DE102021205386A1 - Method for operating a hydraulic cylinder of a working machine - Google Patents

Abstract

A method for operating a hydraulic cylinder of a, in particular mobile, work machine is proposed, with the following steps: - Providing a target value (ṡR) of a movement parameter of the hydraulic cylinder; - Determining a provisional target value (u∗) of a valve control parameter assigned to the hydraulic cylinder Valve unit as a function of the provided setpoint (ṡR) of the movement parameter using an assignment rule (m, 34);- determining a setpoint (u, uFF) of the valve control parameter of the valve unit using a Gaussian process regression (GPR) using◯ the determined provisional target value (u∗) of the valve control parameter of the valve unit and◯ training data ([x,y]) comprising values of the movement parameter and the valve control parameter; and- outputting a signal depending on the determined desired value (u, uFF) of the valve control parameter in order to operate the hydraulic cylinder.

Description

State of the art

The invention relates to a method and a control unit for operating a hydraulic cylinder and a method and a control unit for operating a working unit of a, in particular mobile, working machine that can be operated by means of a hydraulic cylinder. The present invention also relates to a working machine, a computer program and a computer-readable data carrier.

Modern work machines increasingly enable the execution of automated or semi-automated work processes. The content of the functionality is often the automated following of a desired trajectory for the tool center point (TCP) or, in the case of assistance functions, the support of the driver in following a desired trajectory. For example, data-based rule structures are used here.

Since a pressure difference across an orifice of a valve unit associated with the hydraulic cylinder is highly dependent on a load acting on the working unit, the data generation effort for complete coverage of the possible working range of the working unit with a purely data-based model is very high. In particular, it may be necessary to generate training data for different loads.

Disclosure of Invention

According to a first aspect, the present invention relates to a method for operating a hydraulic cylinder of a, in particular mobile, working machine, in particular by means of a control unit, according to independent claim 1.

According to a second aspect, the present invention relates to a method for operating a working unit that can be operated by means of a hydraulic cylinder, in particular an attachment that can be operated by means of a hydraulic cylinder, of an, in particular mobile, working machine, in particular by means of a control unit, according to independent claim 8.

According to a third aspect, the invention relates to a control unit for operating a hydraulic cylinder of a, in particular mobile, working machine according to independent claim 9.

According to a fourth aspect, the invention relates to a control unit for operating a working unit that can be operated by means of a hydraulic cylinder, in particular an attachment, of a working machine, in particular a mobile one, according to independent claim 10.

According to a fifth aspect, the invention relates to a, in particular mobile, working machine according to independent claim 11

According to a further aspect, the invention relates to a computer program or a computer program product, comprising instructions which, when executed by a computer or a control unit, cause this or these to use the method according to the first aspect of the invention and/or the method according to the second aspect of the Execute and / or control the invention, and a computer-readable data carrier on which the computer program is stored. The computer-readable or machine-readable data carrier can, for example, be a storage medium such as a semiconductor memory, a hard disk memory or an optical memory.

The working machine can be a stationary or preferably a mobile working machine. The working machine can be a working machine for construction, agricultural, forestry or logistical purposes. The mobile work machine can be, for example, an excavator, a wheel loader, a bulldozer, an industrial truck or an aerial work platform. The stationary working machine can, for example, be a hydraulically driven industrial robot.

The working unit of the working machine can be a working unit for processing and/or treating an agricultural and/or forestry and/or construction area and/or for transporting a load. The working unit can be an attachment. It is also conceivable that the working tool includes a working arm, a lifting frame or a lifting frame.

The attachment can be, for example, a bucket, a shovel or a work basket. The attachment can be arranged on a working arm, a lifting frame or a mast of the working machine.

The hydraulic cylinder or the hydraulic cylinder is designed to generate a relative movement between the working unit or the attachment and a machine main unit of the working machine. The machine main unit includes, for example, an operator's cab and/or a drive unit of the working unit. The machine main unit can be, for example, an upper structure of an excavator or a rear structure be a wheel loader. The working unit or the attachment can thus be moved relative to the machine main unit of the working machine by means of the hydraulic cylinder. To this end, the hydraulic cylinder includes a housing and a piston. The piston can be moved relative to the housing by means of pressurization of a hydraulic fluid, preferably a hydraulic fluid, in particular it can be moved in and out of the housing.

The valve unit assigned to the hydraulic cylinder is designed to set a predetermined and/or specifiable flow rate of the hydraulic fluid through the valve unit and/or to apply a predetermined and/or specifiable pressure to the hydraulic fluid. The valve unit assigned to the hydraulic cylinder is designed to generate a relative movement between the piston and the housing of the hydraulic cylinder.

The valve unit can include one or more valves. The valve can be designed as a solenoid valve or a pneumatically actuated valve or pneumatic valve. The valve unit can comprise a pilot valve, in particular an electromagnetic pilot valve, and preferably a main valve which is assigned to the pilot valve and can in particular be actuated pneumatically.

A relative movement between the piston and the housing of the hydraulic cylinder can be controlled by controlling the valve unit based on the output signal. Based on an arrangement of the hydraulic cylinder on the working machine, this relative movement can generate a relative movement between the attachment and the machine main unit of the working machine, in particular between the attachment and a working arm and/or a lifting frame of the working machine.

In the context of the present invention, controlling can be understood as controlling in the sense of generating an output variable based on an input variable. Control can further and preferably be understood as a control comprising regulation in the sense of a continuous determination of an actual value of a variable to be controlled and a continuous comparison of the actual value with a setpoint value of the variable to be controlled.

The control unit preferably comprises one or more, in particular cascaded, controllers. The control unit can be arranged on the working machine. Alternatively, the control unit can be arranged away from the working machine, for example in a server backend or a cloud computing system, and can be connected to the working machine by means of a wireless communication link.

