DE3490712C2 - Driverless vehicle control and guidance system - Google Patents

Abstract

A desired route for the vehicle is stored in the vehicle in the form of co-ordinates in a ground reference frame. The 'vectors' between these junction points are divided by successive reference points into incremental vectors, the reference points being generated ahead of the vehicle at regular intervals. A dead reckoning system predicts the position of the vehicle at the end of each interval and this estimate is corrected, using a Kalman filter, by an independent fixed-target detection system using a scanning laser. The error between the estimated vehicle position and the local incremental vector provides a steering angle correction for the vehicle and the vehicle speed is dependent upon the lag of the vehicle behind the generation of reference points.

Description

The invention relates to a vehicle control and control system according to the preamble of claim 1.

Although the position of a vehicle in certain locations thanks to sensors on board and external position markers difficult to determine exactly when trying to move a vehicle to control steady (soft) and economical way precisely. in the in general, an unmanned vehicle can only be on one move predetermined path by either fixed rails used against the wheels, or cables (or similar metallic lines) that are under the one holding track are buried. Such facilities with rails or cables are expensive and inappropriate permanently because the railways (routes) thereby once and for all be stuck.

FR 2 526 181 A1 is an automatic vehicle control system according to the preamble of claim 1, in which a Dead reckoning device periodically by a vehicle transmitter / beacon system is calibrated again. That determines Vehicle periodically its position based on the beacon system. As soon as the vehicle uses the beacon system to determine its position has determined, this is the in a vehicle memory the dead reckoning device superimposed value. There can therefore neither be a comparison nor a comparison Deviation can be determined. To calculate the vehicle position, two bearing angle measurements are required, provided the vehicle does not have three sensors. That kind of he averaging the vehicle position is complex.

The invention has for its object a vehicle tax guiding and control system of the type mentioned at the beginning for a freely maneuverable Unmann t vehicle to indicate the steering of a vehicle allows a desired path in a simple manner.

The inventive solution to this problem is in the patent saying 1 marked.

With this solution, the vehicle position is not external certainly. You only need the bearing angle of the beacon (the Target) to be determined. So there is only one Bearing angle measurement sufficient. This is the actual bearing angle the bearing angle determined by the dead reckoning system compared, and the resulting error (the resulting Deviation) is used to correct both the vehicle course and also used the vehicle position.

Further developments are specified in the subclaims.

A preferred embodiment of a vehicle control and control system according to the invention the drawing described in more detail. Show it:

Fig. 1 is a schematic representation of a railway vehicle on a limited area of a factory floor,

Fig. 2 is a schematic illustration of a vehicle to be guided along a predetermined path and controlled

FIGS. 3 and 4 are diagrams of the position and orientation of the traveling toy natenachsen on the factory floor surface with respect to Fabrikkoordi,

FIGS. 5 and 6 shows schematically the path of a vehicle, the free track width of the vehicle and turns in the web,

Fig. 7 is a block diagram of the system,

Fig. 8 is a block diagram of a Kalman FilterPro zesses of the system,

Fig. 9 represents a graph showing the position and direction errors of the vehicle, and

FIG. 10 is a diagram of a transformation between "Factory Frame" coordinates and "chassis" - coordinates for determining position errors.

Under "dead reckoning device" there should be one Device for navigation depending on the Finding a relative movement between the vehicle and soil are understood.

Fig. 1 shows a factory floor surface 1 with mounted thereon, forming a coordinate system coordinate axes X and Y. The coordinates of a point (or location) on this surface are referred to as "factory coordinates" as opposed to "vehicle coordinates".

A vehicle T, which may be a flat transport vehicle, is to travel unmanned on the area 1 between stations A, B, C and D along a path which is determined by the factory coordinates of these stations. The path shown is only for explanation and can be much more complicated.

According to Fig. 2, the vehicle has drive wheels 3 with a differential gear, if necessary, and a pivotable wheel 5 at one or both ends, the steering angle of which is controllable and the one (not shown) steering angle converter and one measuring the distance traveled Has (not shown) travel converter for feedback of a travel actual value signal in order to compare it with a travel target value.

The drive wheels 3 are speed-controlled and are fed via a transmission and a DC converter from a battery 7 in a manner known per se.

The main features of the vehicle are that it's the steering angle and the speed controlling or regulating drive device points. The mechanical device can fiction be optimized accordingly, but is not critical. Steering can alternatively be done by a dif ferential control of the two drive wheels instead through drive control of a swiveling wheel be effected. The mechanical structure of the vehicle establishment therefore need not be described in detail to become.

