example, when a controller is obtained by a general optimal control design, the operation amount is an optimal solution in an infinite time interval, so if both the operation amounts generated by the controller are not reflected in the control target, the optimality cannot be guaranteed. Of course, stability may not be maintained.

  However, in the case of predictive control, the optimization range is not an infinite time interval but a finite horizon interval, and an operation amount vector ΔU = [Δu (k), Δu (k + 1),..., At each step of the finite horizon interval. Δu (k + Hu−1)] is obtained, but actually, only the operation amount Δu (k) at the current time is used, and the operation amount after the next step proceeds to the next step, and is calculated again and obtained. In addition, an algorithm for recalculating the optimum operation amount according to the latest information sequentially.

  By extending this concept, the accelerator pedal opening and the brake pedal opening generated by the controller are not reflected in the controlled object. Therefore, the actual accelerator pedal operation and the brake pedal opening are performed by the driver, but the next step is to recalculate the optimal operation amount according to the latest information including the change in state due to the driver's operation. This is possible and should be repeated sequentially.

  However, it is guaranteed that the control can be performed without breaking the set constraint conditions when both operating amounts are used, and the constraint conditions may be broken if the accelerator pedal opening and the brake pedal opening are not used. There is.

  However, in the multiparametric programming method, the controller can be arranged only in the region within the constraint condition, and there is no controller in the region outside the constraint condition. Therefore, once it comes out of the area outside the constraint condition, the operation amount becomes 0 and there is a possibility that it may diverge.

  Therefore, another (second) stabilization controller is arranged in a region outside the constraint condition. Since this controller is only for returning the state to the constraint condition, it may be simple. For example, an optimal regulator that sets the speed ratio as the operation amount and returns each state to the origin may be arranged.

  Therefore, the hybrid predictive controller is used within the constraint condition, and when the condition is outside the constraint condition, the control is continued by switching to another stabilization controller. When the stabilization controller returns to the constraint condition again, the hybrid prediction controller is used again. In this way, control without failure is possible.

  As described above, it is possible to perform gear ratio control in consideration of reduction of the driver's operation amount.

  From here, it demonstrates using a specific numerical example. Here, the balance point of the gear ratio corresponding to the vehicle speed is set to (20, 0.60). As a result of deriving a linear model around this equilibrium point and further discretizing it, the following is obtained. Although the discretization is performed using the 0th-order hold here, a primary hold, bilinear transformation, or the like may be used.

Next, the weights Q and R for the evaluation function are set. Here, it sets as follows.
R ₁ = 10, R ₂ = 1000, Q ₁ = 1, Q ₂ = 300

Further, the control horizon Hc and the predicted horizon Hp are set as follows.
Hc = Hp = 3

  Further, the constraint conditions are set as follows.

  Based on the above settings, a hybrid predictive controller obtained by multiparametric programming is shown. The hybrid prediction controller is defined in a polyhedral region obtained by a combination of each state.

  FIG. 13 is a diagram showing how the region is divided with respect to three states of the engine speed, the torque converter output shaft speed, and the vehicle speed among the states. FIG. 13 shows that the area is finely divided according to the state. In the illustrated example, it is divided into 3397 areas. As shown in the table of FIG. 14, the controller is set for each area.

  Therefore, it is possible to perform optimum gear ratio control in consideration of the set evaluation function and constraint conditions by appropriately switching the controller in accordance with the fuel cut OFF / ON or other state changes.

  As shown in the figure, the control law of the controller is described so that the optimal solution group of the input u is given by the product of the state x and the optimal gain K (u = Kx). The control law of the controller shown in FIG. 14 is stored in the ROM of the microcomputer of the shift controller 74 so as to be searchable from the state space area shown in FIG.

  Here, to reiterate the problem of the present invention, as described above, the model predictive control sequentially predicts the future state trajectory, and optimizes the predicted trajectory online based on a certain evaluation index, while reducing the manipulated variable. This is a technique to generate, and the constraint conditions can be taken into account explicitly. However, since it is necessary to solve the optimal control problem online, it is often difficult to apply due to the problem of calculation time, and it is difficult to apply to a relatively fast system such as vehicle control.

  On the other hand, when the optimal control problem is solved off-line by multiparametric programming, the optimal solution can be obtained explicitly in a form that depends on the initial state. The optimal value can be generated immediately. Therefore, application to vehicle control is also possible from the viewpoint of calculation time.