A target position of the working unit can be a spatial relative position of the working unit relative to the working machine comprising the working unit or a spatial position in an external reference coordinate system, for example a global satellite navigation system or a reference coordinate system of a position-detecting sensor unit. The target position of an attachment is preferably a spatial position of the tool center point (TCP) of the attachment. The target position of the work unit or the attachment can be specified, for example, for a work step of a work process to be carried out by means of the work machine. The target position of the attachment can also include a spatial orientation of the attachment, for example relative to the machine main unit of the working machine.

The setpoint value of the movement parameter can be determined as a function of the setpoint position of the working unit using software-based trajectory planning for the working unit or the attachment or for the working machine. The setpoint value of the movement parameter can be determined taking into account at least part of the kinematics of the work machine.

The hydraulic cylinder movement parameter is a parameter of movement of the hydraulic cylinder. The movement of the hydraulic cylinder is preferably a relative movement between the piston and the housing of the hydraulic cylinder. The movement of the hydraulic cylinder can be a uniform or, preferably uniform, accelerated movement.

The movement parameter of the hydraulic cylinder can be a speed and/or an acceleration. The movement parameter is preferably a relative speed and/or a relative acceleration between the piston and the housing of the hydraulic cylinder. The movement parameter can include an amount and a direction of the speed and/or the acceleration.

In the context of the present invention, providing or determining a target value can be understood as providing or determining at least one target value. In this case, the desired value is a predetermined and/or predeterminable value and/or previously determined value. Providing or determining the target value preferably includes providing or determining a plurality or a set of target values. It is conceivable that the plurality or quantity of target values represents a chronological sequence of target values. Providing the target value may include receiving the target value. The target value of the motion parameter is a value or a time accordingly Licher course of the value of the movement parameter according to which a movement of the hydraulic cylinder should take place.

It is conceivable that in the step of providing or determining the setpoint value of the movement parameter, a setpoint value of a movement parameter designed as speed and also a setpoint value of a further movement parameter designed as acceleration of the hydraulic cylinder are provided or determined. Alternatively or additionally, target values for further or higher time derivatives of the speed or acceleration of the hydraulic cylinder can be determined or provided. It is also conceivable that one or more target values of one or more movement parameters of one or more additional hydraulic cylinders are provided or determined, with the one or more additional hydraulic cylinders being assigned to the same valve unit as the hydraulic cylinder or being operated using the same valve unit.

The valve control parameter of the valve unit can be a parameter of one or more valves of the valve unit, based on which the valve or valves of the valve unit are controlled. The valve control parameter can be a valve energization or a current strength of a solenoid valve. The valve control parameter can also be a pressure, by means of which a pneumatically actuable valve is actuated. The valve control parameter can also be a geometric and/or fluidic valve parameter, such as an opening area of a valve orifice of a valve, of the valve unit. It is also conceivable that the valve control parameter is a position of an operating element, in particular a joystick, of the working machine. The operating element can serve as a means for controlling the movement of the hydraulic cylinder. The position of the control of the work machine can be a position or a Z state of the control. The valve unit assigned to the hydraulic cylinder is controlled as a function of the position or the position or the state of the operating element in order to move the hydraulic cylinder. For this purpose, a control signal for controlling the valve unit or a state of the valve unit is assigned to each position of the operating element.

The provisional setpoint value of the valve control parameter is a value determined before the setpoint value of the valve control parameter is determined, on the basis of which the setpoint value of the valve control parameter is determined. The provisional setpoint value of the valve control parameter is therefore an input variable for determining the setpoint value of the valve control parameter. The determined setpoint value of the valve control parameter is preferably different from the determined provisional setpoint value of the valve control parameter.

In the context of the present invention, an assignment rule can be understood to mean a rule that maps the setpoint value of the movement parameter to the provisional setpoint value of the valve control parameter. The assignment specification is in particular a deterministic assignment specification.

The assignment specification advantageously includes a physical model and/or a characteristic diagram and/or a function, in particular a linear function, by means of which the provisional setpoint value of the valve control parameter is assigned to the setpoint value provided for the movement parameter.

The physical model can be a mathematical model or a mathematical algorithm that is designed to map an input variable to an output variable using one or more physical equations or functions. The one or more physical equations represent or approximate a physical law underlying the relationship between the input variable and the output variable.

The, in particular linear, function can include, for example, a straight line through the origin, which represents a proportional relationship between the provisional desired value of the valve control parameter and the provided desired value of the movement parameter.

Advantageously, the setpoint value of the valve control parameter is also determined as a function of, in particular a value of, at least one other parameter of the hydraulic cylinder and/or the work machine. The further parameter is preferably a different parameter from the movement parameter. The further parameter can also correspond to a time derivative of the movement parameter. The additional parameter can be a pressure, for example a pressure of a hydraulic fluid of the hydraulic cylinder. The further parameter is preferably a pressure difference between a pressure on a piston side of the hydraulic cylinder and a pressure on a rod side of the hydraulic cylinder. Alternatively or additionally, the further parameter is a difference between a load pressure of the working unit and a pressure provided by a pump unit for pressurizing the hydraulic fluid. It is also conceivable that the additional parameter is a temperature of a hydraulic fluid in the hydraulic cylinder. It is also conceivable that the white other parameters is a speed of a motor of the working machine.

Within the scope of the present invention, the provisional desired value of the valve control parameter is converted into the desired value of the valve control parameter by means of the Gaussian process regression. Inputs to the Gaussian process regression include the preliminary target value of the valve control parameter and training data.

