JP2001510890A - Method for detecting the rotational state of an autonomous mobile unit and autonomous mobile unit - Google Patents

Abstract

(57) SUMMARY According to the present invention, a method is described for detecting the rotational state of an autonomous mobile unit, such as a small robot performing work in a household or business. The unit is provided with a gyroscope and a distance measuring device for detecting the first and second rotation state changes. These changes in rotational state are compared with each other and the gyroscope is calibrated while the unit is running if the measurement results are confirmed with each other. If the measurement results are not mutually confirmed, it is assumed that an odometry measurement error exists and that the gyroscope is used as a more reliable measuring instrument in a short time. In this manner, a method is presented for the first time to calibrate the gyroscope during travel of the autonomous mobile unit based on the odometry data. The disadvantages of gyroscope or odometry measurements are avoided. This is because accurate data is used as the case may be based on the threshold value for the comparison result. FIG.

Description

DETAILED DESCRIPTION OF THE INVENTION

The present invention relates to a method and an apparatus for detecting the rotational state of a moving mobile system, such as used in an autonomous mobile robot or other autonomous vehicles.

In autonomous mobile systems, there is generally the problem that the system must obtain an image of its surroundings while the system is running. For this purpose, various sensors that sense measurement data from the surroundings are used. A particularly advantageous method for detecting distance measurement data is Odometrie / odometry. This odometry is known from DE 33 15 422 A1. For this purpose, for example, an encoder at the wheel axis is used to measure the wheel rotation. Odometry data obtained by this internal sensor then continuously estimates, for example, the position of this unit. A disadvantage of this form of combined navigation is that position estimation based on odometry data is error-prone. In these errors, a distinction can be made between systematic and non-systematic errors. Systematic errors include, for example, non-uniform wheel diameters or incorrectly measured wheel spacing. However, these systematic errors can be completely eliminated by sophisticated calibration. Thus, only non-systematic errors contribute to degrading the estimated position. Such errors may be due, for example, to wheel slippage on the ground or overcoming unknown objects on the ground such as cables, door sills or tile seams. Rotational state changes can be detected based on odometry data from various distances measured at different wheels of the autonomous mobile unit.

[0003] Another method of detecting a change in rotational state is to use a gyroscope located in an autonomous mobile unit. Using this gyroscope or other acceleration sensor, changes in the rotational state of the autonomous mobile unit are relatively detected. The absolute rotation state is then determined by one or two integrations of the raw data. A disadvantage of this system is that the sensor signal must be integrated to obtain the desired measurand. In this case, the smallest constant error (konstante Fehler) of each of the sensor data results in a drift of the measurand. Since this error increases in a time-dependent manner, these systems are only available for a short time, for example of the order of minutes. After this, an orientation pause must be performed to remove the system offset due to temperature fluctuations and other environmental influences and to calibrate the system. “Fang PD
., Hung JC: Gyrocompassing on a Moving Land Vehicle, Proceedings of t
he 15th Southeastern Symposium on System Theory, 28.-29.March 1988, Hunt
From the prior art of "Sville AL, USA" it is known to calibrate this with a Kalman filter using a gyroscope in an autonomous mobile unit. However, in this method, offset and amplification factors are required for calibration of the gyroscope. In addition to quasi-static parameters such as, dynamic variables such as vehicle orientation are also estimated.

[0004] It is known from the prior art DE 3831166 A1 to use for a short time a further orientation system for the correction of an odometry system used for navigation of autonomous mobile units.

It is known from the prior art DE 3910945 A1 to use two redundant combined navigation systems for the navigation of autonomous mobile units,
Switching is controlled between these two systems. As a redundancy system DE 39 10 945 A1 mentions a combination of an odometry system and a gyroscope system. According to DE 39 10 945 A1, the control of the switching between these two systems is performed as follows. That is, the actual position of the autonomous mobile unit determined by the odometry system and the actual position of the autonomous mobile unit determined by the gyroscope system are calculated from the running course stored in the computer. Is done by comparison with Depending on the results of the two comparisons, a system with a relatively small deviation of each actual value from the target value is used for the navigation of this autonomous mobile unit.

