WO2009145694A1

WO2009145694A1 - Method and device for calculating the reliability of an estimated position

Info

Publication number: WO2009145694A1
Application number: PCT/SE2009/000273
Authority: WO
Inventors: Johan Larsson; Michael Krasser
Original assignee: Atlas Copco Rock Drills Ab
Priority date: 2008-05-30
Filing date: 2009-05-28
Publication date: 2009-12-03
Also published as: SE532430C2; CL2009001309A1; SE0801276L

Abstract

The present invention concerns a method and a device for calculating the reliability of an estimated position for a point on a mining and/or construction machine, wherein estimation of said position involves an estimation based on at least a first parameter value from at least one sensor and a representation of the environment where said machine is located, wherein said machine's position in said environment can be estimated by means of said at least one first parameter value. Said calculation of the reliability involves - calculating, from said estimated position in said representation of the environment, at least one second expected parameter value, - comparing said calculated at least one second expected parameter value with at least one actual parameter value for said machine's position in said environment, and - calculating with the help of said comparison a measure of the reliability of said estimated position.

Description

Method and device for calculating the reliability of an estimated position

Technical field

The present invention concerns position verification for mining and/or construction machines. The invention also deals with a device and a mining and/or construction machine.

PRIOR ART

In many fields there is a constant process of improving the efficiency, productivity and safety, one of which is underground mining. One area in such mining where changes/improvements are taking place to accomplish the above involves automating certain functions for at least some of the vehicles/machines used in the mine. For example, it is desirable that certain machines, such as loading ma- chines, be automatically driven, that is, not only can the machine be driven without a driver, but it can also perform functions totally by itself.

On example of such vehicles where automated operation is desired consists of so- called LHD (loading, hauling and dumping) machines. These machines are often used to remove broken rock, haul it to a particular place where the broken rock is dumped, whereupon the machine returns to the same place for a new load. Thus, these machines often perform the same manoeuvres over and over again, which makes this type of machine/manoeuvre especially well suited to automation.

These machines were previously driven manually by a driver on board the machine or by means of radio control, for example. Due to factors such as driver safety, risk of accident, and labour expenses, however, it is desirable to move such loading machines in an entirely automatic fashion.

One type of existing system to achieve such a fully automatic operation is based on a three step principle, where the machine in a route recording step is first driven manually on the course that will later be driven autonomously, at the same time as signals from various transmitters arranged on the machine are recorded. In step two, a route is generated to create, on the basis of at least some of the above recorded transmitter signals, a system of co-ordinates, covering the region in which the machine will be moving. The route driven by the machine during the recording is described in this system of co-ordinates along with information on, e.g., the suitable speed for various parts of the route. Furthermore, with the help of the transmitter information, a map representation can be created for the pathways in the tunnels where the machine is moving, if such were not previously available.

The third step consists of playback, wherein co-ordinate information as to how the machine was moved manually and the maps generated in step two (or previously generated maps) are used to move the machine autonomously along the same path that the machine travelled manually in step one.

During autonomous playback (tramming) of a route, the machine' s position in the system of co-ordinates in which the route map and the desired route of travel are defined is estimated. When estimating the position, a Kalman filter can be used, which receives as input signals the one or more maps on the route which the machine will travel and data from one or more transmitters, such as the middle angle transmitter and laser range scanners, arranged on the machine, calculating an estimated position from this data. By then comparing data from transmitters against route maps, the Kalman filter can estimate the machine's position, and the estimated position can then be used to automatically steer the machine along the travel path created during the route generating.

One problem with systems containing statistical filters, such as Kalman filters, is that it cannot be guaranteed that the estimated position is correct. If the uncertainty (such as mean error and covariance) is known for all transmitter signals, one can indeed obtain a value of the uncertainty in the estimated position.

It is especially during autonomous (auto) tramming with large, heavy and costly machines, such as LHD machines, that there is a risk of the machine causing damage either to itself or its surroundings is the estimated position is faulty. The distance from the surrounding rock walls is often narrow in relation to the machine's width, so it is very important to be able to move the machine in a safe manner with a good estimate. If the uncertainty of the transmitter signals is too large, then for example when moving along a rather long stretch there will also be great uncertainty as to when the machine reaches the end of the stretch, which in turn can lead to unwanted stopping if the longitudinal uncertainty in the machine's position is so great that the machine has to stop for safety reasons. In an actual application, the uncertainty of the transmitter signals is often not known at all with any adequate accuracy, which means that the value of the uncertainty for the estimated position becomes totally unusable, i.e., the uncertainty far exceeds the acceptable tolerance for the estimated position.

