Pre-control system for a vehicle's powertrain

Abstract

Pre-control system (1) for a powertrain of a vehicle comprising a control unit (2) with a communication device (3) for receiving data from an accelerator pedal (4) of the vehicle, wherein the control unit (2) is configured to, based on the data of the accelerator pedal (4), namely an input signal (u k ), to calculate a trajectory for the feedforward control system (1), wherein the trajectory consists at least partially of a sequence of parametrically modeled curves, namely Bezier curves (b k (τ)), is composed.

Description

Field of invention

[0001] The present invention relates to a pre-control system for a vehicle drive train comprising a control unit with a communication device for receiving data from an accelerator pedal of the vehicle, wherein the control unit is configured to calculate a trajectory for the pre-control system based on the data of the accelerator pedal, namely an input signal, and a vehicle with such a pre-control system. State of the art

[0002] Trajectory planning is a fundamental subtask in many technical fields. For example, it represents the interface between the desired robot movements and their actual execution in industrial robots. In automotive engineering, trajectory planning is used, for instance, in the automation of vehicles or individual vehicle components, as certain technical functions require a type of path calculation along which, for example, a body or a target value can move.

[0003] For example, when considering the calculation of the torque of an electric motor in an electric powertrain using inverter software, the desired torque is currently translated directly to the electric motor via the accelerator pedal. Due to elasticities in the powertrain and the motor connection, this results in an oscillating torque at the vehicle's drive wheels. These oscillations can be detected using speed sensors on the electric motor and the drive wheels, and a counter-torque can be applied to the electric motor. This functionality is known as "Active Damping (AD)." This vibration compensation is necessary because otherwise, it would lead to reduced driver comfort and durability problems in the powertrain.

[0004] With "Active Damping", an initial vibration of the drivetrain is accepted and the "Active Damping" reacts to this initial vibration.

[0005] However, it would be desirable to apply the torque in such a way that oscillation doesn't occur in the first place. The "Active Damping" algorithm attempts to compensate for any relative speed between the electric motor and the drive wheel. However, a twisting of the drivetrain is necessary for torque to be transmitted in the first place. Here, too, the "Active Damping" algorithm counteracts this and prevents faster torque rise times.

[0006] The motor torque can be calculated, for example, from the desired axle torque using an inverse model of the drivetrain. One way to create this inverse model is through the theory of differential flatness. This inverse drivetrain model already incorporates the compliances of the drivetrain, and the torque requirement for the electric motor is such that these compliances are dynamically preloaded.

[0007] For the implementation of such a flatness-based feedforward control of the electric machine, trajectory planning of the flat output is required. This means that the flat output, for example, the angle of rotation between the electric machine and a drive wheel in a conventional electric axle, must follow a specific profile and be continuous over time. Trajectory planning describes the design of this profile and the determination of an interpolating continuous-time function from discrete-time input values. This generally involves a conflict of objectives: On the one hand, the smoothest possible profiles are desirable to avoid dynamic limitations of the feedforward function; on the other hand, the trajectories should be generated with as little dead time as possible to fully exploit the available potential of the feedforward function. Several methods exist for implementing this planning within the software.A known method, as described in German patent application DE 10 2017 010 275 A1, relies on classical filtering followed by numerical differentiation. This achieves sufficiently smooth trajectories (the trajectory of the flat output and its time derivatives) to be used for flatness-based feedforward control.

[0008] Due to numerical differentiation, the known method requires very smooth trajectory profiles to avoid the aforementioned "dynamic limitations." This is achieved through filtering using multiple low-pass filters (proportional transfer characteristics with first- and second-order delays). This automatically introduces a delay (signal dead time) which ultimately has a negative impact on the overall dynamics of the feedforward function.

[0009] Document DE 10 2018 211 738 A1, for example, discloses a method for generating a manipulated variable trajectory for an actuator to influence an output variable of a system, wherein a setpoint of the output variable of the system is fed to a trajectory planning process, which generates from the setpoint a trajectory from limited input values ​​for a filter integrator chain and a trajectory from flat setpoints, wherein the trajectory from limited input values ​​and the trajectory from flat setpoints are fed to a flatness-based feedforward control, which generates the manipulated variable trajectory for the actuator from them, wherein at least one limitation depending on the trajectory from flat setpoints is applied in the trajectory planning process to generate the trajectory from limited input values.