It is an object of the present invention to provide a further method and device for detecting the rotational state of a traveling autonomous mobile unit.

[0007] The object is achieved by a method according to claims 1 to 9 and an autonomous mobile unit according to claim 10.

[0008] Embodiments of the invention result from the dependent claims.

A particular advantage of the method according to the invention is that a change in the rotational state of the autonomous mobile unit is detected by two different measuring means, each of which provides an instantly reliable measurement result for the rotational state detection. Used for

[0010] Advantageously, in the method according to the invention, the distance measurement is performed using odometry. This is because this method is used in a large number of autonomous units and can use advantageous measurement sensors.

[0011] Particularly advantageously, according to the method of the invention, the rotational state is detected as a rotational state measured by a gyroscope or as a rotational state determined from odometry data.

[0012] Particularly advantageously, according to the method of the invention, it is distinguished which of the two measurement systems delivers an instantly reliable result, advantageously using an experimentally determined threshold, This measuring system is used for detecting the state of rotation.

It is particularly advantageous to generate a difference between the absolute values from the two rotational states for the purpose of forming the comparison result. This is because this difference is formed at a low cost in computer technology, and the effect of the positive and negative prefixes is eliminated.

Particularly advantageously, according to the method of the invention, the gyroscope is calibrated while the autonomous mobile unit is running, if the odometry data is regarded as instantaneously reliable. This is determined based on the threshold.

Advantageously, according to the method of the invention, a Kalman filter is used for the estimation of the gyroscope offset. This filter is used in many ways, and there are well-known methods for implementing estimators for time-dependent quantities.

Particularly advantageously, according to the method of the invention, an error counter is implemented. This error counter is counted up depending on a threshold value, and is used for detecting a sensor error after exceeding a preset error limit value.

Particular preference is given to an autonomous mobile unit whose rotational state is detected according to the method of the invention, comprising a measuring means and a gyroscope for detecting distance measurement data. This makes it possible for the first time to calibrate the gyroscope on the fly.

Next, the present invention will be further described with reference to the drawings.

FIG. 1 shows a motion model of the autonomous mobile unit AE. By way of example, the configuration of the mobile system (Konfiguration) is shown in the form of position and rotation in a fixed reference frame. These position indications are determined, for example, relatively to a coordinate system that is running continuously. For example, this arrangement in a periodically performed sampling step k

[0020]

(Equation 1)

[0021] consists of the (x, y) position and orientation of the autonomous mobile unit in the global coordinate system K: (x, y). Of particular importance here is the determination of the rotational state of the vehicle. This is because, even if small, the orientation error causes an ever-increasing translational error during continuous running. In the present invention, this orientation error is reduced by the combined evaluation of the odometry measurements by the sensors OM1 and OM2 at the wheels R1 and R2 and the measurement results from the gyroscope Gy.

For the purpose of positioning and kinetic modeling, one starts here with an autonomous system having, for example, two driven rear wheels R 1 and R 2. Kinematically, this type of drive corresponds to the well-known tricycle movement as found in cars and many mobile robots. Shown is a global coordinate system K: (x, y)
Is a motion model of such a vehicle.

Small changes in position and rotation are estimated by the following equations:

[0024]

(Equation 2)

Where W _s is half the wheel _span, and U _lk and U _rk indicate the change in motion at both wheels. The index k indicates a discrete state. In the relative position determination, sensor information about the external environment is advantageously not used for position estimation. Odometry has the following advantages: 1. There is no offset to detect with drift.

2. A simple incremental generator.

3. Very robust to environmental effects such as temperature, barometric pressure and air humidity.

4. Coordinates and orientation can be determined.

The disadvantages of odometry are: 1. Wheel slip remains unmodelled.