Thus, there is a need to reduce the uncertainty for the estimated position, e.g., during autotramming.

The previously known methods do not meet this need.

Summary of the invention

One objective of the present invention is to provide a method to verify the reliability of an estimated position that solves the foregoing problems.

The present invention pertains to a method and a device for calculating the reliability of an estimated position for a mining and/or construction machine, wherein estimation of said position involves an estimation based on at least a first parameter value from at least one sensor and a representation of the environment where said machine is located, wherein said machine's position in said environment can be estimated by means of said at least one first parameter value. Said calculation of the reliability involves calculating, from said estimated position in said repre- sentation of the environment, at least one second expected parameter value, comparing said calculated at least one second expected parameter value with at least one actual parameter value for said machine's position in said environment, and calculating with the help of said comparison a measure of the reliability of said estimated position.

This has the advantage that uncertainties regarding the precision of the estimate can be replaced to a large extent by a measure of reliability that is based on a comparison between the expected parameter value for the estimated position and the actual parameter value obtained for said machine's position in said environment. This parameter value can constitute a description of the machine's position in relation to the environment, and it can consist of, e.g., one or more determinations of distance in one or more directions to surrounding obstacles such as rock from a range finder placed on the machine, such as a laser range scanner or other range finder. If, e.g., a slight deviation is found between expected and measured parameter values, the position estimation can be considered reliable, even in situa- tions where the uncertainty of the estimation can otherwise be large (e.g., due to uncertainties/inaccuracies in the parameter value used to perform the estimation). This, in turn, means that the present invention enables continued operation in at least some of the situations where otherwise machines come to a stop because the uncertainty of the estimated position is too large, or the machine is halted for safety reasons in a situation where the reliability calculation shows that the position estimation has a large error, but this is not reflected by the uncertainty value.

In a third aspect of the invention, the method according to the invention is used for autonomous driving of a mining and/or construction machine to adjust the machine's speed in relation to a calculated value for the reliability of the ma- chine's estimated position, in a representation of the environment where it is located, regardless of whether the uncertainty of the sensors is known.

Brief description of the drawings

The invention will be explained in more detail by describing various embodiments thereof, based on the accompanying drawings, where

Fig. la-b, is a vehicle from the side and top, respectively, in which the present invention can be used to advantage.

Fig. 2 is an example of a mine where the present invention can be used to advantage.

Fig. 3 is a method according to the present invention, described schematically. Description of alternative embodiments

Fig. 1 A, B, shows a vehicle 100 from the side and top, respectively. The vehicle 100 is part of a loading machine in which present invention can be used to advan- tage. The machine 100 comprises a bucket 101 and wheels 102-105, and a control unit 106, which controls various functions of the machine 100. As shown in Fig. Ib, the machine is part of an articulated vehicle, whose front section 100a is joined to a rear section 100b via a joint 107. The machine also comprises at least one wheel rotation sensor 108, such as an odometer, which can be arranged on the axle emerging from the transmission and puts out signals representing the drive wheel's rotation and/or distance travelled. In an alternative embodiment, one uses a sensor arranged on the machine's drive axle 109, which puts out signals representing the rotation of the drive axle 109 and, thus, the drive wheel. The sensor 108 puts out signals to the control unit 106. Moreover, arranged at the joint 107 is a joint angle sensor 110, which measures the current joint angle and sends these signals to the control unit 106.

The machine 100 further comprises a front 111 and a rear 112 laser range scanner, which is also connected to the control unit 106 and puts out sensor signals representing measured distance, i.e., distance to the closest obstacle which stops the path of the laser beam. The laser range scanners 111, 112 can be designed, e.g., to measure the distance in certain directions in an angle interval, i.e., the distance to the closest obstacle which stops the path of the laser beam The present example uses laser range scanners to measure the distance to the closest object in the forward longitudinal direction of the front section 100a (or the backward longitudi- nal direction of the rear section 100b) and the distance to the closest object (such as rock) for each whole degree ± 90° from the respective longitudinal direction. Thus, each respective laser range scanner measures the distance at 181 respective measurement points. Of course, one can use laser range scanners which measure distance in significantly more directions, as well as those which measure distance in significantly fewer directions. One can also use a single omnidirectional laser instead. In an alternative embodiment, one uses only the scanner which happens to be "pointing" in the direction of travel (i.e., the front one 111 if the vehicle is moving forwards and vice versa). It is in no way essential to the invention for the directions to be measured by laser range scanners, but rather any desired range- finder can be used, as long as it can provide distance measurements with acceptable accuracy. Examples of other type of conceivable range-finder are those based on radar or sonar technology.

