WO2016018170A1

WO2016018170A1 - Localizing a movable object using an inertial measurement system

Info

Publication number: WO2016018170A1
Application number: PCT/RU2014/000573
Authority: WO
Inventors: Artem Vladimirovich OZHIGIN
Original assignee: Siemens Aktiengesellschaft
Priority date: 2014-07-30
Filing date: 2014-07-30
Publication date: 2016-02-04

Abstract

The invention is concerned with localizing a mobile device, e.g. a vehicle or a mobile phone. The invention provides a method comprising the steps of receiving a motion signal (S) from an inertial measurement unit (12) of the device (10) and generating a measured motion pattern (PAT) based on the motion signal (S). The object of the invention is to provide a precise localization. To this end, the inventive method comprises the further step of providing a motion pattern set (38) comprising at least one combination (44) of a predefined motion pattern (PAT1 to PAT6) and location data describing a location (XYZ1 to XYZ6); and the step of finding in the motion pattern set (38) a predefined motion pattern (PAT1) that matches the measured motion pattern (PAT) according to a predefined matching criterion; and the step of providing the location data associated with the matching predefined motion pattern (PAT1).

Description

LOCALIZING A MOVABLE OBJECT USING AN INER I L MEASUREMENT

SYSTEM

DESCRIPTION

The invention relates to a method for localizing a movable object, like e.g. a vehicle or a mobile phone. The localization is performed on the basis of a motion signal as it is provided by an inertial measurement unit that may comprise e.g. an accelerometer and/or a gyroscope. The invention also provides a mobile device designed for performing the inventive method.

Localizing a movable device, i.e. determining the current device location or device position, may be performed on the basis of a measurement of the acceleration and/or changes in spatial orientation. In order to obtain an estimate of the current device location, the information about acceleration of the device must be summed up or mathematically integrated twice to obtain first velocity and then position. Thus, over time small measurement errors that come with each measurement of acceleration or angular rate may accumulate and result in a progressive drift of the estimate with regard to the real device location.

Locating or localizing movable devices on the basis of a motion signal from an inertial measurement unit is especially attractive in cases, where global satellite- based positioning systems do not operate, as it is the case inside buildings due to signal attenuation by building constructions. Here, dedicated systems based on different wireless technologies using electromagnetic waves (radio, infrared, optical) may be deployed in the building to provide location services. However, they require a specific hardware .

Although inertial measurement units based on accelerometers and/or gyroscopes are available in general purpose hardware like mobile phones or tablet PCs, it is difficult to use their motion signal for localizing a movable device due to the integration drift caused by continuous accumulation of measurement errors resulting from the numerical integration of angular rate and accelerations.

It is the objective underlying the current invention to provide means for localizing a movable device on the basis of cheap and readily available hardware. The objective is reached by the sub ect-matter of the independent claims. Embodiments of the invention providing further advantages are given by the features of the dependent claims. According to the inventive method the localization of the mobile or movable device or object is based on an inertial measurement unit that is provided with the device. The method comprises the step of receiving the motion signal from the inertial measurement unit. The inertial measurement unit generates its motion signal in dependence on motions of the device which may be a velocity and/or an acceleration and/or an angular velocity or angular rate. The inventive method further comprises the step of generating a measured motion pattern based on the motion signal. In other words, the motion signal is used to estimate several device locations or positions over time and combine those estimates to give a motion path that forms the motion pattern. The inventive method is characterized by the following steps: At least one predefined motion pattern is provided, i.e. a hypothetical or theoretical motion pattern that might be generated by the movable device. Together with each predefined motion pattern, location data are provided that describe a known location. All combinations of a predefined motion pattern and the corresponding location data together form a motion pattern set. It is now possible to compare the measured motion pattern describing the real motion path that the movable device has moved along, with all the predefined motion patterns comprised in the motion pattern set. In other words, the method comprises the step of finding in the motion pattern set a predefined motion pattern that matches the measured motion pattern according to a predefined matching criterion. The matching criterion may provide a predefined flexibility or tolerance in order to identify a matching predefined motion pattern, although the patterns do not match exactly. Once a matching predefined motion pattern is found the location data associated with that matching predefined motion pattern may be used as a description of the current device location. Thus, a description of the device location is provided that is not only based on the motion signal itself, but also on the predefined locations described by the location data stored in the motion pattern set.