A drive control and steering electronics 9 it holds path and steering angle signals from the pivotable wheel 5 and generates steering angle control signals and Ge speed control signals, as will be described in more detail below.

The device described so far is necessary for a Kop pelnavigation system in which the vehicle location is determined on an incremental basis from the known size of a distance traveled and a direction taken. With such systems, although in some applications where the distances are small, cumulative errors due to wheel slip, uneven surfaces, wear, etc. can occur. According to a feature of the invention, therefore, a monitoring reference and correction system is provided, which is based on the establishment of fixed reference points and the location of the vehicle relative to them. The vehicle is provided with a laser source 11 , which is mounted so that it rotates constantly around a vertical axis. The laser beam has a small width and extends in height, so that a thin vertical line of illumination is created on a target entering the beam path. On the area 1 , several targets 13 are permanently installed in various positions, so that the vehicle can - as far as possible - "see" one or more targets from each location on the surface. Of course, a factory may have facilities and storage devices moved around and obscuring one or more targets from specific locations. In such cases, the dead reckoning system can be "left to its own devices" for at least a short time.

The targets 13 are formed from retro reflectors which reflect incident light in the direction of incidence. They can be made from vertical strips of retroreflective material that are coded in width or by presence and absence to represent their identifier. Expediently, they are arranged on shelves in or above head height to prevent the laser beam from being interrupted by objects on the floor. The scanning laser beam can accordingly be directed upwards and have such a large vertical angular extent that it can detect targets at the distances in question.

The laser scanning system is the subject of the GB patent application dung 8 313 339 and will therefore not be described in greater detail here described.

The vehicle is equipped with a receiver which receives the beam reflected from a target 13 and forms a symbol representing the direction of the reflected beam and thus the target direction relative to the vehicle course.

The latter parameter is the angle that the Longitudinal vehicle axis with the X axis of the factory coordina tensystems includes. The "course" is not necessary way, since the steering wheel in the be does not have to stand straight at the right moment.

The vehicle device also contains a microprocessor and data storage device 19 . The data stored prior to the start of a trip includes a track "map" in the form of the coordinates of points A, B, C and D. The track may have sections where reduced vehicle speed is required and the limit speed for each straight Section (segment or vector) A to B etc. is prescribed and saved.

In addition to speed limits along each orbit vector it can happen that the tole satchel a transverse deviation of the vehicle from the track Change line from vector to vector. So there is practical a "free" (freely available) web width for each vek gate that are not exceeded by the vehicle  allowed, and this web width can at vector transitions to change. The path width is also for each vector saved before each trip.

The above information, path data (coordinates of transition points), permissible speed for each vector and web width of each vector can be stored in the vehicle data memory by manual input of a program or transmitted from a base station 15 to a communication antenna 17 on the vehicle who the, the antenna receives the data and transmits it to the vehicle data storage and microprocessor device 19 . The transmission or communication unit can regularly transmit the current vehicle position to the base station or be queried for the same purpose.

The computing and processing operations carried out here are additionally explained below with reference to the following figures. Fig. 3 shows the vehicle navigation coordinates, the position is given by the "factory" coordinates x and y, the vehicle course of the angle ψ, ie the angle between the vehicle axis and the X axis, the forward speed V and Rotational speed, ie Winkelgeschwin speed of the vehicle, U is.

Fig. 4 also shows the vehicle target coordinates. The target reflector R with the factory coordinates x _i and y _i is determined at an angle θ _i to the vehicle course, the vehicle itself at a certain time t the position coordinates x and y and the course ψ.

Figure 5 shows the transition (junction) of two vectors where the free web widths determined by the stored data are the same. It is essential for "soft" operation that the path at vector transitions is continuous and not angular. Therefore, a curved path or segment (curve) is calculated to adjust the intersection accordingly. Here, the curvature must be as small as possible, that is, the radius of curvature must be as large as possible in order to keep the forces acting on the vehicle in the curve (turning) as small as possible and at the same time ensure that the web width is nowhere smaller than the smaller of the two web widths is. According to FIG. 5, the radius of curvature is therefore equal to half the common seed web width. According to FIG. 6, the vector junctions of variable widths, representing the angepaß th curves at the points E and F, however, each have a radius of pw _a / 2 and _b pw / 2 have.

Fig. 7 illustrates in the form of a block diagram the navigation processing operations performed on the vehicle.

The data representing the coordinate points on the path and the path widths of the different path vectors are received and stored ( 21 ). The processor then calculates ( 23 ) the soft curves required at the vector transitions based on the web widths, as already mentioned. The curves shown in FIGS. 5 and 6 with a constant radius cannot be achieved in practice, since the transition from a straight path to a circular path causes an instantaneous change in the angular velocity from zero to a finite value and thus an infinitely high acceleration and force would mean. With an ideal curve, the curvature should increase and then decrease linearly.