The training data include at least values of the movement parameter and values of the valve control parameter assigned to the values of the movement parameter. The training data can represent determined or measured values of the movement parameter and/or the valve control parameter, for example. The training data can be determined during operation of the hydraulic cylinder, for example during operation of a working machine that includes the hydraulic cylinder. The training data correspond to combinations of values that occur or are present during operation of the hydraulic cylinder.

Determining the target value of the valve control parameter is done by correcting the provisional value of the valve control parameter based on the training data. The higher a correlation of the provisional value of the valve control parameter with the training data, the more the provisional target value is corrected. Conversely, the lower the correlation of the preliminary value of the valve control parameter with the training data, the more the value of the valve control parameter is based on the provisional value of the valve control parameter.

In other words, by means of the Gaussian process regression, a correlation of the provisional desired values determined for the valve control parameter with the training data can be taken into account. Depending on this correlation, the data-based knowledge in the form of the training data and the prior knowledge on which the determination of the provisional desired value of the valve control parameter is based are weighted. In this way, the extrapolation behavior in Gaussian process regression can be precisely defined using prior knowledge. In this way, it is achieved that in the working areas, which are sufficiently represented in the training data, essentially the data-based knowledge is used and thus a high estimation accuracy is achieved, while in the working areas with incomplete training data, prior knowledge continues to dominate the determination of the setpoint of the valve control parameter influenced.

The Gaussian process regression and the assignment rule are preferably part of a pilot control of a controller used to determine the setpoint value of the valve control parameter.

The control unit is preferably arranged on the, in particular mobile, working machine. It is also conceivable that the control unit is arranged away from the working machine and can be connected or is connected to the working machine, preferably by means of a wireless communication connection.

The methods according to the invention and the corresponding control units increase the safety and reliability in the automated operation of a hydraulic cylinder or a working unit of a working machine, in particular a mobile working machine, that can be moved by means of the hydraulic cylinder. By combining an assignment rule and a Gaussian process regression, a work process can also be (partially) automated for a work machine that has complex hydraulic behavior with, in particular, load-dependent pressures or pressure differences and only with particularly great effort through a sufficiently detailed physical model or an extremely laboriously trained data-based model can be fully described.

In particular, the robustness of the method can be increased in work areas that occur less frequently or work areas for which fewer training data are provided. In particular for a hydraulic working machine, for which a possible pressure range in the hydraulic system depends on an interaction between load forces and valve control parameters, untrained input signal combinations can occur, for which the present method enables operation with increased safety. At the same time, simplifications and/or errors in the assignment specification, in particular in the physical model, can be corrected in work areas sufficiently represented by the training data.

It is advantageous if, in the step of determining the setpoint value of the valve control parameter, the training data are determined as a subset of a provided training data record as a function of the provisional setpoint value determined for the valve control parameter. In other words, the size of the training data set provided is reduced by reducing the number of value pairs, value tuples or points contained. This refinement can reduce the need for computing resources and/or a runtime for determining the setpoint value of the valve control parameter.

It is advantageous here if the subset from those training data of Trai ningsdatensatzes exists for which a value of the movement parameter satisfies a nearest-neighbor criterion with regard to the determined provisional desired value of the valve control parameter. The nearest neighbor criterion specifies an arrangement for values or points of the training data, which are taken into account for determining the desired value of the valve control parameter. It is conceivable that with regard to one or more dimensions of the training data, one, two or more nearest neighboring values or

Neighboring points are part of the subset relative to the provisional target value determined. A number of the neighboring values or neighboring points to be taken into account when determining the desired value of the valve control parameter can be different for different dimensions of the training data. For example, the training data may include points each having a velocity value and an acceleration value associated with a value of the valve control parameter. It is conceivable here that a different number of nearest neighbors is taken into account with regard to the determined desired value for the speed and a determined desired value for the acceleration.

Alternatively, it is advantageous if the subset consists of those training data of the training data record for which a value of the movement parameter satisfies a distance criterion with regard to the provisional setpoint value of the movement parameter that was determined. The distance criterion specifies a maximum value for a distance for values or points of the training data, which are taken into account for determining the setpoint value of the valve control parameter. The maximum value for the distance can be different for different dimensions of the training data.

A determination and in particular an inversion of covariance matrices or covariance vectors takes place based on the determined subset of the training data set. For this purpose, the method can include a step of determining the neighboring values or neighboring points relative to the determined provisional setpoint value of the valve control parameter and in particular a setpoint value of a further movement parameter, e.g. an acceleration, assigned to the determined provisional setpoint value of the valve control parameter.

With this configuration, the computing effort and the computing time for operating the hydraulic cylinder can be significantly reduced. At the same time, in comparison to a Gaussian process regression with k-means clustering known to those skilled in the art, the modeling quality or the modeling accuracy is increased and prediction errors are reduced. As a result, the method can be executed online on a control unit, i.e. during the runtime of the working machine, and in particular the training data record can also be adapted during the runtime.

It is also advantageous if the training data set is determined by mapping the values of the movement parameter of an overall training data set comprising values of the movement parameter and the valve control parameter to grid values of a predetermined grid of values of the movement parameter. In other words, this means that the overall training data set is converted into the training data set by discretization to a predetermined grid. The grid can have a spatially constant resolution. It is also conceivable that the grid has different spatial resolutions along different dimensions of the training data. It is also conceivable that the grid has a spatially varying resolution, i.e. areas with different resolutions, along one or more dimensions of the training data. The step of mapping can take place during the runtime of the method, in particular the operation of the hydraulic cylinder, for example in the step of determining the setpoint value of the valve control parameter, or offline, i.e. before the operation of the hydraulic cylinder. By discretizing using the specified grid, outliers in the training data can be excluded and a physically or work machine-side permissible value space can be defined.