  However, it is known that as the prediction problem becomes complicated by taking a longer prediction interval (horizon) or setting more constraint conditions, the number of regions for switching the controller increases rapidly, and as a result, the gain is maintained. The storage area will be enlarged. In addition, the amount of calculation for searching for the region also increases. In particular, when considering application to an embedded system such as vehicle control, it is not desirable that the (optimal) gain storage area is enlarged because the usable storage area is limited.

  Therefore, a controller is provided for each state space area according to the multiparametric programming method, and the increase in the number of areas is suppressed, and the gain is reduced to prevent the storage area from being enlarged, and the increase in the calculation amount is suppressed as much as possible. I tried to do it.

  This will be described below.

  In the controller design by the multiparametric programming method, it is possible to design the controller by imposing constraint conditions on each state and input. Now consider a simple example.

  For example, as shown in FIG. 15, there is a model having two states of x1 and x2, and the movable ranges thereof are assumed to be x1min ≦ x1 ≦ x1max and x2min ≦ x2 ≦ x2max. Therefore, this range may be set as a constraint condition in the controller design.

  However, in actuality, it is assumed that the hatched portion in FIG. 15 is a combination that cannot be physically obtained. Even in that case, since there is only a restriction on each state at the controller design stage, a controller is assigned to each grid including the hatched portion, and is deleted after the design.

  In the above-described embodiment, the vehicle speed, the engine speed, the torque converter speed, and the pulley gear ratio are selected as the states. Therefore, the controller is switched according to the combination of these values.

  For example, when the drive wheel (tire) W is directly connected from the engine 10, the relationship between the vehicle speed and the engine speed is determined by the gear ratio as shown in FIG. 16. Since the variable range of the gear ratio is finite, it cannot enter the shaded area in FIG.

  Therefore, in this embodiment, a combination that cannot be physically taken as indicated by the shaded portion in FIG. 15 is a controller that describes a state x that cannot be physically generated in the state space area. Removed.

  In this embodiment, as described above, in the control device for the vehicle 14 provided with at least the engine 10, more specifically, in the control device (shift controller 74) for controlling the driving force of the vehicle 14 through the gear ratio, A control system (speed ratio control system, FIG. 9) that is modeled and described by a state equation consisting of an input u and a state x and that controls at least the engine 10 and performs feedback control so that the output y matches the input u. ) And at least off-line generated in advance so that the change in the state x is optimal according to an evaluation function J (from Equation 19 to FIG. 22) that evaluates at least the change from the current time of the state x to n steps in the future. A controller group in which the optimal solution group of the input u is described by u = Kx, which is the product of the state x and the optimal gain K, is stored in the state space region including the state x. Storage means (ROM of the microcomputer of the shift controller 74) and the corrected value of the input u obtained by searching the optimum solution group of the input u from the state x as the manipulated variable from the control system to the control object. And an operation amount applying means for applying (processing performed in accordance with Equations 28 to 31) and a controller that describes a state x that cannot physically occur in the state space region is deleted. An increase and an increase in calculation amount can be suppressed as much as possible.

  That is, the controller obtained by the hybrid predictive control is in a form of switching the optimum controller according to the combination of states, but the combination of states is generated exhaustively, and therefore includes a region that does not physically exist. By deleting the controller in such an area, it is possible to reduce the storage area for holding the gain and reduce the amount of calculation.

  In addition, in developing a vehicle control device, it was difficult to perform adaptive control according to changes in the situation, in addition to requiring a huge amount of man-hours in order to respond to changes in driving characteristics and driving environments. Such inconvenience can also be eliminated.

  17 and 18 are explanatory views showing the configuration of the vehicle control apparatus according to the second embodiment of the present invention.

  As described in the first embodiment, the controller design based on multiparametric programming solves the optimization problem, so that a controller having slightly different values may be generated in order to satisfy mathematical optimization. .

  However, in many cases, the actual controlled object hardly reacts with such a slight difference. Or it may be rounded by the restriction | limiting of the number of digits, when mounting in an on-board microcomputer etc.

  Therefore, in the second embodiment, when the ratio to the (optimal) gain of adjacent regions is sufficiently small, the number of regions is reduced by replacing the average value of gains by assuming that they are the same. The storage area for storing was reduced. Also, the search time for the area is reduced by reducing the number of areas.

  For simplicity, consider a two-dimensional example as shown in FIG.

  In FIG. 17, since the ratio of adjacent gains is within a predetermined value (for example, 100 ± 1%), the portion surrounded by the thick line can be regarded as almost the same. In the illustrated example, since there are a plurality of gains for one region, it is assumed that the gains are the same when the ratios of all the gains are sufficiently small.