An approximation to this curve is given by Ver implementation of the following equation:  

Where α is the fraction into which the turn is divided, ie tenths, fifteenths or any other fraction,
₂ a vector representing the coordinates of a transition point,
₁ and ₃ vectors representing the coordinates of the points at the beginning and end of the curve, and
the running vector representing the coordinates of the points on the curve after every fraction α.

By inserting one tenth, two tenths, three tenths, etc. for α, you get the coordinates on successive points on the curve. This process is carried out at 23 in FIG. 7 with a value for α which is determined in the manner explained below.

This principle of an incremental composition of the Curves also apply to the straight sections, with for the coordinates of successive points on the ge straight section applies

where x₀ and x₁ are the coordinates of points at the beginning and end of the straight section.

While the length of the straight sections is about 20 to 30 m, each incremental vector can be one Have a length of about 5 cm. The fraction α is then 1/500.

The incremental vectors are formed in block 25 from the initial data ( 27 ), which represent the speed limit values for the various path vectors.

Each incremental vector is by the coordinates of the point (the "reference point") at its front End determined. The reference point is determined once per base time period (at about 20 Hz) that by a clock pulse generator of the system is determined. Given the (maximum) speed for each vector is determined and the time period is fixed be the maximum length of each incremental vector Right. At a speed of 1 m / s this is The length of the incremental vector is therefore 5 cm. After this the maximum length has been determined in this way it can be reduced slightly to an even number to get incremental vectors in the orbit vector from the α represents the reciprocal.

If the length of the incremental vector is less than their maximum permissible value is selected, reduced the speed of the vehicle in this Ab cut according to the path. This way a dynamic regulation of the speed.

When composing the curve in the manner mentioned by calculating successive reference points, the incremented reference course ψ _{r is determined} as the angle which the line between successive reference points includes the factory X-axis.

The reference points calculated in this way, the desired path in contrast to the path actually tracked len, are generated once per base time period and for a comparison with those taken by the vehicle an actual position is saved. These actual positions are practical estimates by the Koppelelnaviga tion system, after testing and correction by the laser Target frame of reference.  

If the generated reference points due to inertia units etc. run in front of the vehicle, the Posi difference between the last reference point and the Vehicle can be regarded as a path error that the driving is trying to reduce stuff. The greater this path error the faster the driving speed should be be stuff (within the range for the railway gate or the limit in question), to decrease it. Conversely, the speed should be speed can be reduced if the generated reference points immediately in front of the vehicle position. This too the latter case occurs when the vehicle approaching a stop at the end of a path vector. The Number of reference points that are "on hold", is therefore a measure of the speed required and the vehicle drive motor is ge accordingly controls.

The generated reference points and the associated incremental reference course ψ are passed to determine position and course errors ( 29 ), one pair at a time.

The estimated value of the vehicle position and the vehicle course is determined by a Kalman filter predictor process 37 shown in FIG. 7. Target determination input signals θ _i ( FIG. 4), which were formed by the laser reflector system 33 , are used. The path covered by the vehicle and the steering angle are signals are formed by the aforementioned converter 35 of FIG. 7. The Kalman predictor process itself ( 37 ) is shown in greater detail in FIG. 8.

According to FIG. 8, the first process to be performed is the estimation of the position and of the course at the end of a basic time period (At) from the actual or abge estimated forward and rotational speeds of the driving zeugs during this time and the position on at scavenging this period.

That through the converter in the position prediction process entered signals represent the by the swiveling Wheel measured steering angle Φ and by the swivel bare wheel steered wheel traveled, in the form of counting represented path impulses. The path increment and the increment Δt gives the speed in the direction of the swiveling wheel from which the forward speed speed V of the vehicle in its course direction by multi plication determined with the cosine of the steering angle Φ becomes. The angular velocity U of the vehicle is by multiplying the steering wheel speed by Sine of the steering angle Φ according to the geometry of the Vehicle wheels determined. The speeds V and U are calculated in every period under the assumption net that the speeds and the steering angle in this (short) time period Δt are constant.

The equations realized in the process of position predictor 39 of FIG. 8 are derived below. Thus, for reference to Figure 3 for the change speeds of the position coordinates x and y, and specify the heading angle Ψ the following equations.:

By integrating these three equations over time Δt the following equations result:  

According to the first of these equations, the value ψ is Time (t + Δt) equal to the value of ψ at time t plus that Angle by which the vehicle moves in the time span Δt turned. Have in the second and third equations t and (t + Δt) have the same meaning as in the first Equation.