2. Wheel rotation based on overcoming door sills or cable or tile seams leads to erroneous estimates of the state of rotation.

FIG. 2 illustrates, by way of example, one of these disadvantages, the degradation of the quality of the distance measurement results.

FIG. 2 shows by way of example how a wheel gets over a small obstacle H. This shows the effect of non-systematic errors on localization. In FIG. 2, a wheel R having a radius r has just begun to overcome an unknown obstacle H having a height h. Ideally, for example, it is assumed that the wheel R moves over the contact point C precisely and without slippage. In this case, the wheel center point M rotates through the contact point C until it reaches the position M 'exactly above the contact point C. During this movement, the wheel encoder advantageously measures the wheel rotation ψ. This wheel rotation is advanced traveling section

[0033]

[Outside 1]

Is interpreted as However, the actual horizontal movement of the wheel in the movement direction Bew is d _hor .

If only one wheel of the unit runs over this obstacle, the distance change Δd

[0036]

(Equation 3)

Is determined. The factor of 2 comes from the fact that the wheel has to climb on this obstacle and get off the obstacle again.

For a wheel radius r and an obstacle height h, the path difference is

[0039]

(Equation 4)

Is determined by

If only one wheel of a tricycle kinematics robot gets over this obstacle, the change in the distance of the wheel Δd will result in an angular error Δθ during position detection.

[0042]

(Equation 5)

Causes the following. Here, W _s is half of the wheelbase.

Examples of calculations based on the dimensions of the experimental robot will clarify these effects.

[0045]

(Equation 6)

If a wheel with a radius r = 6 cm gets over a h = 1 cm threshold, an angle error Δθ = 0.75 ° results. Due to this angle error, for example, in the next straight running of dist = 10 m

[0047]

(Equation 7)

Causes a translation error ΔT of approximately 13 cm. This angle error is a very serious error. This is because the position error resulting from this error increases indefinitely with further driving.

Therefore, in order to improve the position detection and the rotation state detection of the autonomous mobile unit,
Advantageously, a second measuring means in the form of a gyroscope or accelerometer without such errors is used, for example. Thereby, the rotation of the robot is relatively determined. One or two integrations of the measured raw data determine the state of rotation. A disadvantage of this system is that the sensor signal must be integrated to obtain the desired measurand. In this case, each minimum constant error in the sensor data results in the required measurand drift.

Therefore, errors in detecting the rotation state increase endlessly with time. Therefore, if only these systems are utilized, these systems guarantee accurate detection of orientation for only a few minutes.

The gyroscope sends, for example, a voltage U _out to the output side. This voltage U _out is proportional to the angular velocity ω. If this voltage is sampled periodically every T _G , the rotation state is

[0052]

(Equation 8)

Is determined by Euler-Vorwaerts-Integration. Here, K is a proportional coefficient depending on a gyroscope having K = 22.2 mV / deg / sec (10).

A major problem in cheap gyroscopes that appears to be particularly suitable for consumer applications such as domestic robots or mail distributors is offset. However, this offset mainly depends on temperature, air humidity, switch-on time and other factors. In the case of a commercially available gyroscope, the offset changes by almost 100 mV due to a temperature change of 3K. this is

[0055]

(Equation 9)

Means a continuous error of This continuous error is caused only by the offset shift.
Therefore, the offset should be re-calibrated with experimental findings, preferably at a standstill every minute when using only a gyroscope for orientation. In this case, Offset = U _out (12) holds.

If the robot has to perform, for example, a transport mission, the robot is not allowed to stop every minute. Therefore, we propose a method to calibrate / check the offset while the robot is running.

An advantage of the gyroscope lies in its slip-independent detection of its rotational state. The disadvantages of the gyroscope are: 1. The rotational speed is detected, and this rotational speed must be integrated in order to detect the rotational state. Thus, the state of rotation is very strongly dependent on the exact determination of the zero.

2. Zeros are subject to very large drifts that are further dependent on changing environmental influences. Therefore, the zeros must be constantly recalibrated. The conventional solution is that the robot stops for 2 seconds every minute.