Moreover, the sample embodiment shown here uses a range scanner to measure distance in only one plane (the horizontal plane of the machine). Yet it will be obvious that range scanning can occur in more than one plane, e.g., also in a vertical plane to measure tunnel or local height, or another plane lying between the horizontal and vertical plane, thereby refining the possibilities for a correct position estimation. In yet another alternative embodiment, one or more scanners pointed at the sides can be used instead or in addition.

Moreover, the aforesaid sensors put out sensor signals to the control unit 106 at appropriate times, such as continuously or every 40 ms or more often or more sel- dom. The control unit 106 then uses the signals received as will be described below.

Figure 2 shows an example of a mine where the present invention can be used to advantage. In the example shown, the vehicle 100 uses the bucket 101 to load rocks at site A and then hauls the load for dumping at site B. When the machine 100 is set up for autotramming, the above described three-step principle can be used, i.e., a route recording is first done, where a recording of signals is activated.

The loading, hauling, dumping and return procedure can be designed to run as a single route, or with movement from A to B as a first separate route and the movement from B to A as a second separate route. Thus, the sensor signal re- cording is activated for recording of a route from A to B, whereupon an operator with the vehicle parked at point A backs up to point C and turns the vehicle around, whereupon hauling along the broken line is then done to point B, where the route recording is stopped.

Based on the recorded transmitter signals the route is then created, i.e., how the vehicle should be moved and at what speed the vehicle should be moved on different parts of the route. As mentioned above, the sensor signals can be read off every 40 ms, for example. If each sensor signal reading will constitute a route point, the number of route points becomes very large. For this reason, the route points can instead be signals determined for every half meter of the machine's movement, for example. The data stored for the route is preferably the position, the vehicle's direction and desired speed. One thus gets a route which in theory consists of a number of points, indicating for each point where the vehicle is supposed to be, what direction it should have, and the speed it should be moving at in the subsequent autotramming.

When the vehicle is then moving autonomously on the route, it is normally not enough to just use this information to carry out the desired movement, for example, because uncertainty in the sensor signals means that the end position in all likelihood will deviate from the calculated one, so that also the starting position for the next route will deviate from the original one. For this reason, one also uses a representation of the surroundings, such as route maps, in order to compare sig- nals measured during autotramming with the map and thereby determine with more certitude the vehicle's position.

The route maps can consist, e.g., of a system of co-ordinates which can advantageously be local for the specific route and which can also be created on the basis of the recorded sensor signals. Thus, the system of co-ordinates need only include the area where the machine will be moving, and it can have its origin at the point on the machine which constitutes the reference for the positioning when the recording is begun, such as the middle of the machine's front axle.

The path driven by the machine during the route recording can then be described in this system of co-ordinates along with information on, e.g., appropriate speed for different parts of the stretch.

The representation of the surroundings (the route maps) should include information on which parts of the system of co-ordinates are broken up surfaces and which are rock. This can be generated beforehand, or by means of the transmitter information. The representation of the surroundings should consist of a represen- tation in at least two dimensions, where a distance between two points in the representation has a known relation to the corresponding distance in the said sur- rounding. An example is a metrical representation, i.e., a representation where a distance between two points in the representation is directly proportional to the corresponding distance in reality.

If the maps are generated during the route recording, the map can be assumed at the start, e.g., to consist entirely of rock, but then rock portions are later "erased" whenever the range scanners detect a free path. When the route reaches its end, a representation corresponding to that in Fig. 2 will have been generated, which can then be used in subsequent route playback (autotramming). Of course, other types of representations of the environment can be used, e.g., one where rock walls are represented by line segments.

For a route from point A to point B, as shown in Fig. 2, normally does not make use of a single map, as the figure indicates, but often the route maps consist of several consecutive map segments.

The use of several map segments instead of a single map has the advantage that if one or more sensors is giving faulty signals during the map generating or putting out signals with great uncertainty in the precision, this will have significantly less impact if the error is "zero reset" with shorter intervals than if a single map is used.