  Therefore, in the second embodiment, as shown in FIG. 18, these areas are combined and an average value of each gain before the combination is assigned. As a result, since the original gain of 16 was reduced by five, the storage area for storing the gain could be reduced by about 30%. In addition, since the number of areas is reduced, the area search time can also be reduced.

  As described above, in the second embodiment, a state equation which is modeled and includes the input u and the state x in the control device (shift controller 74) (controlling the driving force) of the vehicle 14 including at least the engine 10. And a control system (gear ratio control system, FIG. 9) that performs feedback control so that at least the engine 10 is to be controlled and the output y matches the input u, and at least n from the current time of the state x The optimal solution group of the input u generated in advance offline is set as the state x so as to optimize the change of the state x according to the evaluation function J (Equation 19 to FIG. 22) for evaluating the change to the step future. Storage means for storing a controller group described by u = Kx consisting of the product of the optimum gain K in the state space area consisting of the state x (the shift controller 74 (ROM of the microcomputer) and an operation amount applying means (several number) for applying the corrected value of the input u obtained by searching the optimum solution group of the input u from the state x to the control target from the control system as an operation amount When the ratio of the optimum gains K1 and K2 between adjacent regions in the state space region is equal to or smaller than a predetermined value (for example, 100 ± 1%), the two regions are Combined to form one region, more specifically, the two regions are combined into one region, and the average value of the optimum gains K1 and K2 is assigned. And an increase in calculation amount can be suppressed as much as possible.

  In other words, the hybrid predictive controller can be obtained as a switching controller as described above. However, in order to satisfy mathematical optimality, a controller with a slightly different value is generated even if it can be regarded as substantially the same controller. There are many things to do. Therefore, when the ratio with the (optimal) gain of adjacent areas is sufficiently small, the storage area for holding the gain can be reduced and the amount of calculation can be reduced by assuming that the ratio is the same.

  In the above description, the ratio of the gains K1 and K2 is described as being a predetermined value or less. However, the difference between the gains K1 and K2 may be a set value, for example, 1% or less.

  19 and 20 are explanatory views showing the configuration of the vehicle control apparatus according to the third embodiment of the present invention.

  The state feedback controller has a form u = Kx, and a plurality of states x = [x1, x2,... Xn] are multiplied by corresponding gains K = [k1, k2,. As a result, the output (operation amount) u is described. That is, u = k1 x1 + k2 x2 +... + Kn xn.

  In this case, if the absolute value of the gain product kpi · xi corresponding to a certain state is small compared to all other terms, there is no problem even if the gain is set to 0 because the control effect is not affected. However, doing this for a typical state feedback controller is problematic.

  For simplicity, consider a two-dimensional example. Assume that the two gains are k1 = 500 and k2 = 1, respectively. First, consider the case where the state is x1 = 1 and x2 = 1. In this case, k2 × 2 = 1, which is sufficiently smaller than k1 × 1 = 500, so it seems to be negligible.

  However, considering the case where the state is x1 = 0.01 and x2 = 5, it becomes the same value as k1x1 = 1 and k2x2 = 1, and cannot be ignored. In the case of general state feedback, the same gain is applied to all states, so when considering a combination of states that is opposite to the magnitude of the gain, each term is close and cannot be ignored. Become.

  A hybrid predictive controller based on multiparametric programming is also a kind of state feedback controller, but differs in that a corresponding gain is always selected according to the state, not the same gain. Therefore, the maximum value and the minimum value of the ratio of the absolute value | knxn | of each term of the manipulated variable can be obtained in advance for each region. In particular, when the region does not include the zero axis, the maximum value and the minimum value can be obtained at the end points of the region.

  Therefore, the absolute value of the product of the gain corresponding to the state in each region is always less than or equal to a predetermined value (for example, 1/100) with respect to the absolute value of the product of the gain corresponding to all other states in the region. In this case, the gain is sufficiently small in the region, so it can be safely ignored and is set to zero. In this way, the gain storage area can be reduced, and the amount of calculation can be reduced by eliminating the calculation for the term having a gain of zero.

  Specifically, using the two-dimensional numerical example shown in FIG. 19, there are two states x1 and x2 in the illustrated example, and the end points of the regions are (1, 2), (2, 2), (1 , 3), (2, 3).