3. Only the rotation state can be detected.

The present invention advantageously exploits the advantages of odometry and gyroscope measurements without having to accept the disadvantages of odometry and gyroscope measurements.

The present invention ensures the combined use of odometry and a gyroscope for rotational state detection of a mobile system without the orientation pause required for offset compensation in the gyroscope. The purpose is the sampling time (k +
In 1), the angle change Δθ _{Odometrie / Gyroskop} is detected, and the rotational state of the mobile unit is thereby changed to θ _{k + 1} = θ _k + Δθ _{Odometrie / Gyroskop} (13).
). For this purpose, the following method is advantageously used.

A) Odometry (index Odo) and gyroscope (index Gy
ro) is used to detect an angle change at time (k + 1).

[0064]

(Equation 10)

B) Compare both angle changes according to the following scheme.

[0066]

[Table 1]

C) Calibration of gyroscope offset If the difference between the gyroscope angular velocity value and the odometry angular velocity value is smaller than, for example, an experimentally determined threshold, both the odometry measurement and the gyroscope measurement have errors. Not assumed. In addition, the vehicle can be
It is unlikely that odometry has been negatively impacted since it has been driven over obstacles such as those illustrated in FIG. Therefore, odometry is used for detecting the rotation state.

The gyroscope measurements and odometry values are advantageously used in the present invention for the calibration / calibration check of the gyroscope offset during the travel of the mobile unit. FIG. 3 shows an example for estimating a gyroscope offset by a filter.

As shown in FIG. 3, a Kalman filter K is used for determining a gyroscope offset. Since the gyroscope offset to be estimated varies very much with time, iterative or recursive estimators are used for calibration. The estimator should advantageously be able to take into account system instability and measurement noise, as the measurement data is additionally strongly noisy (eg due to A / D conversion).

The use of the Kalman filter K is advantageous in such a case. The use of this filter will now be described.

However, it should be noted here that, in principle, other filters, and possibly even fuzzy rules or neural networks, may be suitable for the method of the invention. The prior art has already reported calibration using a Kalman filter. In contrast to estimating dynamic variables such as vehicle orientation in addition to quasi-static parameters such as offset and gain, the present invention provides gyroscope sensors such as offset and gain, especially offset. Only the parameters have to be optimized. This is because the offset mainly affects the rotation state detection. Here, the detection of the vehicle orientation can also be integrated into this filter. This will, of course, make the case distinction implemented below difficult.

The parameters that characterize the gyroscope as states for the Kalman filter, especially

[0073]

[Equation 11]

Is used.

State vector

[0076]

(Equation 12)

Thus, the system equation becomes as follows.

[0078]

(Equation 13)

Here, it is the system matrix A (E = unit matrix).

Depending on whether the angular velocity measurement M provided by the gyroscope is greater or less than the instantaneous zero, the angular velocity is estimated via the Kalman filter measurement equation as follows:

[0081]

[Equation 14]

This is then preferably the angular velocity determined from the odometry data

[0083]

[Outside 2]

And the difference is used for the estimation of the gyroscope parameters. A Kalman filter for gyroscope parameter estimation is constructed from these equations and the associated covariance matrix.

With this method, it is possible to calibrate the gyroscope so that the gyroscope orientation estimate matches the odometry orientation estimate in the normal case when there is no obstruction. If an odometry fault is detected, the vehicle orientation is reliably determined by the gyroscope.

D) Test: Sensor Error The methods described so far have always assumed a correctly functioning sensor.
Of course, sensor failures can always occur. For this reason, in the method of the invention, the odometry and the gyroscope are advantageously monitored for their data match.

FIG. 5 shows an example of a flowchart for detecting a sensor error based on an error counter. If the sensor measurement does not match continuously and repeatedly above F_max, then either the gyroscope or the odometry has a high validity error. Advantageously, in this case the robot interrupts its mission and stops.