[0010] Document DE 10 2015 222 988 A1, for example, discloses a method for controlling the speed of a drive motor in a drive train that can be operated in various operating modes and / or operating states, in particular in a hybrid drive train comprising at least two different drive motors, using speed control of a control loop with a manipulated variable limit. To improve the control behavior when regulating the speed, a planned target trajectory is corrected using a model as soon as the manipulated variable limit is violated by a feedforward control, whereby an output of the feedforward control is limited as soon as the output of the feedforward control exceeds the manipulated variable limit.

[0011] Document DE 10 2021 116 878 A1 describes, for example, a device for controlling the longitudinal and / or lateral guidance of a vehicle, wherein the device includes a trajectory planner. The trajectory planner is configured to determine target trajectory parameters for a trajectory of the vehicle, wherein the target trajectory parameters comprise at least a first trajectory state and a second trajectory state, the second trajectory state corresponding to a temporal change of the first trajectory state. A feedforward control system is configured to calculate a feedforward parameter from the target trajectory parameters and a model of the dynamic behavior of the vehicle, wherein the feedforward parameter and the first trajectory state are parameters of the same type.The device is set up to determine a manipulated variable for a controlled system within the framework of longitudinal and / or transverse guidance, depending on the input control variable, wherein the manipulated variable and the input control variable are quantities of the same type. Summary of the invention

[0012] One object of the invention is to provide an improved feedforward control system for a vehicle's drivetrain and thus an improved vehicle.

[0013] This need can be met by the subject matter of the present invention according to independent claim 1. Advantageous embodiments of the present invention are described in the dependent claims.

[0014] The inventive pre-control system for a vehicle drive train comprises a control unit with a communication device.

[0015] According to the invention, the communication device serves to receive data from a vehicle's accelerator pedal. It should be noted that bidirectional data exchange between the control unit's communication device and an external component, such as a "server," is also possible.

[0016] According to the present invention, the control unit is configured to calculate a trajectory for the pilot control system based on the data from the vehicle's accelerator pedal, namely an input signal.

[0017] According to the invention, the trajectory is composed at least partially, namely section by section or in its entirety, of a sequence of parametrically modeled curves, namely Bezier curves.

[0018] Furthermore, the control unit is preferably configured to plan a parametrically modeled curve for a trajectory segment based on a current accelerator pedal value of the vehicle and on a previous parametrically modeled curve for a previous trajectory segment in such a way that a continuously differentiable transition between the two trajectory segments is ensured.

[0019] The control unit is preferably configured to determine at least three control points based on the data from the accelerator pedal received via the communication device, i.e. an input signal, which define the course of a parametrically modeled curve in a time / input signal plane, then to establish a function equation for the parametrically modeled curve defined by the control points, and subsequently to evaluate the function equation with any temporal resolution in a defined time interval.

[0020] The control unit is also preferably configured to determine the control points in the time / input signal plane such that the time coordinates are equidistant and that a first control point is determined based on an end value of the preceding trajectory segment, a third control point is determined based on the input signal, and a second control point, namely a control point between the first control point and the third control point, is calculated from the first control point and the derivative of the function describing the preceding trajectory segment at the end time, so that the derivatives of the two trajectory segments coincide at the point where they collide or are combined, i.e., the derivatives with respect to the curve parameter are continuous.

[0021] A vehicle according to the invention comprises a pilot control system according to the invention.