The inventive method provides a significant increase in accuracy of localizing a movable device on the basis of a motion signal generated by an inertial measurement unit. A further advantage is that the method works without additional infrastructure. It may rely exclusively on the inertial measurement unit and a processing unit that provides the motion pattern set and performs the search or comparison of the measured motion pattern with the predefined motion patterns in the motion pattern set. This makes it possible to perform the localization based on the hardware of general purpose devices which are already on the market, thus opening the application of localization to a wide range of technical devices like mobile phones, portable and hand-held PCs, robots and/or vehicles. Also, low-cost and simple hardware may be provided for embedded positioning devices for manned and robotic vehicles.

In one embodiment of the inventive method the motion signal comprises the sensor signal of at least one accelerometer and/or at least one gyroscope. Preferably the inertial measurement unit has at least three accelerometers to measure acceleration along vertical, longitudinal and latitudinal axis. Alternatively or additionally, the inertial measurement unit may comprise up to three gyroscopes to measure angular rates. The use of sensor signals of accelerometers and/or gyroscopes provides the advantage that the sensors can be completely enclosed in a casing and do not depend on the measurement of external signals. In contrast to this, directly measuring, e.g. a velocity of the device, requires measuring a movement of an element, e.g. a wheel. Such sensors are sensitive to deterioration, erosion or dirt. Accelerometers and gyroscopes may be designed much more robust with regard to deterioration.

Localizing the movable device solely on the described pattern matching method may not be sufficient in cases where for a longer period of time no pattern may be recognized. For such cases one embodiment of the method comprises the step of estimating a current device location in dependence on the motion signal itself. In other words, the known method of estimating the current device location by integration may also be comprised in the inventive method. Additionally, this embodiment further comprises the step of calculating a respective distance to each location described by the location data in the motion pattern set. In other words, it is determined how far the current estimated location is away from any of the predefined known locations that are described by the location data. The term "distance" may relate to the direct geometrical distance. The distance may also consider obstacles like walls, i.e. the distance may be a path length calculated along possible motion paths that may be followed by the movable device. Based on the estimated current device location and the calculated distances to the known predefined locations, those motion pattern/location data combinations from the motion pattern set are selected that comprise location data of a location in a predefined vicinity of the current device location. The vicinity may be defined as a maximum distance value. In other words, only those combinations of predefined motion pattern/location data are considered that belong to a location whose relative distance to the current device location is smaller than the maximum distance value. Further, for matching the measured motion pattern with the predefined motion patterns only the selected combinations are used. This allows to speed up the process of finding the matching predefined motion pattern. In other words, a complex and/or resource-demanding method for performing the pattern-matching may be employed.

Another embodiment of the invention comprises repeating the two steps of finding the matching predefined motion pattern and providing the location data in periodic intervals and/or at predefined points in time. This ensures a regular re-calibration of the localization. The repetition may also be triggered in response to a trigger signal of a sensor or of an observation unit. A trigger sensor may be, e.g., a touch screen of a mobile phone such that the localization is re-calibrated whenever the user of the mobile phone begins using his phone. An observation unit may be, e.g., a central computer for observing the locations of several movable devices, like robots in a warehouse, wherein the central computer performs an update of a list of the locations of each movable device.

One embodiment of the invention is concerned with finding suitable predefined motion patterns. The embodiment comprises the step of providing a map of a predefined area, e.g., a warehouse or an office. In general, the map comprises data of possible motion paths of the device in the area. Such data may comprise a description of the building construction, e.g., the positions of walls and other hindering objects, and/or the course of free corridors. The embodiment further comprises a step of defining the at least one combination of a predefined motion pattern and location data based on the data of the map. Such characteristic patterns may include narrow turns, long straight passages, a passage through partitions via, e.g., doors .

In the case that the motion signal also provides information regarding an upward and/or downward motion, one embodiment of the inventive method comprises the step of providing the predefined motion pattern for at least one of the following motion paths: vertical ascend and/or vertical descend, especially of an elevator, escalator, a ladder, staircases. These motion paths provide especially characteristic motion patterns that may be easily distinguished from horizontal motions.