In summary, the following advantages of the method of the invention are obtained: Combining the advantages of odometry and gyroscopes.

Enables “on-line” calibration of the gyroscope while driving.

For the first time, meaningful use of inexpensive gyroscopes (with unstable offsets) is possible.

• Enable identification of sensor failures.

The relationships and important parameters of the method OV of the present invention are illustrated as a block diagram in FIG.

[Brief description of the drawings]

FIG. 1 exemplarily shows units in a reference coordinate system.

FIG. 2 illustrates non-systematic errors in odometry measurements.

FIG. 3 shows an example of a Kalman filter.

FIG. 4 shows a block diagram as an example of the method of the invention.

FIG. 5 illustrates a method for detecting a sensor error based on a threshold and an error counter.

[Explanation of symbols]

 AE Autonomous mobile unit OM1 sensor OM2 sensor Gy Gyroscope R1 wheel R2 wheel K Kalman filter

──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Wolfgang Renken, Germany Munich Zing Tauer Strasse 86 A F-term (reference) 2F029 AA08 AB01 AC01 AC04 AD03 2F105 AA02 AA06 AA10 BB04 BB17 5H301 AA01 BB14 GG12 GG17 9A001 JJ17

Claims (10)

[Claims] 1. A method for detecting the rotational state of an autonomous mobile unit, comprising: a) a first movement of the unit using at least a distance measurement at a first point and a second point of the unit. A first travel section in which the first location is advanced and a second travel zone in which the second location is advanced, in the form of first and second distance measurement data, Determining a first rotational state change of said unit from a known position of said location as well as a known trigonometric function; b) using a gyroscope coupled to said unit, depending on a first movement of said unit, Measuring the rotational state change of the unit in the form of rotational state measurement data and determining it as a second rotational state change of the unit; c) comparing the first and second rotational state changes with each other; Comparison Forming a result; d) detecting a rotational state of the autonomous mobile unit depending on the comparison result;
A method for detecting the rotational state of an autonomous mobile unit. 2. The method of claim 1, wherein the distance measurement is performed using odometry. 3. The method according to claim 1, wherein the rotation state is detected as a first rotation state. 4. The method according to claim 1, wherein the rotation state is detected as a second rotation state. 5. The comparison result is compared with a threshold value, and the rotation state is detected depending on whether the threshold value is greater than, less than or equal to the comparison result.
The method according to one of the preceding claims. 6. The first rotation state change or the first and second distance measurement data and the second rotation state change or the rotation state measurement data are converted into numerical values, and the first rotation state is obtained as a comparison result. 6. The method according to claim 1, further comprising forming an absolute value of a difference between the change and the second rotational state change. 7. The gyroscope according to claim 2, wherein the gyroscope has an offset and the offset is calibrated during running of the unit using the first rotational state change. Method. 8. The method of claim 7, wherein the Kalman filter is used as an iterative estimator of the offset for calibration, and the first and second rotational state changes are supplied to the iterative estimator.
The described method. 9. An error counter value is counted depending on whether it is above or below a threshold value. If the error counter value is above or below an error limit value defined according to the counting method, a measurement error is detected. 9. The method according to one of claims 5 to 8, which is detected. 10. An autonomous mobile unit having rotation state detection, comprising: a) the autonomous mobile unit comprises means for measuring a distance of a first movement of the unit at a first location and a second location. The means includes a first travel section where the first location is advanced and a second travel zone where the second location is advanced, as first and second travel sections.
B) said autonomous mobile unit has a gyroscope for measuring a second change in rotational state; c) said autonomous mobile unit comprises said first and second autonomous mobile units. E) an evaluation means for calculating a first rotation angle state change from the distance measurement data using a known trigonometric function; and d) the autonomous mobile unit has the first rotation state change and the second rotation state. E) the autonomous mobile unit further comprises detection means for detecting a rotation state of the autonomous mobile unit depending on a result of the comparison, e. Autonomous mobile unit with detection.