It is advantageous here if, when mapping the values of the movement parameter of the overall training data set, the grid values are replaced by a mean value of the values of the movement parameter mapped to the respective grid value. In this case, the values of the valve control parameter assigned to the respective movement parameter can also be replaced by an average value of the values of the valve control parameters. Discretization errors when determining the desired value of the valve control parameter can be reduced by this configuration.

It is also advantageous if, in the step of determining the setpoint value of the valve control parameter, a variance of the determined setpoint value of the valve control parameter is also determined and the signal is also output as a function of the determined variance. It is conceivable that, depending on the determined variance, a speed of the hydraulic cylinder, in particular a relative speed between a piston and a housing of the hydraulic cylinder, is limited and/or reduced by means of the control signal. For this purpose, if a threshold value for the determined variance is exceeded, a cor correction request to a module, e.g reduce.

It is alternatively or additionally conceivable that, when a threshold value for the ascertained variance is exceeded, a determination of new training data is requested in order to promote targeted active learning in poorly trained working areas of the cylinder or the working machine. It is also conceivable that when a threshold value for the determined variance is exceeded, the output control signal controls an acoustic and/or optical and/or haptic output unit in order to draw the attention of an operator or supervisor of the work machine to a less trained work area. With this configuration, a certainty or accuracy of the determined setpoint value of the valve control parameter can be taken into account when the control signal is output and in particular when the hydraulic cylinder is operated. As a result, the hydraulic cylinder and in particular the work machine can also be operated safely in work areas with little training data and low quality of the assignment rule.

Furthermore, it is advantageous if the setpoint value of the valve control parameter is determined using a speed controller, with the speed controller including a pilot control based on the assignment rule and the Gaussian process regression. With this configuration, the method can provide a speed control of a hydraulic cylinder of a work machine, which is characterized by low training effort and at the same time high accuracy even in untrained or little trained work areas.

character list

• 1 eine schematische Darstellung einer Steuereinheit zum Betreiben einer Arbeitseinheit einer Arbeitsmaschine;
• 2 eine schematische Darstellung einer Vorsteuerung für die Steuereinheit gemäß 1;
• 3A,B eine schematische Darstellung eines Gesamttrainingsdatensatzes für die Vorsteuerung gemäß 2;
• 4A,B eine schematische Darstellung eines Diskretisierens des Gesamttrainingsdatensatzes gemäß 3;
• 5 eine schematische Darstellung eines Reduzierens des Gesamttrainingsdatensatzes gemäß 3; und
• 6A,B je ein Ablaufdiagramm eines Verfahrens zum Betreiben eines hydraulischen Zylinders und eines Verfahrens zum Betreiben einer Arbeitseinheit einer Arbeitsmaschine.

The invention is explained in more detail below by way of example with reference to the attached drawings. Show it:

• 1 a schematic representation of a control unit for operating a working unit of a working machine;
• 2 a schematic representation of a pre-control for the control unit according to FIG 1 ;
• 3A,B a schematic representation of an overall training data set for the pre-control according to FIG 2 ;
• 4A,B a schematic representation of a discretization of the overall training data set according to FIG 3 ;
• 5 a schematic representation of a reduction of the total training data set according to FIG 3 ; and
• 6A,B a flowchart of a method for operating a hydraulic cylinder and a method for operating a working unit of a working machine.

1 shows a schematic representation of a control unit 10 for operating a working unit, which can be operated by means of a hydraulic cylinder 26, of a working machine, in particular a mobile working machine. The working machine includes the working unit, the hydraulic cylinder 26, a valve unit 24 assigned to the hydraulic cylinder 26 and the control unit 10. The working unit is operated by means of the hydraulic cylinder, in particular moved relative to a machine main unit of the working machine.

The working machine can be an excavator, for example, which includes a working unit designed as a working arm. The working arm includes a boom, an arm, and a bucket, and a boom cylinder, an arm cylinder, and a bucket cylinder. Associated with each of the cylinders is a respective valve unit for moving the cylinder.

The control unit 10 is set up to receive a target position r _R of the working unit. Here, the target position r _R can also include an indication of a spatial orientation of the work unit. For example, the control unit 10 can be set up to receive target spatial coordinates of a tool center point (TCP) of the bucket of the excavator and an angle of the bucket relative to a predefined reference direction.

The target position r _R can be part of a target trajectory of the working unit. It is conceivable that the setpoint trajectory is determined by an operator of the working machine and/or by means of a computing unit 12 for generating a setpoint trajectory for the working machine and is made available to the control unit 10 .

The target position r _R of the working unit is preferably made available to a pose module 14 of the control unit 10 . The pose module 14 is preferably designed as part of a cascade controller and set up to receive the target position r _R of the working unit and also the actual position r of the working unit from a sensor unit 28 detecting the actual position r of the working unit. Next, the pose module 14 is set up based on a Deviation from target position r _R and actual position r of the working unit to determine a target speed ṙ _R of the working unit, by means of which the actual position r can be converted into the target position r _R . For example, the target position r _R of the TCP and the actual position r of the TCP are received by means of the pose module 14 and a target speed ṙ _R of the TCP is calculated on the basis of the deviation. The pose module 14 can also be set up to determine a target acceleration _r̈ R of the working unit based on the deviation between the target position r _R and the actual position r of the working unit or the determined target speed ṙ _R .

The control unit 10 includes an inverse kinematics module 16 which is set up to determine a setpoint value ṡ _R of a movement parameter of the hydraulic cylinder 26 based on the setpoint speed ṙ _R of the working unit. The setpoint value ṡ _R of the movement parameter is preferably determined taking into account the kinematics of the working unit of the working machine, in particular by solving the inverse kinematic problem for the working machine. For example, based on the target speed ṙ _R of the TCP of the excavator, a target speed ṡ _R is calculated for each of the boom cylinder, the arm cylinder, and the bucket cylinder. For example, the target value ṡ _R of the movement parameter of the hydraulic cylinder 26 can be a target speed ṡ _R of the hydraulic cylinder 26 . It is conceivable that the inverse kinematics module 16 is also set up, alternatively or additionally, to determine a target acceleration s̈ _R of the hydraulic cylinder 26 as a target value of a particular further or additional movement parameter.