[0022] By means of the design of the feedforward control system according to the invention, in particular the control unit of the feedforward control system, analytically differentiable target trajectories can be easily calculated based on the current accelerator pedal value of the vehicle by interpolation using Bézier curves, and numerical differentiation is not required, resulting in smoother and less delayed signal responses. When the feedforward control system is applied, for example, in a flatness-based feedforward control system for a vehicle's powertrain, its design according to the invention results in improved driving dynamics and greater driving comfort. Brief description of the drawings

[0023] The invention is described below by way of example with reference to the drawings. Fig. Figure 1 shows a block diagram for a feedforward control system according to an embodiment of the invention. Fig. Figure 2 schematically shows a strategy for trajectory planning by a control unit. Fig. Figure 3 schematically shows a strategy for determining one Bézier curve per time step for the in Fig. 2. Trajectory planning by a control unit is shown. Fig. Figure 4 shows the schematic course of a trajectory composed of Bezier curves. Detailed description of the invention

[0024] The following describes the inventive feedforward control system 1 for a vehicle powertrain in the context of a flatness-based feedforward control of an electric machine of an electric powertrain. However, the application of the feedforward control system 1 to precisely this application is not a prerequisite and serves only as an example.

[0025] Fig. Figure 1 shows a block diagram of a feedforward control system 1. The feedforward control system 1 has a control unit 2. The control unit 2 has a communication device 3 for receiving data from an accelerator pedal 4 of a vehicle.

[0026] The control unit 2 is configured to, based on the data from the vehicle's accelerator pedal 4, namely an input signal u k , to calculate a trajectory for the feedforward control system 1, wherein the entire trajectory consists of a sequence of parametrically modeled curves, namely Bezier curves b k (τ), is composite. That is, from the calculation of individual Bézier curves b k (τ) and their piecewise assembly yields the trajectory and its temporal derivatives ( Fig. 2).

[0027] In the block diagram of Fig. 2 and Fig. In section 3, an optional prediction function TP is also specified, which can be executed by the control unit 2 of the feedforward control system 1. The prediction function TP serves for the short-term prediction of the input signal u. k - The control unit 2 is thus designed with a prediction function TP such that a future input signal u k is calculated. The result consists of both a time value ("x-value") and a function value ("y-value"). This can be clearly explained using an accelerator pedal curve: Even before the driver has fully depressed the accelerator pedal 4, the control unit 2 provides information about the time and the objective function value of the accelerator pedal movement via the prediction function TP, to a first approximation ( Fig. 2).

[0028] In the subsequent calculation of the Bezier curves b k (τ) these results, i.e. the short-term target value forecasts (time value and function value), can be included.

[0029] Subsequently, with u k denotes the value of the time-discrete input signal to be approximately interpolated by a continuous trajectory at time t=kT, k ≥ 0, where T represents the sampling time or the time step size.

[0030] Interpolation is performed by piecewise assembling Bezier curves b k (τ) of degree n ≥ 2. bk(τ)=[xk(τ)yk(τ)],τ∈[0,1] denotes the Bezier curve b to be planned in the k-th time step. k (τ) with the curve parameter τ, defined by the control points to be determined b 0,k , b 1,k , b 2,k , ..., b n,k in the xy-plane ( Fig. 2, Fig. 3) The x-coordinate corresponds to time t and the y-coordinate to the input signal u. k or the input value.

[0031] The relationship between checkpoints b 0,k , b 1,k , b 2,k, ..., b n,k and Bezier curve b k (τ) is through bk(τ)=∑i=1nBin+(τ)bi,k given, whereby Bin(τ) so-called Bernstein polynomials.

[0032] The x-coordinates of the control points b 0,k , b 1,k , b 2,k , ..., b n,k will be equidistant with a distance of 12T chosen: x0,k=0,x1,k=12T,…,xn,k=n2T, dh b0,k=[0y0,k],b1,k=[12Ty1,k],b2,k=[Ty2,k],…,b4,k=[n2Ty4,k]⋅

[0033] Because of the equidistance, the following applies: bk(2nTt)=[tyk(2nTt)], dh between the curve parameter τ and the x-coordinate of the Bezier curve b k (τ), i.e., time t, there is a linear relationship, described by the factor 2nT.

[0034] It should be noted that the respective Bezier curve b k (τ) always only for the interval τ∈[0,2n] The evaluation is based on the time interval [0, T] or, in absolute terms, the time interval [kT, (k + 1)T].