The maps/map segments used according to the present invention can be generated beforehand or advantageously are generated by means of the signals recorded dur- ing the route recording. Such map generating can be done in such a way that the map is represented by a relatively fine-meshed route network, for example, with a resolution of 1 cm or 1 dm per route, wherein at the start of the map generating the entire surface is marked as rock, but as the laser range scanners (or other range finders) detect free distance the routes on the map are marked as open space. Alternatively, the distances detected by the laser range scanners can be used to generate line segments representing walls, for example. Thus, maps of the type shown in Fig. 2 can be generated by means of transmitter data gathered during the route recording. Later, when the route, and perhaps maps have been generated, these can be used by the machine's control system to autonomously move the machine along the same stretch as the operator previously drove the machine manually.

During autonomous tramming, the system then plays back the route profile generated per above, wherein sensor data from the aforesaid sensors and map data is used by navigation algorithms to estimate the sideways and lengthways error of the route profile and that of the vehicle's speed, for example, with the same time intervals as used during route recording. The control system then corrects for errors, such as sideways displacement of the machine relative to the desired position so that the machine follows the determined route at the desired reference speed.

During playback, the machine's placement at the start of route recording is used as the assumed starting position, and depending on the extent of the map material and the computing power of the machine there may be limits placed on how much the machine can vary from its placement at the start of the route recording in order to still be capable of orienting itself and following the route. For example, one conceivable margin of error may be that the machine is within 10 meters of the starting position of the recorded route, while at the same time being oriented within 30°, for example, from the lengthways direction of the original position. As noted above, a statistical filter such as a Kalman filter can be used to estimate the position during playback. Preferably, one uses a so-called unscented Kalman filter, which uses a deterministic sampling technique.

Examples of transmitter signals which can be fed in to the Kalman filter are map data and signals from middle angle transmitters (assuming the machine is articulated, otherwise, or in addition, one can use, e.g., steering angle transmitters for steerable front and/or rear wheels) in order to measure the machine's centre angle, odometers to measure the distance travelled by the machine, and laser range scanners. In the present example, double scanners are used, i.e., scanning occurs both forward and backward at the same time, but an alternative sample embodiment uses only the scanner which happens to be "pointing" in the direction of travel. Yet another sample embodiment can use instead or in addition one or more scan- ners pointing toward the sides. After the above data has been fed into the filter, the filter then puts out an estimated position for the machine, in which a search algorithm can be used to determine which of the route points in the route profile comes closest to the estimated position of the vehicle, whereupon the desired po- sition is extracted from this route. The direction error and the sideways error of the vehicle in relation to the route are then determined from the estimated position and the desired position, wherein this data can be used to adjust the steering angles (and/or angle of articulation) and speed (e.g., in the form of gas pedal and gear shift) in suitable manner to get back on the recorded route.

One problem with this way of estimating the machine's position is that the margins of error of the different sensors are often indicated incorrectly or totally unknown. As explained above, even if mean error and covariance are known for the different sensors, one only gets a measure of the uncertainty in the estimated position. Thus, one gets no value for the reliability of the estimated position, i.e., no measure of how "good" the position estimate actually is. This also means that, if the machine is moving over a long stretch with no distinct landmarks such as intersections (e.g., section D-E in Fig. 2), the uncertainty can build up to an ever larger value so that the machine may ultimately stop because the uncertainty is so great that the machine no longer knows where it is. This is exemplified by Fig. 2, where a machine 201 at the indicated position can have so great an uncertainty in the estimated position that it must stop for safety reasons.

The present invention at least mitigates the above problem by calculating a measure for the reliability of the estimated position. A sample method 300 according to the present invention is shown in Fig. 3. The process starts with step 301, where data is read from laser range scanners, odometers and middle angle transmitters, per above, while in step 302 the position is estimated by means of the Kalman filter with inputs per above. Instead of jumping directly after the position estimation to step 306 to calculate the adjustment of the gas pedal, the steering angle and the gear shift per above, the process of the invention continues to step 303, where information from the laser range scanners is used to verify that the position estimated by the Kalman filter is correct.

This is done by comparing actual data from the laser range scanners at the esti-

RECTlFlED SHEET (RULE 91) mated position (e.g., the data supplied to the Kalnian filter during the position estimation, or data from the laser range scanners in the immediate vicinity of data that was used for the position estimation) with expected data from the scanners so as to get a measure of the reliability for the estimated position. The expected data from the laser distance scanners is calculated using the estimated position of the machine from the Kalman filter and the map(s) making up the route. For example, the calculation can be done by so-called beam tracking, that is, at the position estimated by the Kalman filter in a representation of the environment, in this case, the map, a simulated light beam's path is followed from a source (i.e., in this case, the scanner's estimated position on the map, which can differ from the machine's estimated position, since the machine's estimated position might be dictated by a determination of the middle of the front axle's position and not the laser's position, so that a correction has to be done for this case) until an obstacle halts its propagation.