The control unit 10 is also set up to use a speed module 18 to determine a target value u of a valve control parameter of the hydraulic cylinder 26 as a function of the target value ṡ _R of the movement parameter.

The speed module 18 is preferably designed as a speed controller 18 and set up to receive the target value ṡ _R of the movement parameter and the actual value s _R of the movement parameter of the hydraulic cylinder 26, preferably also the target value s̈ _R of the further movement parameter. The speed module 18 is also set up to determine the setpoint u of the valve control parameter based on a deviation between the setpoint ṡ _R and the actual value ṡ of the movement parameter.

• - Istwert s̈ und/oder Sollwert s̈_R einer Beschleunigung des hydraulischen Zylinders 26 und/oder eines oder mehrerer weiterer derselben Ventileinheit 24 zugeordneten hydraulischen Zylinder,
• - Istwert eines Drucks an einer Zylinderstange des hydraulischen Zylinders 26,
• - Istwert eines Drucks an einem Zylinderkopf des hydraulischen Zylinders 26,
• - Differenz Δp_cyl zwischen Istwert des Drucks an der Zylinderstange und Istwert des Drucks am Zylinderkopf 26,
• - Istwert eines Lastdrucks der Arbeitseinheit,
• - Istwert eines Pumpendrucks einer dem hydraulischen Zylinder 26 zugeordneten Pumpeneinheit zur Druckbeaufschlagung einer den hydraulischen Zylinder 26 bewegenden Hydraulikflüssigkeit,
• - Differenz Δp_sys zwischen Istwert des Lastdrucks und Istwert des Pumpendrucks,
• - Istwert einer Temperatur der Umgebung und/oder eines Motorenöls in einem Ölkreislauf der Arbeitsmaschine.

The speed module 18 is preferably set up to also take into account one or more of the following influencing variables when determining the setpoint u of the valve control parameter:

• - Actual value s̈ and/or setpoint value s̈ _R of an acceleration of the hydraulic cylinder 26 and/or one or more other hydraulic cylinders assigned to the same valve unit 24,
• - Actual value of a pressure on a cylinder rod of the hydraulic cylinder 26,
• - Actual value of a pressure on a cylinder head of the hydraulic cylinder 26,
• - Difference Δp _cyl between the actual value of the pressure on the cylinder rod and the actual value of the pressure on the cylinder head 26,
• - Actual value of a load pressure of the working unit,
• - Actual value of a pump pressure of a pump unit assigned to the hydraulic cylinder 26 for pressurizing a hydraulic fluid moving the hydraulic cylinder 26,
• - Difference Δp _sys between the actual value of the load pressure and the actual value of the pump pressure,
• - Actual value of a temperature of the environment and/or an engine oil in an oil circuit of the working machine.

The speed module 18 preferably includes a PI controller with a pre-control 18a (feedforward control). As a result, even with complex and non-linear behavior, the PI controller can be relieved to the extent that its feedback 18b (feedback control) only has to take into account disturbances and model errors.

The pilot control 18a represents an inverse behavior of the working unit relating to the relationship between movement parameters of the hydraulic cylinder 26 and valve control parameters of the respective valve unit 24. In other words, the pilot control 18a determines a value u of the movement parameter of the hydraulic cylinder 26 based on a predetermined value ṡ _R of the valve control parameter. In this case, the pilot control 18a takes place using an allocation rule, in particular a physical model, and a Gaussian process regression.

For the excavator with boom cylinder, stick cylinder and shovel cylinder, the control unit 10 includes a speed module 18 designed as a PI controller for determining a setpoint value u of the valve control parameter for the valve unit assigned to the respective cylinder. Each of the PI controllers has its own pilot control 18a.

The desired value u of the valve control parameter is output as the manipulated variable of the speed module 18 . Here, the setpoint u of the valve control parameter is in particular a sum of an output variable u _FB of the feedback 18b and an output variable u _FF of the pilot control 18a.

The model-based pre-control 18a is based on the assignment rule m, in particular the physical model, and the Gaussian process regression GPR as exemplified in 2 shown. In other words, the setpoint value u to be determined for the valve control parameter is determined based on the received setpoint value ṡ _R of the speed parameter of the hydraulic cylinder 26 using the assignment rule m, in particular the physical model and the Gaussian process regression GPR.

The control unit 10 is also set up to output the determined setpoint value u of the valve control parameter to the system 20 . The system 20 includes a control module 22, the valve unit 24 and the hydraulic cylinder 26. The system 20 represents a controlled system of the speed controller 18.

The control module 22 is set up to operate the valve unit 24 based on the signal output by the control unit 10 . The valve unit 24 is designed to adjust a flow of a hydraulic fluid through the valve unit 24 in response to the signal output to the control module 22 in order to move the hydraulic cylinder 26 . In other words, the hydraulic cylinder 26 is operated and the working unit is thus moved by outputting the signal to the valve unit 24 as a function of the determined desired value u of the valve control parameter.

One or more of the sensors 28 arranged on the working machine are designed to measure the actual position r of the working unit, an actual position or actual position s of the hydraulic cylinder 26, and preferably the influencing variables listed above, in particular ṡ s̈, Δp _cyl , Δp _sys . In this case, the sensors 28 are preferably assigned suitable model-based filters 30 .