These equations can be rewritten for the estimates become:

The circumflex over a parameter denotes one Estimated value, and the symbol "| t" means "evaluated at time t ".

As can be seen, these equations are sufficient to estimate the position and course of the vehicle at time (t + Δt) if the estimated position and course at time t are known. The output signals of block 39 are therefore the estimated coordinates x and y and the estimated course ψ of the vehicle after a further period of time .DELTA.t, which, apart from input variable 41, are based only on the dead reckoning navigation system of the path covered and the angle passed.

According to FIG. 4, the estimated angle may be a target reflector R in the form of the vehicle heading ψ and the coordi nates of the vehicle and the target by monitoring the following slip are determined:

In this equation, _{i is} the estimated target angle at time (t + Δt), evaluated at time t. The coordinates x _i and y _{i of} the target are predetermined by the arrangement of the targets on the factory floor. The estimated values x and y of the vehicle and the estimated value ψ of the course are determined in the process 39 . The equation is thus processed in a target bearing predictor 43 , which forms the output variable _i .

The output of the laser targeting system 33 is an accurate representation of the target angle θ _i , which is compared to the estimate _i in a process 45 to determine a target estimate error θ _i - _i .

This error signal is processed by the Kalman filter 47 , which multiplies the error by Kalman amplification factors k _r , k _x and k _y . The correction products are then added in a process 49 to the dead reckoning predictions of process 39 according to the following equations to obtain corrected estimates of vehicle heading and position:

The Derivation of the Kalman Correction Products and the Us The Kalman filter is described in the book "Optimi sation of Stochastic Systems "by M. Aoki ben by Academic Press 1967.  

In this way, corrected estimates of the course and position at time (t + Δt), evaluated at time t + Δt, are determined from the estimated values of the course and position, which were determined at time t and corrected by the Kalman filter process . These best estimates are entered in the error determination process 29 according to FIG. 7 and also as "instantaneous" input variables 41 in the position predictor 39 according to FIG. 8 in order to predict the next reference point from them. Therefore, if one or more target reference corrections fail, for example because the targets are covered, then the next reference point is calculated in advance by means of coupling navigation from the last reference point at which a target determination correction was carried out.

The determination of path and course errors is described below with reference to FIG. 7. The two inputs to process 39 are a) the generated sequence of reference points that define the ideal trajectory, and b) the best estimate of the course and the position of the vehicle averaged by the Kalman process of FIG. 8.

FIG. 9 shows successive incremental vectors IV1 and IV2 and their associated reference points RP1 and RP2, which were determined after process 25 ( FIG. 7). Errors in the navigation of the vehicle T are shown in the form of the vertical distance d _{e of} the vehicle center from the local incremental vector and as an angle error θ _e between the vehicle course and the direction of the local incremental vector.

These errors d _e and θ _e are measured by transforming the actual value of the vehicle position in factory coordinates into a position in a vehicle reference frame, in which the zero point with the reference point of the local incremental vector and the new X axis with the local incremental vector. This transformation is shown in Fig. 10, in which X and Y are the factory coordinate axes, X * and Y * are the vehicle frame axes, x _r and y _{r are} the coordinates of the local incremental vector reference point in the factory frame, x and y are the Coordinates of the vehicle in the factory frame and x * and y * are the vehicle coordinates in the vehicle frame.

The angle ψ _r is the course or the direction of the incremental vector relative to the factory frame. The transformation equations according to the following can be derived from FIG. 10 by simple geometric considerations:

x * = (x - x _r ) cosψ _r + (y - y _r ) sinψ _r
y * = (y - y _r ) cosψ _r - (x - x _r ) sinψ _r .

In this vehicle reference frame, the distance or path error d _{e is} the y * coordinate of the vehicle center and the angle error θ _{e is} the direct transformation angle ψ _r .

When the vehicle passes the local reference point x _r , y _r , the sign of x * changes from minus to plus. This change triggers a deletion of the current vehicle reference frame and its re-training in such a way that its zero point coincides with the following reference point and the x * axis with the next incremental vector. It can therefore be seen that the vehicle frame progresses synchronously with the vehicle.

The error values d _e and θ _e are therefore used to derive a steering angle setpoint signal θ _d as a direct function of these error values (directly proportional to these error values). The setpoint value of the angular velocity U _d is first determined as:

U _d = K₁ d _e + K₂ θ _e

in the K₁ and K₂ gain functions (gain factors) are caused by the dynamics of the vehicle are determined themselves. The target value of the steering angle is then from the vehicle geometry and the target value of the pre forward speed of the vehicle is calculated.