In the present case, the obstacle halting the light beam is constituted by the tunnel walls on the maps and the result of a beam tracking of a single laser beam thus consists of the distance to the first wall on the map in the light beam's direction from the simulated light source's position (i.e., the position for the machine's laser range scanners when the machine is at the position estimated by the Kalman filter). This beam tracking can be done for any desired number of directions, e.g., it can be done for all the directions used in the route recording and/or autotram- ming, or alternatively fewer (or more) directions can be used. In the present example, one uses thirty of the 181 directions used per above, which means that a beam tracking is thus done for every sixth degree of the laser range scanner's an- gle interval.

At step 304, one then determines the difference between the expected distance as determined by beam tracking in the various directions and the corresponding distance actually measured from the laser range scanner. A measure of how well the estimated position agrees with the actual position can then be determined by comparing the expected value with the measured value, e.g., using the difference between expected and measured values. One way of calculating a measure of correctness of the position estimation is to compute the RMS (root mean square) for the difference of expected and measured distance in the various directions, i.e., the mean value of the square root for the difference between expected and measured distance in the different directions. This can be done by equation 1,

where a = measure of correctness of the position estimate, z = vector with distance data from laser scanner, r = vector with expected distance data from laser range scanner (i.e., beam tracking data) corresponding to the vector z, and n = number of distance values being evaluated (in this case, 30).

Instead of using the RMS, one can of course use other methods to determine a measure of correctness of the position estimate. For example, one can use the sum of the absolute magnitude of the differences as a measure, i.e.,

(2)

If the estimated position is correct, the measure determined at step 304 will be low, whereas an inaccurate estimate of the position will result in a higher meas- ure, unless sensor data from a reading at the inaccurately estimated position is identical with sensor data read at the actual position. Thus, the size of the measure will grow with increased error in the position estimation. In step 305, the measure computed in step 304 is then compared with a threshold value representing a maximum acceptable deviation between expected and measured distance. If the measure is less than the threshold value found, the process goes on to step 306, where appropriate modulation of the forward drive of the machine is computed per above to follow the desired route, and the process returns to step 301 to once again determine the measure as described at a later time, when updated sensor data is available. Thus, the present invention can largely reduce the problem of machines stopping as described in connection with Fig. 2, since the above described measure of uncertainty can be replaced or supplemented by the measure of the present invention. If, on the other hand, the measure determined for the estimated position's reliability exceeds the threshold value, the process goes on to step 307, where in one embodiment the machine is halted to avoid collision with the surrounding rock. In an alternative embodiment, however, the machine is not halted in step 307, but in- stead the machine's speed is reduced depending on the magnitude of the deviation of the measure from the threshold value. The greater the deviation, the more the speed is reduced. If the deviation then decreases, the speed can be increased again as the error diminishes.

In certain cases, even slight errors in the estimated position can give rise to large differences between expected distance and measured distance in certain directions, such as near corners/intersections or other conspicuous irregularities. If, then, only one determined measure is used for the above regulating process in step 307, this can result in needless stops/speed reductions. For this reason, it may therefore be advantageous not to base the judgement of the reliability of the local- isation only on a single measure, but rather on a series of consecutive measures. For example, one can use an intersection of the last x measures determined (e.g., 10, 20, 30, 50 or any other number) as a measure of the reliability. Another alternative is to use a filter, such as a minimum filter, where the lowest measure obtained during a certain time is returned. For example, the lowest measure of the last x measures (e.g., 10, 20, 30, 50 or any other desired number), or the measures determined during an interval of time, such as the last 2 seconds, or during a certain distance, can be returned from the minimum filter and used for comparison with the threshold value.

This also means that a slower (softer) speed adjustment (decrease) can be achieved as the error increases, if such is the case. On the other hand, if the error increases and then decreases, the speed can again be increased, thus entirely avoiding a stoppage.

The threshold value, for example, can be found experimentally or theoretically, and it can also depend on the type of sensors on board the vehicle, the reliability of the map, and the nature of the environment (such as the width of the site in relation to the vehicle's width, the number of intersections the vehicle will pass along the route, and so on). The present invention thus has the advantage over the prior art that one can obtain a reliable value of how correct is the estimation of the machine's position during autotramming regardless of whether the uncertainty of the transmitters is known or unknown, wherein the machine can be halted if the error becomes too large, and/or continue with reduced speed until the error of the estimate diminishes.