In order to provide for an effective pattern matching, one embodiment of the method comprises the step of finding the matching predefined motion pattern on the basis of a clustering method and/or a Bayesian network, especially a hidden Markov model. For performing a clustering method, the coordinates describing the measured motion pattern may be aligned to form one vector that describes a point in a multi-dimensional mathematical space. This mathematical space may be sub-divided into regions, wherein each region represents one predefined motion pattern. A Bayesian network, especially a hidden Markov model, may be trained by means of measurement data generated from recorded motion signals. With regard to a Bayesian network, the matching criterion may be a predefined minimum value for the confidence, i.e. a minimum likelihood value that must be reached or topped.

The inventive method may also be flexibly combined with additional sensors. One embodiment comprises the step of receiving a sensor signal of at least one sensor outside the inertial measurement unit, especially a light sensitive sensor and/or a sensor of velocity and/or a clock. Finding the matching motion pattern in the motion pattern set is then also based on the respective additional sensor signal. The additional sensor signal of the additional sensor may be used to verify whether a certain match is plausible. For example, if an ascend of the mobile movable device is detected on the basis of the motion signal of the inertial measurement unit, the movable device may be inside an elevator moving upwards. However, if - at the same time - the additional sensor signals a forward motion it is very unlikely that the movable device is inside the cabin of an elevator. Using a light sensitive sensor may, e.g., provide the information whether the movable device is inside or outside a certain place like a building or a pocket. On the basis of a clock signal, it may be verified whether it is very likely that the movable device is within a certain location. If the location is not accessible a certain time, like, e.g., a supermarket during night, this location may be ruled out .

The invention also provides a mobile device comprising an inertial measurement unit of the type already described. The mobile device also provides a processing unit coupled to the inertial measurement unit. The inventive mobile device is characterized in that the processing unit is designed to receive the motion signal of the inertial measurement unit and perform an embodiment of the inventive method .

According to different embodiments of the invention, the mobile device may be a robot, a fork lift, a mobile phone, a hand-held PC, a vehicle, like, e.g., a car, or a wearable device, like a bracelet or a jacket.

A further aspect of the invention is concerned with a tangible data storage medium. The inventive storage medium comprises a program code executable by a processing unit of a mobile device, wherein the program code is designed to perform an embodiment of the inventive method, when executed by the processing unit. The tangible data storage medium according to the invention provides the advantage that existing mobile devices, like mobile phones or handheld PCs or robots, may be equipped with program code to perform an embodiment of the inventive method.

In the following, a preferred embodiment of the invention is explained in more detail. The accompanying drawing illustrates the embodiment. The figure shows a mobile device 10 comprising an inertial measurement unit 12, a location estimator 14, and a correction unit 16. The mobile device can be a vehicle, like a ground conveyor or a fork lift, an autonomously moving robot, a hand-held device, like a smartphone or a hand-held PC, or a wearable device, like a jacket.

The inertial measurement unit 12 can comprise accelerometers 18, 20, 22, wherein the accelerometers 18, 20, 22 may be configured to measure acceleration along different axes: The accelerometer 18 may be configured to measure along a vertical axis, the accelerometer 20 may be configured to measure along a longitudinal axis, and the accelerometer 22 may be configured to measure along a latitudinal axis. The inertial measurement unit may comprise more than or less than the shown three accelerometers 18, 20, 22. Additionally or alternatively the inertial measurement unit may comprise gyroscopes 24, 26, 28 to measure angular rates in different planes. The inertial measurement unit may comprise more than or less than the shown three gyroscopes 24, 26, 28. Each of the sensors 18 to 28 may be provided as a electronic device, e.g. an MEMS (micro electro-mechanical system). The sensor signals of the accelerometers 18, 20, 22 and/or the gyroscopes 24, 26, 28 may together form a motion signal S that is provided to the localization module 14.

The localization module 14 may be designed to estimate a current device location XYZ on the basis of the motion signal S. The localization module 14 may be a program module executed by a processing unit of the mobile device 12. The estimation of the current device location XYZ may be based on a double integration in order to estimate a current velocity VXYZ based on the first accumulation or integration 30 and then to estimate the location XYZ based on a second accumulation or integration 32.

The mobile device 10 may be part of an application foreseen in a warehouse management, a guided tour, a mall, an airport, or a store. For navigating or localizing the mobile device 10, the indoor positioning or localization may be performed wirelessly on the basis of the estimate of the device location XYZ.