The model-based filters 30 are set up to output the actual position r of the working unit to the pose module 14 and to the inverse kinematics module 16, to output the actual position or actual setting s of the hydraulic cylinder 26 to the inverse kinematics module 16, and preferably to output the influencing variables listed above, in particular ṡ s̈, Δp _cyl , Δp _sys , to the speed module 18 .

2 shows a representation of an embodiment of the pilot control 18a for the in 1 described control unit 10. The pilot control 18a includes the assignment rule m and the Gaussian process regression GPR.

According to this exemplary embodiment, the assignment rule m includes a physical model 34 which is set up to assign a provisional setpoint u _∗ of the valve control parameter to the setpoint ṡ _R of the movement parameter of the hydraulic cylinder. For this purpose, the physical model 34 includes a first module 36, which is set up based on the setpoint ṡ _R of the movement parameter, taking into account or neglecting a pressure build-up dynamics within the hydraulic cylinder, a setpoint Q _vlv,R of a volume flow of a hydraulic fluid through the hydraulic cylinder associated valve unit 24 to determine. For example, a target value Q _Hd,R of a volume flow rate of the hydraulic fluid at the head of the hydraulic cylinder 26 and a target value Q _Rd,R of a volume flow rate of the hydraulic fluid at the cylinder rod of the hydraulic cylinder 26 can be calculated.

If the pressure build-up dynamics of the hydraulic cylinder 24 are neglected, a steady-state relationship between the cylinder speed ṡ _R and the volume flow Q _vlv,R results, which depends on a piston area of the hydraulic cylinder 26 .

The physical model 34 includes a second module 38 that is set up to determine a setpoint value A _vlv,R of an opening area of an orifice of the valve unit 24 based on the determined setpoint value Q _vlv, R of the volume flow.

The physical model 34 is preferably set up so that, when determining the target value A _vlv,R of the opening area of the orifice, a pressure difference Δp _cyl between the actual value of the pressure on the cylinder rod and the actual value of the pressure on the cylinder head and/or a pressure difference Δp _sys between the actual value of the load pressure and the actual value of the pump pressure must be taken into account. Alternatively, the physical model 34 can be set up to take into account a specified and/or specifiable value for the pressure difference Δp _cyl and/or pressure difference Δp _sys when determining the target value Δ _vlv,R of the opening area of the orifice, e.g. if there is no pressure sensor is available.

The physical model 34 also includes a third module 40, which is set up to determine the provisional setpoint value u _* of the valve control parameter based on the determined setpoint value A _vlv,R of the opening area of the orifice. The third module 40 of the physical model 34 can be set up to take valve dynamics of the valve unit 24 into account when determining the provisional desired value u _* of the valve control parameter.

The Gaussian process regression GPR is set up to convert the provisional desired value u _∗ of the valve control parameter provided by the physical model 34 into the desired value u _FF of the valve control parameter. For this purpose, training data [X,y] are provided, which include values X of the movement parameter and associated values y of the valve control parameter.

In this case, a covariance matrix K _y is determined with regard to the values X of the training data [X,y] provided. Furthermore, a covariance matrix k _* is determined between the values X of the provided training data [X,y] and the target value ṡ _R of the movement parameter, preferably also the target value s̈ _R of the further movement parameter. Furthermore, a variance k _∗∗ of the target value ṡ _R of the movement parameter is determined. Various kernel functions are known to those skilled in the art for determining the covariance matrices. A squared exponential kernel function is preferably used.

Hierbei bezeichnet $K_y^{-1}$

die Inverse Matrix der Kovarianzmatrix K_y. Der Ausdruck m(X) bezeichnet eine Ausgabe der Zuordnungsvorschrift m bei Anwendung auf die Eingabegröße X.The setpoint u _FF of the valve control parameter is determined according to: $${\mathrm{and}}_{ff}={\mathrm{and}}_{*}+k_{*}^T{K}_y^{-1}\left(y-m\left(X\right)\right)$$

referred to here $K_y^{-1}$

the inverse matrix of the covariance matrix K _y . The expression m(X) denotes an output of the assignment rule m when applied to the input quantity X.

In addition, a variance of the setpoint u _FF of the valve control parameter can be determined according to: $$va\mathrm{right}\left({\mathrm{and}}_{ff}\right)=k_{**}-k_{*}^T{K}_y^{-1}{k}_{*}$$

When determining the target value u _FF of the valve control parameter, the in 3 and 4 described method steps for discretizing an overall training data set and/or preferably the in 5 method steps described for reducing the total training data set are used.

The manipulated variable of the speed controller 18, namely the setpoint u of the valve control parameter, can be adjusted based on the output variable u _FF of the model-based pilot control 18a by means of the feedback u _FB . Preferably, u=u _FF +u _FB applies. This means that without taking into account the feedback 18b, the target values u _FF and u are identical.

3 shows a representation of an overall training data set [X, y] _G for the pre-control 18a according to FIG 2 . The overall training data record [X,y] _G comprises a multiplicity of combinations of a value ṡ of the movement parameter, a value s̈ of a further movement parameter and a value u of the valve control parameter assigned to these values s, s of the movement parameters. The motion parameter represents a speed of the TCP. The further movement parameter represents an acceleration of the TCP. The total training data set [X, y] therefore includes _G points (ṡ,s̈), each of which is assigned a value u of the valve control parameter.

3A and 3B each show the same total training data set [X,y] _G . A horizontal axis of 3A , 3B represents the speed of TCP. A vertical axis of 3A , 3B represents the acceleration of the TCP. If the overall training data set [X,y] _G contains a value u of the valve control parameter for a combination of a speed value ṡ and an acceleration value s̈, this is independent of its height as a point in 3A and 3B shown.