[0035] The y-coordinates of the control points b 0,k , b 1,k , b 2,k , ..., b n,k will now be determined as follows: • y 0,k is determined in such a way that there is a continuous transition between the previous and the new Bezier curve b k (τ) gives, i.e. bk(0)=bk-1(2n) This will be determined by the election y0,k=yk−1(2n) reached. • y 1,k is determined in such a way that there is a continuously differentiable transition between the preceding and the new Bezier curve b. k (τ) gives, i.e. ddτbk(0)=ddτbk-1(2n) In other words, this means that the two Bezier curves b k (τ) should have the same slope at the point where they are combined. This is achieved by choosing y1,k=y0,k+1nddτyk−1(2n) reached. • y 2,k is determined according to the current value of the input signal to be interpolated u k chosen, i.e. y 2,k = u k .

[0036] For the case n > 2, the remaining points are determined by a linear extrapolation based on the points b 0,k and b 2,k certain, i.e. yi,k=y0,k+i2(y2,k−y0,k).

[0037] Since the checkpoints b 0,k , b 1,k , b 2,k , ..., b n,k now determined, the trajectory for the time interval t ∈ [kT, (k + 1)T] is via the relationship bk(τ)=∑i=1nBin(τ)bi,k,τ∈[0,2n] clearly defined and can be evaluated with any desired temporal resolution.

[0038] It should be noted again that the Bezier curve b k(τ) has the curve parameter τ. The x-coordinate corresponds to the time t relative to the beginning of the current sampling interval, i.e., t - kT, and the y-coordinate to the function value.

[0039] Since the trajectory is available as an analytical function, it can also be differentiated for flatness-based feedforward control. However, since the time t and the curve parameter τ vary by a scaling factor, as mentioned previously. 2nT It should be noted that the derivative with respect to time and the derivative with respect to the curve parameter τ also differ by this factor. That is, the time derivative is obtained from the derivative with respect to the curve parameter τ by multiplying by the factor. 2nT. Because of this factor, the interval corresponds to [0,2n] of the curve parameter τ the time interval [0, T] or, in absolute terms, [kT, (k + 1)T]. This means that in the subsequent evaluation, only the part of the Bézier curve segment with τ∈[0,2n] is used instead of τ ∈ [0,1].

[0040] In order to plan the trajectory for the time interval t ∈ [kT, (k + 1)T], k ≥ 0, in addition to the value of the discrete-time input signal u, one needs k because of y0,k=yk−1(2n) and y1,k=y0,k+1nddτyk−1(2n) also information about the trajectory at the end of the preceding time interval [(k - 1)T, kT].

[0041] For initialization, it is therefore assumed that the function value, i.e., the y-value, of the trajectory is identically 0 for t < 0. With y−1(2n)=0 and ddτyk−1(2n)=0 For planning the first trajectory segment at time k = 0, y is therefore given. 0,0 = 0 and y 1,0 = 0 and further y 2,0 = u0 and yi,0=i2y2,0, i = 3, ..., n.

[0042] In Fig. 4 is one of individual Bezier curves b k (τ) composite trajectory represented, where a single Bezier curve b k (τ) fourth degree, that is, based on five control points, namely a first control point b 0,k , a second checkpoint b 1,k , a third checkpoint b 2,k , a fourth checkpoint b 3,k and a fifth checkpoint b 4,k , shown in more detail (dashed line). Time t is plotted on the x-axis. The input signal u is plotted on the y-axis. k , i.e., a reading from the accelerator pedal.

[0043] In Fig. 4 are the input signal u kand the trajectory determined via control unit 2 of the feedforward control system 1 is shown. The "steps" shown represent the accelerator pedal values. Up to time x = kT, the trajectory determined up to that point from individual Bézier curves b k The trajectory determined by (τ) is represented by a continuous curve. The five control points calculated at time x = kT are shown below. 0,k , b 1,k , b 2,k , b 3,k , b 4,k are marked as black diamonds. The route through these five control points b 0,k , b 1,k , b 2,k , b 3,k , b 4,k defined Bezier curve b k (τ) is represented by a dashed line. Two essential properties can be seen in this representation, namely that the trajectory is not exactly defined by the input signal u. k The trajectory must run past defined control points and that the determined trajectory changes by a whole time step of the input signal u. kis delayed because a value of the input signal must always be available. k "From the future" are available for planning.