  Since this area does not include the zero axis, the absolute value of each term of the manipulated variable at each end point is calculated as (| k1x1 |, | k2x2 |) = (500, 2), (1000, 2), (500, 3), (1000, 3). The ratio is (250, 500, 166.7, 333.3), and the smallest ratio is 166.7. Accordingly, | k2 × 2 | is always sufficiently smaller than | k1 × 1 | in this region, and therefore the gain k2 = 0 is set. What is necessary is just to repeat similarly about another area | region.

  A design method for reducing the gain to zero will be described with reference to FIG.

  First, the absolute value of each term is calculated at each end point of the region, and the ratio is obtained (S10).

  Next, it is determined whether or not all the values (1) to (4) are larger than 1 in the ratio obtained in S10 (S12). If the result is affirmative, the end point having the minimum value is selected (S14).

  On the other hand, when the result in S12 is negative, it is determined whether or not all the values (1) to (4) are smaller than 1 in the ratio obtained in S10 (S16). Is selected (S18). If the result in S16 is negative, the gain cannot be ignored, and the process ends.

  When the end point having the minimum value is selected in S14, it is determined whether the ratio of the points is 100 or more (S20). If the result is affirmative, the gain corresponding to the smaller term can be ignored. The gain is set to 0 (S22).

  Further, when the end point having the maximum value is selected in S18, it is determined whether or not the ratio of the points is 1/100 or less (S24). If the result is affirmative, the gain corresponding to the smaller term is ignored. Considering no problem, the gain is set to 0 (S22). The above processing is repeated in the same manner for other areas.

  As described above, in the third embodiment, a state equation that is modeled and includes the input u and the state x in the control device (shift controller 74) (controlling the driving force) of the vehicle 14 including at least the engine 10. And a control system (gear ratio control system, FIG. 9) that performs feedback control so that at least the engine 10 is to be controlled and the output y matches the input u, and at least n from the current time of the state x The optimal solution group of the input u generated in advance offline is set as the state x so as to optimize the change of the state x according to the evaluation function J (Equation 19 to FIG. 22) for evaluating the change to the step future. Storage means for storing the controller group described by u = Kx, which is the product of the optimum gain K, in the state space area consisting of the state x (the shift controller 74 (ROM of the microcomputer) and an operation amount applying means (several number) for applying the corrected value of the input u obtained by searching the optimum solution group of the input u from the state x to the control target from the control system as an operation amount 28), and the product of the state x and the optimum gain K is a predetermined value, more precisely the absolute value of the product of the optimum gain corresponding to the state in each region. Since the optimum gain K is set to zero when the absolute value of the product of the optimum gains corresponding to all other states is 1/100 or less, the increase in the number of regions and the calculation amount are similarly performed. Can be suppressed as much as possible.

  In other words, since the hybrid predictive controller is given an operation amount by the product sum of a plurality of (optimum) gains and states, if the product of a gain corresponding to a certain state is small, it is assumed that the control effect has no effect and is treated as 0 As a result, it is possible to reduce the storage area for holding the gain and reduce the amount of calculation.

  21 and 22 are explanatory views showing the configuration of the vehicle control apparatus according to the fourth embodiment of the present invention.

  As described above, in the design of hybrid predictive control, a prediction section called horizon is set. Then, in the prediction section obtained by multiplying the sampling time by the horizon, the operation amount is sequentially determined while predicting each state.

  In vehicle control, particularly in a control system for a vehicle power plant, control is generally performed at a control cycle (sampling time) on the order of milliseconds. Therefore, when trying to set a prediction interval of the second order, a horizon of 100 times or more is required. However, when such a long horizon is set, it becomes a very complicated prediction problem, resulting in a controller having an extremely large number of regions. Or even numerical prediction problems cannot be solved.

  Therefore, in the fourth embodiment, the control cycle (sampling time) of the hybrid predictive control is extended so that a prediction section having the same length can be obtained with a small number of horizons. By reducing the number of horizons, the prediction problem can be simplified and the number of regions can be greatly reduced.

  On the other hand, by extending the control cycle (sampling time) of predictive control, the update amount of the manipulated variable is also lowered, and there is a concern that the control performance is lowered.

  Therefore, in the fourth embodiment, instead of a configuration in which the output of the hybrid prediction controller as shown in FIG. 21A is an input (operation amount) that is directly given to the control target, as shown in FIG. The original target value (input) given to the prediction controller is handled as a corrected target value corrected in view of the state and constraint conditions at that time or the evaluation function J described above.

  Then, a lower-layer controller (local controller, second control system) that performs feedback with the operation amount as a target value is arranged, and this controller performs control in the original sampling time.