Opposite to 3A is in 3B additionally shown a grid with grid points, represented by circled crosses. A grid point comprises a first grid value with respect to the movement parameter, namely the speed of the TCP, and a second grid value with respect to the further movement parameter, namely the acceleration of the TCP.

In this exemplary embodiment, the grid has the same spatially constant resolution with regard to the movement parameter and the further movement parameter. In other words, this means that a distance between adjacent raster points along the vertical or the horizontal axis is the same for all raster points.

Based on the overall training data set [X,y] _G , a training data set [X,y] is determined by mapping or discretizing points es, s) of the overall training data set [X,y] _G onto the grid points. Here, as in 4 shown, the speed value ṡ and the acceleration value s̈ are each mapped to or assigned to a grid value of a grid point.

4 shows the mapping of points of the overall training data set [X,y] _G to grid points as an example for the discretization with respect to the value ṡ of the motion parameter. Here shows 4A Points es, s) of the overall training data set [X,y] _G with the respectively associated value u of the valve control parameter in a representation in the (s, u) plane.

The nearest grid point is determined for each point (ṡ, s̈) of the overall training data set [X, y] _G. Furthermore, for each point (ṡ, s̈) of the overall training data set [X,y] _G , a distance to the nearest grid point is determined. If the distance to the nearest grid point is less than or equal to a predetermined maximum value, the point (ṡ, s̈) of the total training data set [X,y] _G is assigned to the nearest grid point. If the distance to the nearest grid point is greater than a predetermined maximum value, the point (ṡ, s̈) of the total training data set [X,y] _G is discarded or not taken into account. The distance can be determined using any suitable mathematical norm.

4B shows a replacement of the raster points by an average of the points (ṡ, s̈) of the motion parameters mapped to the respective raster points. That is, in other words, a point (ṡ, s̈) of the training data set [X,y] results as the mean of points of the total training data set [X,y] _G . The value u of the valve control parameter assigned to a point (ṡ, s̈) of the training data set [X,y] is determined as the mean value of the values u of the valve control parameter of the points (ṡ, s̈) mapped onto the grid point.

5 FIG. 12 shows an illustration of reducing the total training data set [X,y] according to _{G. FIG} 3 for the pilot control 18a according to 2 . Here, in addition to the grid represented by circled crosses, according to 3B a point (s, s) represented by a white cross is represented according to a determined desired speed ṡ and a determined desired acceleration s̈.

A subset [X,y] _U of training data of the training data set [X,y] is determined for this point (ṡ, s̈). The subset consists of the four points of the grid that are neighboring with respect to ṡ and s̈. These four nearest neighboring points or nearest neighboring points are in 5 shown as circled crosses with a white border. In other words, the training data includes training data in the form of a subset [X,y] _U of the training data set [X,y] .

The Gaussian process regression is set up, the target value u using the training data [X,y] _U , ie using a reduced training data set [X,y] _U in the form of the determined subset [X,y] _U consisting of the four nearest neighboring points. As a result, the computing effort for determining, and in particular for inverting, the covariance matrices K _y , f _* can be significantly reduced.

6 shows in 6A a flow chart of a method for operating a hydraulic cylinder of a, in particular mobile, working machine. The method is provided with the reference number 100 in its entirety.

The method 100 includes a step of providing 110 a target value of a movement parameter of the hydraulic cylinder.

The method 100 further includes a step of determining 120 a provisional target value of a valve control parameter of a valve unit assigned to the hydraulic cylinder as a function of the provided target value of the movement parameter using an assignment rule.

The method also includes a step of determining 130 a target value of the valve control parameter of the valve unit by means of a Gaussian process regression using the determined provisional target value of the valve control parameter of the valve unit and training data comprising values of the movement parameter and the valve control parameter. In this case, the values of the valve control parameter are in particular each assigned to a value of the movement parameter.

The method also includes a step of outputting 140 a signal as a function of the determined setpoint value of the valve control parameter in order to operate the hydraulic cylinder, in particular based on the signal output

6 shows in 6B a flowchart of a method for operating a work unit that can be operated by means of a hydraulic cylinder, in particular an attachment, of a work machine, in particular a mobile work machine. The method is provided with the reference number 200 in its entirety.

The method 200 includes a step of providing 210 a target position of the working unit, in particular the attachment. The method further includes a step of determining a target value of a movement parameter of the hydraulic cylinder as a function of the provided target position of the working unit.

The method 200 further includes a step of determining 220 a provisional setpoint value of a valve control parameter of a valve unit assigned to the hydraulic cylinder as a function of the determined setpoint value of the movement parameter using an assignment rule.

The method 200 also includes a step of determining 230 a target value of the valve control parameter of the valve unit by means of a Gaussian process regression using the determined provisional target value of the valve control parameter of the valve unit and training data comprising values of the movement parameter and the valve control parameter.

Method 200 also includes a step of outputting 240 a signal as a function of the determined setpoint value of the valve control parameter, in particular directly or indirectly to the valve unit, in order to operate the working unit, in particular the attachment, of the, in particular mobile, working machine by operating the hydraulic cylinder .