[0044] A key property of Bezier curves b k (τ) consists of the global influence of each control point b 0,k , b 1,k , b 2,k , b 3,k , b 4,k on the entire Bezier curve b k (τ) . This means: any control point b 0,k , b 1,k , b 2,k , b 3,k , b 4,k If the Bezier curve is moved or added, the entire curve changes. k (τ). This property makes it possible to derive at least one predicted control point from the prediction function TP ( Fig. 2, Fig. 3) advantageously included in the calculation of the individual Bezier curves b k to incorporate (τ).

[0045] The term “1 / z” in the block diagram of the Fig. 3 describes a delay of one time step.

[0046] In Fig. 4 is a Bezier curve defined by the dashed line. k (τ) shown, which was determined via the control unit 2 without a prediction function TP.

[0047] The use of filters can be dispensed with by means of the feedforward control system 1, because trajectories are generated instead by means of the control unit 2 of the feedforward control system 1, which are modeled using parametrically modeled curves (Bezier curves b). k (τ)). These curves are generated using analytical functions from predefined values ​​(input signal u). k ) and the support points calculated from them (checkpoints b) 0,k , b 1,k , b 2,k , ..., b n,k) described, so that numerical differentiation can be dispensed with. This reduces the signal dead time to a minimum. Furthermore, control unit 1 allows support points further in the future to be included in the planning, thereby further improving both the trajectories and the feedforward control function. Reference symbol list 1 Input control system 2 Control unit 3 Communication device 4 Accelerator pedal u k Input signal (accelerator pedal value) T sampling time or time step size t time k time step b k (τ) Bezier curve (Bezier curve to be planned in the k-th time step) τ Curve parameters b 0,k First checkpoint b 1,k Second checkpoint b 2,k Third checkpoint b 3,k Fourth checkpoint b4,k Fifth checkpoint b n,k (n+1)th checkpoint x x-axis y y-axis

Claims

[1] Pilot control system (1) for a drive train of a vehicle comprising a control unit (2) with a communication device (3) for receiving data from an accelerator pedal (4) of the vehicle, wherein the control unit (2) is designed to, based on the data of the accelerator pedal (4), namely an input signal (u k ), to calculate a trajectory for the pilot control system (1), wherein the trajectory is at least partially composed of a sequence of parametrically modelled curves, namely Bezier curves (b k (τ)). [2] Pilot control system (1) according to claim 1, characterized by that the control unit (2) is designed to generate a Bezier curve (b k (τ)) for a trajectory segment, based on a current accelerator pedal value and on a previous Bezier curve (b k-1(τ)) for a previous trajectory segment in such a way that a continuously differentiable transition between the two trajectory segments is guaranteed. [3] Pilot control system (1) according to claim 1 or 2, characterized by that the control unit (2) is set up to, based on the data of the accelerator pedal (4) received via the communication device (3), i.e. an input signal (u k ), at least three checkpoints (b 0,k , b 1,k , b 2,k , ..., b n,k ) that represent the course of a Bezier curve (b k (τ)) in a time / input signal plane, then define a functional equation for the signal passing through the control points (b 0,k , b 1,k , b 2,k , ..., b n,k ) defined Bezier curve (b k (τ)) and then evaluate the functional equation with any temporal resolution in a defined time interval. [4] Pilot control system (1) according to claim 3, characterized by that the control unit (2) is set up to control the control points (b 0,k , b 1,k , b 2,k , ..., b n,k ) in the time / input signal plane so that the time coordinates are equidistant, a first control point (b 0,k ) is determined based on a final value of the previous trajectory piece, a third control point (b 2,k ) based on the input signal (u k ) and a second control point (b 1,k ), namely a checkpoint between the first checkpoint (b 0,k ) and the third checkpoint (b 2,k ), from the first checkpoint (b 0,k ) and the derivative of the function describing the previous trajectory piece at the end time, so that the derivatives of the two trajectory pieces agree at the point where they collide. [5] Vehicle with a pilot control system (1) according to one of claims 1 to 4.

Images