  This lower layer controller may be simple, for example, a PID controller may be used. By doing so, it is possible to achieve both the reduction in the number of regions and the reduction in the amount of calculation due to the reduction in the prediction frequency of the hybrid prediction control, and the update frequency of the operation amount.

  The above will be described using a simple example shown in FIG.

  The example shown in FIG. 22A is an example in which control is performed with a control cycle (sampling time) = 1 second, horizon = 6, and prediction interval = 6 seconds. In this case, since the prediction and the operation amount are updated every second, the control performance is improved. However, since it is necessary to perform prediction for six steps every second, the number of regions of the hybrid prediction controller increases.

  On the other hand, the example shown in FIG. 22B is an example in which the control period (sampling time) = 3 seconds and horizon = 2, the prediction interval is the same 6 seconds, and the horizon is reduced by increasing the sampling time. is there. In this case, the number of regions can be reduced while maintaining the prediction interval, but since the amount of operation is updated every 3 seconds, there is a concern about a decrease in control performance.

  Therefore, as shown in FIG. 22 (c), similarly to (b), the control period (sampling time) = 3 seconds, horizon = 2, the prediction interval is set to 6 seconds, and the output of the hybrid prediction controller itself is 3 Update every second.

  However, instead of the operation amount, it is treated as a corrected target value for the original target value, the actual trajectory is fed back by the local controller, and the operation amount is updated every second. Thereby, the update frequency of the operation amount can be performed every second while the number of regions of the hybrid prediction controller is reduced with the prediction frequency being seconds.

  In the case of the gear ratio control described in the first embodiment, the target gear ratio after correction is obtained by the gear ratio control system by the hybrid predictive controller with respect to the representative gear ratio obtained from the vehicle speed or the like. It is possible to reduce the prediction frequency and the amount of calculation.

  As described above, in the fourth embodiment, the state equation that is modeled and includes the input u and the state x in the control device (shift controller 74) (controlling the driving force) of the vehicle 14 that includes at least the engine 10. And a first control system (FIG. 9) that performs feedback control every control cycle kT (for example, 3 seconds) so that at least the engine is a control target and the output y matches the input u, Optimum of the input u generated in advance offline so that the change of the state x is optimized according to the evaluation function J (Equation 19 to FIG. 22) that evaluates the change of the state x from the current time to the future n steps. Storage means for storing a controller group in which a solution group is described by u = Kx, which is a product of the state x and the optimum gain K, in a state space region consisting of the state x (shift A microcomputer ROM of the controller 74) and a correction value calculation means for calculating a correction value of the input u obtained by searching the optimum solution group of the input u from the state x (processing performed according to equations 28 to 31) ) And the control system and the control target, and the control cycle kT is multiplied by 1 / k (for example, k: 3) so that the output y matches the correction value of the input u. Since the second control system (FIG. 21B, local controller) that performs feedback control every control cycle T (for example, 1 second) is configured, the increase in the number of regions and the increase in the calculation amount are performed in the same manner. It can be suppressed as much as possible.

  In the above description, the device for controlling the driving force of the vehicle via the gear ratio is described as an example of the vehicle control device. However, the present invention is not limited to this and may be a constant speed traveling control device or the like.

  Moreover, although CVT26 was used as a transmission, a stepped transmission may be used.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic diagram showing a vehicle control apparatus according to a first embodiment of the present invention as a whole by taking a driving force control apparatus as an example. FIG. 2 is a block diagram generally showing modeling of a controlled object, explaining a design of a shift controller and the like of the apparatus shown in FIG. 1. FIG. 3 is a block diagram of the engine controller model shown in FIG. 2. FIG. 3 is a block diagram when the fuel cut of the engine model shown in FIG. 2 is OFF. FIG. 3 is a block diagram when the fuel cut of the engine model shown in FIG. 2 is ON. FIG. 3 is a block diagram of the torque converter model shown in FIG. 2. FIG. 3 is a block diagram of the transmission (CVT) model shown in FIG. 2. FIG. 3 is a block diagram of the vehicle body model shown in FIG. 2. FIG. 3 is a block diagram showing a configuration of a transmission ratio control system (shift controller) that controls the configuration shown in FIG. 2. 10 is an explanatory graph showing characteristics of a fuel efficiency optimum operating line used in the configuration shown in FIG. 9. It is explanatory drawing explaining the model prediction control to which the structure shown in FIG. 9 is applied. It is explanatory drawing which shows the model prediction control by the multiparametric programming method to which the structure shown in FIG. 9 is applied. It is explanatory drawing which shows the state space area | region of the controller of the multiparametric programming method shown in FIG. FIG. 14 is an explanatory diagram of a control law of the controller shown in FIG. 13. It is explanatory drawing which shows the example of allocation of the controller shown in FIG. It is explanatory drawing which shows the example of deletion of the controller shown in FIG. It is explanatory drawing which shows the area | region and the allocated gain which show the structure of the control apparatus of the vehicle which concerns on 2nd Example of this invention. Similarly, it is explanatory drawing which shows the area | region and the allocated gain which show the structure of the control apparatus of the vehicle which concerns on 2nd Example of this invention. It is explanatory drawing which shows the numerical example of a gain which shows the structure of the control apparatus of the vehicle which concerns on 3rd Example of this invention. Similarly, it is a flowchart showing a design method for making the gain zero, showing the configuration of the control device for a vehicle according to the third embodiment of the present invention. It is the same block diagram as FIG. 9 which shows the structure of the control apparatus of the vehicle which concerns on 4th Example of this invention. Similarly, it is explanatory drawing similar to FIG. 11 which shows the structure of the control apparatus of the vehicle which concerns on 4th Example of this invention.