Claims (13)

Method (100) for operating a hydraulic cylinder (26) of a, in particular mobile, working machine, with the following steps: - providing (110) a setpoint value (ṡ _R , s̈ _R ) of a movement parameter of the hydraulic cylinder (26); - Determining (120) a provisional target value (u _∗ ) of a valve control parameter of a valve unit (24) assigned to the hydraulic cylinder (26) as a function of the provided target value (ṡ _R , s̈ _R ) of the movement parameter using an assignment rule (m, 34) ; - Determining (130) a target value (u, u _FF ) of the valve control parameter of the valve unit (24) by means of a Gaussian process regression (GPR) using ◯ the determined provisional target value (u _∗ ) of the valve control parameter of the valve unit (24) and ◯ training data ([X,y], [X,y] _G , [X,y] _U ) comprising values of the motion parameter and the valve control parameter; and - outputting (140) a signal depending on the determined desired value (u, u _FF ) of the valve control parameter in order to operate the hydraulic cylinder (26). Method (100) according to claim 1 , characterized in that the assignment rule (m, 34) comprises a physical model (34) and/or a characteristic diagram and/or a function, in particular a linear function, by which or which the provided setpoint value (ṡ _R , s̈ _R ) of the Movement parameter of the provisional setpoint (u, u _FF ) is assigned to the valve control parameter. Method (100) according to claim 1 or 2 , characterized in that in the step of determining the target value (u, u _FF ) of the valve control parameter, the training data ([X, y] _U ) as a _subset ([X, y ] _U ) of a provided training data set ([X,y], [X,y] _G ) can be determined. Method (100) according to claim 3 , characterized in that the subset ([X,y] _U ) consists of those training data of the training data set ([X, y], [X, y] _G ) for which a value of the movement parameter is a nearest-neighbor criterion with regard to the determined provisional target value (u _∗ ) of the valve control parameter is met. Method (100) according to claim 3 or 4 , characterized in that the training data set ([X,y]) is determined by mapping the values of the movement parameter of an overall training data set ([X,y] _G ) comprising values of the movement parameter and of the valve control parameter to grid values of a predetermined grid of values of the movement parameter. Method (100) according to claim 5 , characterized in that when mapping the values of the movement parameter of the overall training data set ([X, y] _G ), the grid values are replaced by an average of the values of the movement parameter mapped to the respective grid value. Method (100) according to one of the preceding claims, characterized in that in the step of determining the setpoint (u, u _FF ) of the valve control parameter, a variance of the determined setpoint (u, u _FF ) of the valve control parameter is also determined and the signal is also dependent is output from the determined variance. Method (200) for operating a working unit that can be operated by means of a hydraulic cylinder (26), in particular an attachment, of an in particular mobile working machine, with the following steps: - providing (210) a target position (r _R ) of the working unit, in particular of the attachment; - Determining (220) a target value (ṡ _R , s̈ _R ) of a movement parameter of the hydraulic cylinder (26) as a function of the provided target position (r _R ) of the working unit; - Determining (230) a provisional target value (u _∗ ) of a valve control parameter of a valve unit (24) assigned to the hydraulic cylinder (26) as a function of the determined target value (ṡ _R , s̈ _R ) of the movement parameter using an assignment rule (m, 34) ; - Determining (240) a target value (u, u _FF ) of the valve control parameter of the valve unit (24) by means of a Gaussian process regression (GPR) using ◯ the determined provisional target value (u _∗ ) of the valve control parameter of the valve unit (24) and ◯ training data (([X,y], [X,y] _G , [X,y] _U ) comprising values of the movement parameter and the valve control parameter; and - Outputting (250) a signal depending on the determined setpoint (u, u _FF ) of the valve control parameter in order to operate the working unit, in particular the attachment, of the, in particular mobile, working machine by operating the hydraulic cylinder (26). Control unit (10) for operating a hydraulic cylinder (26) of a, in particular mobile, working machine, the control unit being set up - to receive a target value (ṡ _R , s̈ _R ) of a movement parameter of the hydraulic cylinder (26), - a preliminary target value (u _∗ ) of a valve control parameter of a valve unit (24) assigned to the hydraulic cylinder (26) as a function of the received target value (ṡ _R , s̈ _R ) of the movement parameter using an assignment rule (m, 34), - a target value (u , u _FF ) of the valve control parameter of the valve unit (24) by means of a Gaussian process regression (GPR) using ◯ the determined provisional desired value (u _∗ ) of the valve control parameter of the valve unit (24) and ◯ training data ([X,y], [ X,y] _G , [X,y] _U ) comprising values of the movement parameter and the valve control parameter, and - a signal depending on the determined desired value (u, u _FF ) of the valve spend expensive parameters to operate the hydraulic cylinder (26). Control unit (10) for operating a working unit that can be operated by means of a hydraulic cylinder (26), in particular an attachment, of a working machine, in particular a mobile one, the control unit (10) being set up, - a target position (r _R ) of the working unit, in particular of the attachment , to receive, - to determine a target value (ṡ _R , s̈ _R ) of a movement parameter of the hydraulic cylinder (26) as a function of the received target position (r _R ) of the working unit, - to determine a preliminary target value (u _∗ ) of a valve control parameter of one of the hydraulic to determine the valve unit (24) assigned to the cylinder (26) as a function of the determined desired value (ṡ _R , s̈ _R ) of the movement parameter using an assignment rule (m, 34); - a target value (u, u _FF ) of the valve control parameter of the valve unit (24) by means of a Gaussian process regression (GPR) using ◯ the determined provisional target value (u _∗ ) of the valve control parameter of the valve unit (24) and ◯ training data ([X ,y], [X,y] _G , [X,y] _U ) including determining values of the motion parameter and the valve control parameter, - outputting a signal as a function of the determined desired value (u, u _FF ) of the valve control parameter, in order to be operated of the hydraulic cylinder (26) to operate the working unit, in particular the attachment, of the, in particular mobile, working machine. Working machine, in particular mobile working machine, with a working unit, in particular an attachment, at least one hydraulic cylinder (26) for moving the working unit, in particular the attachment, and a control unit (10) for operating the working unit, in particular the attachment claim 10 . Computer program, comprising instructions which, when executed by a computer or a control unit (10), cause this or these to carry out the method (100) according to one of steps 1 to 7 or the method (200). claim 8 execute and/or control. Computer-readable data carrier on which the computer program claim 12 is saved.

Images