Claims

1. Method for calculating the reliability of an estimated position for a point on a mining and/or construction machine, wherein estimation of said position comprises an estimation based on at least a first parameter value from at least one sensor and a representation of the environment where said machine is located, wherein said machine's position in said environment can be estimated by means of said at least one first parameter value, wherein the method comprises

- calculating, from said estimated position in said representation of the environ- ment, at least one second expected parameter value,

- comparing said calculated at least one second expected parameter value with at least one actual parameter value for said machine's position in said environment, and

- calculating with the help of said comparison a measure of the reliability of said estimated position.

2. Method according to claim 1 , wherein said second expected parameter value corresponds to said at least one first parameter value, and wherein said actual parameter value at said position consists of said first parameter value.

3. Method according to claim 1 or 2, wherein the step of performing said calculation of the reliability of said estimated position comprises calculating a measure of reliability of said estimated position based on the difference between said at least one actual parameter value and said calculated at least one second expected parameter value.

4. Method according to any one of claims 1 or 3, wherein said parameter value consists of a representation of the machine's position in relation to the surroundings.

5. Method according to any one of claims 1-4, wherein said parameter value consists of the distance from said machine to surrounding obstacles in one or a plurality of directions seen from one or more points on said machine.

6. Method according to any one of the preceding claims, wherein said parameter value is determined from at least one sensor arranged on said machine.

7. Method according to any one of claims 1-6, wherein it further comprises the step of performing said estimation of the machine's position by use of a statis- tical filter, wherein said at least one first parameter value consists of input data to said filter.

8. Method according to claim 7, wherein said statistical filter is a Kalman filter, an enlarged Kalman filter or an unscented Kalman filter.

9. Method according to any one of the preceding claims, wherein during said estimation, said parameter value consists of data from at least two of the group: range finder, steering angle sensor, articulation angle sensor, wheel rotation sensor, information from said representation of the environment.

10. Method according to any one of the preceding claims, wherein a plurality of measures are determined for a plurality of consecutive time points and/or posi- tions, whereupon said measure of reliability is based on said plurality of measures.

11. Method according to claim 10, wherein said measure of reliability consists of the lowest value of said plurality of consecutive measures, or a mean value of said plurality of consecutive measures.

12. Device for calculating the reliability of an estimated position for a point on a mining and/or construction machine, wherein estimation of said position is designed to involve an estimation based on at least a first parameter value from at least one sensor and a representation of the environment where said machine is located, wherein said machine's position in said environment can be estimated by means of said at least one first parameter value, characterised in that said device for calculation of the reliability comprises means of

- calculating, from said estimated position in said representation of the environment, at least one second expected parameter value, - comparing said calculated at least one second expected parameter value with at least one actual parameter value for said machine's position in said environment, and

13. Device according to claim 12, wherein said first at least one actual parameter value consists of a set of distances running in one or more directions emerging from one or more points adjoining said machine to surrounding obstacles in said environment, and wherein said at least one second expected parameter value is a set of distances in corresponding directions from corresponding points to surrounding obstacles in the representation of the environment.

14. Device according to claim 13 or 14, wherein it is arranged to compute a reliability for one or a plurality of consecutive time points and/or positions, wherein the reliability of an estimated position is based on determinations for said plurality of time points and/or positions.

15. Device according to claim 14, wherein said reliability is designed to consist of the least value calculated for said plurality of consecutive time points and/or positions, or a mean value of calculations for said plurality of consecutive time points and/or positions.

16. Device according to any one of claims 2-15, wherein said mining and/or construction machine is a self-propelled machine.

17. Device according to any one of claims 12-16, wherein it furthermore comprises means of diminishing and/or halting the speed of said machine if said computed measure of reliability exceeds a first threshold value.

18. Device according to any one of claims 12-17, wherein said representation of the environment is a representation where a distance between two points in the representation has a known relation to the corresponding distance in reality.

19. A mining and/or construction machine, wherein it comprises a device according to any one of claims 12-18.

20. Use of a method according to any one of claims 1-11 for autonomous driving of a mining and/or construction machine to regulate the machine's speed in relation to a calculated value for the reliability of the machine's estimated position, in a representation of the environment where it is located, regardless of i whether the uncertainty of the component sensors is known.