However, the estimated device location XYZ may comprise a progressive drift that can guickly cause inacceptable positioning error. The device 10 does not have this problem. The estimated device location XYZ may be corrected by the correction unit 16 to give a corrected device location XYZcorr. The correction unit 16 may be another program module executed by the processing unit of the mobile device 10. The correction unit 16 may comprise a pattern generator 34, a map 36, a motion pattern set 38, and a pattern matcher 14.

The pattern generator 34 may receive several successive estimated device locations XYZ from the location estimator 14 and may combine these estimates to generate a motion path 42 describing the path along which the mobile device 10 has moved within a predefined amount of time, wherein the amount of time may be in the interval from one second to one hour, for example. The motion path 42 or a section or part of it may be transmitted to the pattern matcher 40 as a measured motion pattern PAT.

The motion pattern set 38 may comprise a list of pairs or combinations 44, wherein each combination 44 comprises a predefined motion pattern PAT1 to PAT6 together with corresponding or associated location data describing a known predefined location XYZl to XYZ6. The predefined motion patterns PAT1 to PAT6 and/or the locations XYZl to XYZ6 may be derived from map 36. The predefined motion patterns PAT1 to PAT6 may describe the following motion paths (in this order) : a double turn, a loop, a motion along a long passage, an ascend, a descend, a U-turn. In the figure, corresponding symbols are shown. The predefined location XYZl to XYZ6 is visualized in the corresponding pattern PAT1 to PAT 6 by a cross x.

The map 36 may describe an area in which the mobile device 10 is to be localized. In the example underlying the figure the area may comprise two floors FLOOR0, FLOOR1 of a building. The two floors FLOOR0 and FLOOR1 may be connected by an escalator 46. The map 36 may also comprise data describing walls 48 of the buildings. From the map 38 possible motion patterns could be derived as the predefined motion patterns PAT1 to PAT6.

The matching unit 40 may be designed to compare the measured motion pattern PAT with one or more or all of the predefined motion patterns PAT1 to PAT6 of the pattern motion set 38. The comparison may be based on, e.g., a clustering method and/or a Bayesian estimator method, like a hidden Markov model.

In order to limit the amount of predefined motion patterns PAT1 to PAT6, which need to be compared with the measured motion pattern PAT, the matching unit 40 may first select a sub-set from the pattern motion set 38. This may be performed by first checking where in the map 38 the estimated device location XYZ is positioned. In the visualized example the estimated device location XYZ points to a position on FLOOR1. The matching unit 40 may then find all those predefined locations XYZl to XYZ6 that lie within^' a predefined vicinity 50. The vicinity 50 may be defined as a circle with radius 52 around the estimated device location XYZ. In the example only the predefined locations XYZl and XYZ2 lie within the vicinity 50. The matching unit 40 may therefore select only the combinations 44 comprising the two predefined XYZl and XYZ2 for the pattern matching.

As is visualized in the figure, the motion path 44, i.e. the measured motion pattern PAT matches the predefined pattern PAT1. The matching unit 40 may therefore output the predefined location XYZl as the corrected device location XYZcorr .

The predefined patterns PAT1 to PAT6 may also comprise additional data, for example velocity data describing the motion velocity that was observed while the mobile device 10 moved along the motion path forming the corresponding motion patterns PAT1 to PAT6. For example, the estimated velocity VXYZ may be used or a measured velocity provided by an additional sensor.

The corrected device location XYZcorr may be used to provide localization information of the mobile device 10, i.e. an object like a robot or a vehicle, or a person carrying the mobile device 10.

The localization estimator 14 and the correction unit 16 are preferably provided in connection with an indoor positioning system (IPS) . Such a system then does not require any external infrastructure and can be realized as a system embedded into vehicle (a fork lift or another transport vehicle) or a person vehicle, or as a hand-held device or wearable device or even purely software solution in a general purpose device like a mobile phone or a handheld PC of any kind. The main disadvantage of such systems in the prior art is the accumulation of error due to the integration of motion parameters measured by sensors (mainly accelerations and angular rates) .