Explanation of symbols

  10 engines (internal combustion engines), 14 vehicles, 16 DBW mechanisms, 24 torque converters, 26 transmissions (CVT), 30 forward / reverse switching devices, 30a forward clutches, 60 engine controllers, 74 shift controllers, W drive wheels (tires)

Claims (4)

In a control device for a vehicle equipped with at least an engine,
a. A control system that is modeled and described by a state equation consisting of an input u and a state x, and that controls at least the engine and performs feedback control so that the output y matches the input u;
b. An optimal solution group of the input u, which has been generated off-line in advance so as to optimize the change of the state x in accordance with an evaluation function that evaluates at least the change of the state x from the current time to n-step future, Storage means for storing a controller group described by u = Kx, which is a product of the optimal gain K, in a state space area consisting of the state x;
c. An operation amount applying means for applying the correction value of the input u obtained by searching the optimal solution group of the input u from the state x as an operation amount to the control object from the control system;
And a controller that describes a state x that cannot physically occur in the state space region is deleted. In a control device for a vehicle equipped with at least an engine,
a. A control system which is modeled and described by a state equation consisting of an input u and a state x, and which controls at least the engine;
b. An optimal solution group of the input u, which has been generated off-line in advance so as to optimize the change of the state x in accordance with an evaluation function that evaluates at least the change of the state x from the current time to n-step future, Storage means for storing a controller group described by u = Kx, which is a product of the optimal gain K, in a state space area consisting of the state x;
c. An operation amount applying means for applying the correction value of the input u obtained by searching the optimal solution group of the input u from the state x as an operation amount to the control object from the control system;
The vehicle control device is characterized in that when the ratio of the optimum gains K1 and K2 between adjacent regions in the state space region is equal to or less than a predetermined value, the two regions are combined into one region. . In a control device for a vehicle equipped with at least an engine,
a. A control system which is modeled and described by a state equation consisting of an input u and a state x, and which controls at least the engine;
b. An optimal solution group of the input u, which has been generated off-line in advance so as to optimize the change of the state x in accordance with an evaluation function that evaluates at least the change of the state x from the current time to n-step future, And storage means for storing a controller group described by u = Kx, which is a product of the optimal gain K, in a state space area consisting of the state x,
c. An operation amount applying means for applying the correction value of the input u obtained by searching the optimal solution group of the input u from the state x as an operation amount to the control object from the control system;
And a vehicle control apparatus that sets the optimum gain K to zero when the product of the state x and the optimum gain K is equal to or less than a predetermined value. In a control device for a vehicle equipped with at least an engine,
a. A first control system that is modeled and described by a state equation consisting of an input u and a state x, and that controls at least the engine and performs feedback control every control cycle kT so that the output y matches the input u; ,
b. An optimal solution group of the input u, which has been generated off-line in advance so as to optimize the change of the state x in accordance with an evaluation function that evaluates at least the change of the state x from the current time to n-step future, And storage means for storing a controller group described by u = Kx, which is a product of the optimal gain K, in a state space area consisting of the state x,
c. Correction value calculating means for calculating a correction value of the input u obtained by searching the optimal solution group of the input u from the state x;
d. Feedback control is performed every control cycle T, which is inserted between the control system and the control target, and the control cycle kT is multiplied by 1 / k so that the output y matches the correction value of the input u. A second control system to
A vehicle control apparatus comprising: