WO2008037649A1

WO2008037649A1 - Programmable logic controller having kalman filter

Info

Publication number: WO2008037649A1
Application number: PCT/EP2007/059966
Authority: WO
Inventors: Bernd-Markus Pfeiffer; Michael Brämig
Original assignee: Siemens Aktiengesellschaft
Priority date: 2006-09-25
Filing date: 2007-09-20
Publication date: 2008-04-03

Abstract

A programmable logic controller comprises a kalman filter.

Description

description

Programmable controller with Kalman filter

The invention relates to a programmable logic

Control, in particular for a load transport device.

In container ports, load-handling devices are used in the form of rubber-tired gantry cranes (RTGs), which can freely change their position in the port. For easier operation while an automatic web guide (Autosteering) is used.

In every container port, countless RTGs stack the containers of trucks on the storage yard and vice versa. In the big harbors, hundreds of devices are busy allocating thousands of containers every day to hundreds of uniform stacks. Automatically transmitted orders of a warehouse management system (for example NAVIS) with storage bin number and container number must be processed by the operator manually error-free and fast. The five-digit coded storage location (Section, Row, Bay, Stack, Slot: 01 A 23 B 4) must be approached manually and the ten-digit container number (ABCD123456) compared. These activities often result in errors during a shift, which cost a lot of time and therefore money. In particular, the manual web guidance of the RTG is extremely stressful. The driver must keep the RTG within a narrow lane at full speed by constant correction interventions, and he must remain in an unfavorable physical position to be able to see the lane markings at all.

Individual components for the different control and control technology tasks are from different manufacturers in the

Offer. However, a one-off system is not available. US 2005/0033514 A1 discloses a load transport device in which position signals obtained by means of GPS are fed to a main processor as a control device. This then processes control signals which are supplied to a programmable logic controller as a control device.

Based on this, the present invention seeks to provide a load transport device that can be modular and connect with other components.

This object is achieved by the inventions specified in the independent claims. Advantageous developments of the invention will become apparent according to the subclaims.

Accordingly, a programmable logic controller (PLC) has a Kalman filter. The Kalman filter is preferably implemented by a corresponding programming device of the PLC.

By implementing the Kalman filter in the programmable logic controller, the overall control of the load handling device can be built modularly with components from the automation environment. This makes it possible to rely on a large number of reliable components.

Preferably, the Kalman filter is a non-linear Kalman filter (extended Kalman filter).

Normally, due to the associated computational burden, it would be impossible to implement a Kalman filter on a PLC having a model for a complex dynamic system, especially a load handling device.

Therefore, the Kaiman filter is preferably based on a state space representation from kinematics, which has the change of deviation from a desired size. In particular, for a load transport device, it is recommended to select the desired path as a target size.

Among other things, in the event that the PLC as

Control means for generating control signals for controlling the load transport device or parts of the load transport device is formed on a web direction and between the web direction and the speed of the load transport device is a course angle, it is further preferred for the Kalman filter is preferably based on a state space representation from the kinematics that shows the change in the price angle.

The representation becomes particularly compact if the state space is completely unfolded by changing the deviation from the nominal path and by changing the course angle.

The control device is designed in particular as a programmable logic controller, more preferably as a control device for generating control signals for controlling a load transport device, for position determination and optionally container positioning. For position determination, position signals generated by position determining means are supplied to the control device via a connection, which in particular represent the position and / or a change in the position of the load transport device.

Advantageously, the load transport device is mobile, freely movable, i. e.g. not rail-bound, rubber-tired, a crane and especially a gantry crane. The position determining means determine the position, in particular by means of GPS.

Advantageously, the load transport device in addition to the control device to a control device which the Drives (converters, motors) of the load transport device taking into account the control signals of

Control device controls. This control device is advantageously designed as a programmable logic controller.

In addition to the position determination means, the load transport device has in particular additional position determination means for fine position determination and / or position correction, the additional ones

Generate position signals which are supplied to the control device.

The additional positioning means work in particular by means of odometry and have, for example transponder on.

The control device is preferably set up for automatic web guidance / autosteering of the load transport device taking into account the position signals and / or the additional position signals and / or the use of the Kalman filter. This facilitates, in particular, straight-line driving for the operator.

The load of the load transport device is, for example, a container or a plurality of containers. The load transport device may have means for identifying or individualizing the load, which are designed, for example, as means for recognizing the container number and for supplying the container number to the control device.

As part of the modular design, the load transport device may include means for receiving job data from a storage bin management system and for supplying the job data to the controller.

In particular, when the load transport device is movable on tracks, the control device with a Memory device, for example, be connected in the form of a memory card, on which a plurality of tracks is stored. This may be necessary because some programmable logic controllers do not have sufficient memory to store a plurality of lanes. The memory card can be changed depending on the port or warehouse. For using multiple

Load handling devices on the same port or warehouse can use memory cards with the same contents.

The load transport device can be constructed modularly in a particularly simple manner, in that the storage device is connected to the position determining means via an interface, the interface being standardized in terms of its design and / or bus. Analogously, by setting up interfaces, it is also possible to proceed with regard to the additional position determination means of the load identification means and the means for receiving order data.

In a method of controlling the load transfer device, a programmable controller is used as a control device to which position signals are supplied from position determining means. Advantages and embodiments of the method are analogous to those of the device and vice versa.

Further exemplary advantages and details emerge from the description of an embodiment with reference to the drawing. Showing:

1 shows a signal flow plan for course control of an RTG at a glance;

FIG. 2 nomenclature for kinematics;

FIG. 3, floor plan of the RTG. With the help of a Kalman filter, which contains a dynamic model of the controlled system, it is possible to distinguish physically possible and plausible movements of the RTG from interference signal components which are caused by the elastic oscillations of the setup or measurement errors of the GPS system.

The path control works with a cascade of position PI and course P-controller. Standard controller function blocks of a PLC with an additional controlled adaptation of the reset time of the position controller as a function of the travel speed are used.

FIG. 1 shows an overview of the overall system and the solution concept on the basis of a signal flow plan. The nominal trajectory runs along the x-axis, while the y-coordinate

Transverse deviations from the nominal trajectory are described, a is the heading angle and v the speed, where the indices L and R refer to the left and right side of the vehicle.

In manual mode of operation, a crane driver specifies the desired speed v and, as a steering command, the percentage speed difference Av between the left and right sides. The steering principle thus corresponds to a tracked vehicle (for example, excavator or tank). The controller calculates target speeds for the left and right side of the vehicle.

In a drive controller, the speed setpoint values are generated therefrom by means of a ramp function, namely with a sufficiently flat gradient, so that the required speeds of rotation can actually be achieved taking into account the inertia of the RTG and the drive power of the motors.

The crane kinematics describes the changes in position of the vehicle over time, i. the trajectory.

Since the antenna of a GPS is usually at the upper end of the crane, the actual trajectory of the Antenna from a superposition of the trajectory of the chassis with the elastic vibrations of the structure.

The coordinates provided by the GPS do not correspond exactly to the trajectory of the antenna, but are additionally subject to stochastic disturbances.

So the considered system model has

- as inputs the target speeds of the left and right sides, and as (measurable) output quantities the coordinates supplied by the GPS.

However, for the orbit control in a cascade of attitude and heading controllers, the actual position of the chassis and the actual heading angle are needed. Both quantities can not be directly measured and regarded as internal states of the system.

The task of the Kalman filter or observer is to calculate estimates for the non-measurable internal state variables on the basis of the measured output variables and a dynamic process model.

The task of the position controller is to control the transverse deviation y from the setpoint path to the setpoint zero. Output variable of the position controller is a setpoint for the course controller, which calculates a steering manipulated variable in the form of a speed difference from the course deviation.

FIG. 2 shows the nomenclature for kinematics.

All considerations are done in a fixed orbit coordinate system. The x-direction is the web direction, while the y-direction indicates deviations from the desired web. The chassis of the container crane is considered as a rigid body in which three points are interesting: the center point with the location vector (vector from the coordinate origin to the center)

and the two wheels (centers of wheelsets) left and right with the location vectors

The instantaneous speed (of the center point) is

The angle between the x-axis and the velocity vector v is called the heading angle a. Because of the rigid connection, the velocity vectors of all three points are parallel:

v v ₁

The crane is steered by different speeds on both wheels, i. like a tracked vehicle. The speed of the center is an average

v = - {v _L + v _R )

The transverse axis of the vehicle (from left to right) is the difference vector between the two position vectors to the wheels

q = ^r κ -r _τ and because of the rigid connection has a constant amount ("length"), namely the width of the vehicle

b = q = const.

The "course" is the tangent of the heading angle, and since the lateral axis is perpendicular to the velocity vector, it can be read from the components of the velocity or from those of the transverse axis:

tan (α) = - ^ - = ^■

V "-i

FIG. 3 shows the floor plan of the RTG. The drives are numbered 1 to 4 counterclockwise.

The control is supplied on the input side with the measured value for the path deviation y, which is calculated from the unfiltered GPS measured data in a north-east coordinate system by conversion to the local xy coordinate system of a taught-in path. The control intervention is additive to the steering commands of the driver. For this purpose, the required off course controller speed difference Av _to be distributed among the drives of both sides.

For this purpose, first a correction factor

which is to be added to the speed setpoint on the left-hand drive side and subtracted on the right-hand side. If Av ₈₀₁₁ traversed the entire range -1 <Δv _soll <1 then the correction factor would be +0.5 to -0.5. With a setting range of -0.3 <Δv _soll <0.3 of the course controller, only the range -0.13 <fk _o rr <0.17 will be traversed. In the drive control, all speed values are normalized to a range up to 16384 corresponding to 100% (maximum speed). One Steering command of the driver takes place with a maximum of 190%. The calculated value is thus normalized

f _k "= f _k∞I * 1.9 * 16384 / 0.17

that at the maximum setting intervention of 0.17 exactly 190% are output. In the power converter for the electric drives, the speed setpoint for each of the drives is, in principle, according to the formula

V ₁₁ SOlI = ^V soll ( ^{1 + f} koJ

determined, with the drive 1 and 4 of the left, the drives 2 and 3 of the right side of the vehicle (when driving in the positive x direction) are assigned. This results in a ratio of the target speeds of

Δv 1 + -

V _so ii _ v _soll (l - f) _ _{2 +} Δv _ 2 + Δv + Δv _

VL, let V soll (1 + f) be ₁ Δv 2 + Δv-Δv

2 + Δv as required in the definition. In the real inverter, however, the travel command v _soll is delayed by the joystick in the driver's cab via a ramp function with a rise time of 6s from 0 to 100%. The correction values f are limited to ± 190%, delayed by a ramp with a rise time of 2 s, and then multiplied by a factor of 5%. There are also other, minor ones

Correction values, e.g. due to the cross position of the cat with load.

Since an encoder is mounted on only one of the 4 drives, the actual speed differences are calculated from frequencies that are calculated inside the converter for the purpose of internal frequency control of the asynchronous motor (ie not really measured!): -I _* f 2 calculated

1, calculated

The drive 2 is representative of the right, the drive 1 for the left side. A check with a trace function in the inverter has shown that the calculated frequencies are almost congruent with the wheel speeds measured by the encoder, so that the installation of additional encoders does not promise any benefit.

Surprisingly, it has been found that the assumption of elastic deformation of the crane and harmonic vibration caused thereby is not realistic. Fluctuations in the y values measured by the GPS are based, on the one hand, on roll movements of the vehicle due to unevenness of the web and, on the other hand, on inaccuracies in the actual, satellite-based measuring method. The effect of long-wave ground unevenness on the path control can best be compensated by archiving not only start and end points but, for example, 10 intermediate points when learning a path.

A portion of the vehicle model, which is intended to describe the elastic vibration, however, can be removed.

However, the transition from the linear Kalman filter to the nonlinear extended Kalman filter (EKF) has proven to be advantageous, because in the extended Kalman filter the nonlinear equations of motion can be coded directly with physical parameters, so that

a direct parameterization on the basis of technical data (eg vehicle width, maximum speed), but above all - a continuous adaptation of model parameters such as the current driving speed is possible. This ensures that the observer, for example, even with negative

Speeds, i. Drive in negative x-direction, works correctly. The calculated heading angles are to be interpreted as counterclockwise backwards from the negative x-axis.

The filter can and should also continue to run in the background at standstill or manual drive in order to keep its state variables up to date. However, important is a correct initialization. When launching the Caymanian filter, the crane should actually be in the middle of its lane, with zero track and zero heading.

The EKF is based on the nonlinear state space representation from kinematics:

For purposes of discretization, the time derivative is replaced by forward difference quotient, e.g.

The following pair of state and output equation is obtained, which is implemented in the EKF:

y _k + t _o vsin (α _k) y _{k +} α _{k +} α i + _k t ₀ _is -.DELTA.V

f (* iΛ) y _k = [io] ^χ _k vh (x, u) This results in the following Jakobi matrices

Sf ₁ d ^

Sf Sx ₁ Sx ₂ 1 t ₀ vcos (α) Sh Sh Sh

= [1 0]

Sf, Sf, 0 1 Sx ^" Sx ₁ Sx,

Sx ₁ Sx,

which are also needed in the EKF.

The variance R _{k of} the output y is estimated from the measurement data online over a sliding time window of 20 samples. The covariance matrix of the process noise as the only remaining tuning parameter is given in diagonal form with relatively small, constant values, while maintaining the ratio of the two diagonal elements determined in simulations, but dividing both by a factor of 1000:

2e-8 0

^Q _k = 0 le-9

Thus, the filter relies more on its inputs in the vector u and the internal dynamics model than on the noisy measurements y.

If the GPS can no longer maintain its precise RTK mode and then yields very inaccurate values, the measured variance of the output is increased by a factor Ie9, so that, in principle, a pure process simulation without comparison with measured values is continued ,

Because of the reverse drive is outside the Kalman filter nor the track deviation

y _H = y + csin (α) the rear wheel is calculated. C describes the distance between the GPS antenna and the rear wheel in the x-direction.

Claims

claims

1. Programmable logic controller, characterized in that the programmable logic controller comprises a Kalman filter.

2. Programmable logic controller according to claim 1, characterized in that the Kalman filter is an extended Kalman filter.

3. Programmable logic controller according to one of the preceding claims, characterized in that for the Kalman filter a state space representation from the kinematics is used, which has the change in the deviation from a desired size, in particular of a desired path.

4. Programmable logic controller according to one of the preceding claims, characterized in that the programmable logic controller is designed as a control device for generating control signals for controlling a load transport device or parts of a load transport device on a web direction and between the web direction and the speed of the load transport device is a course angle.

5. Programmable logic controller according to claim 4, characterized in that for the Kalman filter a state space representation from the kinematics is used, which has the change of the course angle.

6. Programmable logic controller according to claims 3 and 5, characterized in that the state space is completely opened by the change in the deviation from the nominal path and by the change in the course angle.

7. Load transport device comprising a

Programmable logic controller according to one of Claims 1 to 6.

8. A method of controlling a load handling device using a programmable logic controller having a Kalman filter.