As already mentioned, the vehicle is speeding up speed by measuring the number of incremental vectors between the vehicle and the last generated ge controls. The last one created is always in front of the drive stuff and seems to "move on" the vehicle as if an elastic band the vehicle with the last one telten reference position connects.

The reference points become a stationary position at a linear speed from the "rest" state tion formed away, so that they mentioned the "elastic Practically stretch the belt ". The vehicle is adjusted accordingly the distance from the last reference point and depending speed of its inertia and performance so that the error is gradually reduced, and until a steady state is reached in which the vehicle speed equal to the reproductive rate speed of the reference points.

The reference point speed and thus the vehicle speed can be during a trip during a predetermined path can be changed dynamically, namely  by changing the increment length (i.e. by changing of α in equation (1)) between successive ones Reference points.

If the generation of increment vectors and reference point ten is completed, the last incremental ends Vector at the last stop position, which is determined by the ur original course is prescribed, which in turn in Coordinate position determination path vectors predefined is. The vehicle speed decreases with decreasing distance or path error. By influencing solution of the ratio of distance error to speed speed setpoint is regulated so that the end point is reached without overshoot. Also here the vehicle frame is directly applicable again bar because the x * coordinate of the to be returned in the frame way is and this can be done without further calculation be watched over.

The electromagnetic bearing device is at obi According to the embodiment, a laser system through which Direction measurement using an azimuthal scanning thin NEN laser beam is effected. Instead you can Radar beams are also used, which are accurate Enable bearing by phase comparison. Furthermore, the reflectors can be transponders (Ant word devices) with coded broadcasts.

Claims (7)

1. Vehicle control and guidance system with a motion drive device ( 3 , 7 , 9 ) for driving a vehicle (T), a steering device ( 5 , 9 ) for controlling the path of the vehicle (T), a dead reckoning navigation device ( 35 , 39 ) for calculating the position and course of the vehicle (T) on an incremental basis, a device ( 21 ) for storing a desired path (ABCD) of the vehicle (T), a device ( 29 ) for controlling the vehicle drive and steering device ( 3 , 5 , 7 , 9 ) to drive the vehicle (T) along the desired path (ABCD), and a device ( 19 ) for storing the position of one or more fixed reference targets ( 13 ), characterized by a processing device ( 43 ) for predicting the bearing angle of the target or targets ( 13 ) from the calculated vehicle position and the calculated vehicle course, an electromagnetic direction finder ( 33 ) for determining the actual bearing angle of the target or targets ( 13 ), means ( 45 ) for comparing the actual bearing angle (θ _i ) of the target or targets ( 13 ) with the predicted bearing angle ( _i ) and means ( 47 , 49 ) for correcting the control ( 29 ) the vehicle drive and steering device ( 3 , 5 , 7 , 9 ) depending on a bearing angle error (θ _i - _i ). 2. System according to claim 1, characterized in that the correction device ( 47 , 49 ) comprises: a Kalman-Fil tereinrichtung ( 47 ) for generating correction products from the bearing angle error (θ _i - _i ) and the Kalman gain factors with respect to the Position coordinates and the vehicle course and a device ( 49 ) for summing the correction products and correspond to the estimated values of the position and the course, which are formed by the dead reckoning device ( 35 , 39 ), the sums forming the best estimates of the position and the course of the vehicle . 3. System according to claim 1 or claim 2, characterized by a device ( 19 ) for determining a desired path (ABCD), a storage device ( 21 ) for storing the desired path in the form of a sequence of rectilinear segments (E, F) and Has means ( 23 ) for converting the transition of the straight-line segments (E, F) into continuous, curved transition segments. 4. System according to claim 3, characterized by a device for storing a permissible maximum vehicle speed ( 27 ) and a permissible path width for each straight and curved segment depending on the local path conditions. 5. System according to claim 4, characterized in that the means for converting the transition (E, F) into curved transition segments means for calculating the radius of curvature (r _a , r _b ) of a curved segment as not less than half the value of Web width (pw _a ) and, if there are different web widths (pw _a , pw _b ) at the transition (E), has half the value of the larger of the two web widths (pw _a , pw _b ). 6. System according to claim 2, characterized in that the dead reckoning device ( 35 , 39 ) means for estimating the position and the course of the vehicle (T) after each sequence of time and distance increments from the position and the course of the vehicle (T ) at the beginning of the increment and depending on the forward and rotational movement of the vehicle (T) during this increment. 7. System according to claim 6, characterized by means for predicting the bearing angle of the reference target ( 13 ) from the vehicle (T) on the basis of the estimated position of the vehicle (T) and the last determined position of the reference target ( 13 ).