The described periodic correction of such errors by the correction unit 16 can significantly improve the performance of the system. For the error correction some additional sensors can be used, if they are present in the device. The main feature of the invention is, however, that for the purpose of error correction the measurement from the same sensors of the inertial measurement unit 12 can be applied. In order to implement automatic error correction, the indoor positioning system shall preferably have a spatial map 36 of the building. In this map 36 there shall be introduced characteristic motion patterns for certain locations XYZ1 to XYZ2 within the building. Every time the system detects an actual motion pattern PAT matching one pattern PAT1 to PAT6 stored in the map 36 in some vicinity 50 of the current calculated location XYZ, it can automatically compensate accumulated error by immediate fixation to the proper map location XYZ1 to XYZ6. Such characteristic patterns may include narrow turns, long straight passages, vertical ascends and descends in elevators, escalators, ladders and staircases; a passage through partitions via doors or similar. The example illustrates how the invention provides an inertial indoor positioning system with error correction.

Claims

1. Method for localizing a mobile device (10), comprising the steps of:

- receiving a motion signal (S) from an inertial measurement unit (12) of the device (10), wherein the inertial measurement unit (12) generates the motion signal (S) in dependence on motions of the device (10);

- generating a measured motion pattern (PAT) based on the motion signal (S);

characterized by the steps of:

- providing a motion pattern set (38) comprising at least one combination (44) of a predefined motion pattern (PAT1 to PAT6) and location data describing a location (XYZ1 to XYZ6) ;

finding in the motion pattern set (38) a predefined motion pattern (PAT1) that matches the measured motion pattern (PAT) according to a predefined matching criterion;

- providing the location data associated with the matching predefined motion pattern (PAT1).

2. Method according to claim 1, wherein the motion signal (S) comprises the sensor signal of at least one accelerometer (18, 20, 22) and/or at least one gyroscope (24, 26, 28) .

3. Method according to any of the preceding claims, comprising the steps of

- estimating a current device location (XYZ) in dependence on the motion signal (S),

- calculating a respective distance to each location (XYZ1 to XYZ6) described by location data of the motion pattern set (38); - selecting those combinations (44) from the motion pattern set (38) that each comprise location data of a location (XYZ1, XYZ2) in a predefined vicinity (50) of the current device location (XYZ) ;

- using only the selected combinations (44) for finding the matching predefined motion pattern (PAT1).

4. Method according to any of the preceding claims, comprising repeating the steps of finding the matching predefined motion pattern (PAT1) and providing the location data in periodic intervals and/or at predefined points in time and/or in response to a trigger signal of a sensor or of a observation unit.

5. Method according to any of the preceding claims, comprising the step of providing a map (36) of a predefined area (FLOOR1, FLOOR2), the map (36) comprising data of possible motion paths of the device (10) in the area (FL00R1, FLOOR2) , and defining the at least one combination (44) of a predefined motion pattern (PAT1 to PAT 6) and location data based on the data of the map (36) .

6. Method according to any of the preceding claims, comprising the step of providing the predefined motion pattern (PAT1 to PAT6) for at least one of the following motion paths: turn (PAT6), straight passage (PAT3), passage through a partition (PAT1).

7. Method according to any of the preceding claims, comprising the step of providing the predefined motion pattern (PAT1 to PAT6) for at least one of the following motion paths: vertical ascend ( PAT4 ) and/or vertical descend (PAT5), especially of an elevator, an escalator, a ladder, staircases.

8. Method according to any of the preceding claims, comprising the step of finding the matching predefined motion pattern (PAT1) is performed on the basis of a clustering method and/or a Bayesian network, especially a Hidden Markov model.

9. Method according to any of the preceding claims, comprising the step of receiving a respective sensor signal of at least one sensor outside the inertial measurement unit, especially a light sensitive sensor and/or a sensor of velocity and/or a clock, wherein finding the matching motion pattern is also based on the respective sensor signal.

10. Mobile device (10) comprising an inertial measurement unit (12) and a processing unit coupled to the inertial measurement unit, wherein the inertial measurement unit (12) is designed to generate a motion signal (S) in dependence on motions of the device (10),

characterized in that

the processing unit is designed to receive the motion signal (S) and perform a method according to any of the preceding claims.

11. Mobile device (10) according to claim 10, wherein the device (10) is a robot, a fork lift, a mobile phone, a handheld PC, a vehicle or a wearable device.

12. Tangible data storage medium comprising a program code executable by a processing unit of a mobile device (10) , wherein the program code is designed to perform a method according to any of the claims 1 to 9 when executed by the processing unit.