Method for determining the position of an automatically moving device

By dividing the map into sub-areas and using calculated distance numbers to exclude irrelevant positions, the method enhances the efficiency and accuracy of self-localization for automatically movable devices, reducing computational effort and improving positioning precision.

DE102011000536B4Active Publication Date: 2024-12-05VORWERK & CO INTERHOLDING GMBH
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
DE102011000536
Authority / Receiving Office
DE · DE
Patent Type
Patents
Current Assignee / Owner
Filing Date
2011-02-07
Publication Date
2024-12-05
Estimated Expiration
2031-02-07

AI Technical Summary

Technical Problem

Existing methods for determining the position of automatically movable devices in known environments, such as cleaning or transport devices, are inefficient and computationally intensive, particularly when using obstacle detection and mapping systems.

Method used

The method involves dividing the environment map into sub-areas using a grid, calculating distance variables from obstacles for each sub-area, and using an algorithm to assign a number representing the average distance, which is compared with theoretically calculated numbers to narrow down possible positions, reducing computational effort through exclusion of irrelevant areas and utilizing SLAM for precise localization.

Benefits of technology

This approach significantly reduces computational load and improves positioning accuracy by limiting the number of positions to check, enabling efficient self-localization with reduced processing power requirements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0001_ABST
    Figure 00000000_0001_ABST
  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

Method for determining the position of an automatically moving device (1), in particular an automatically moving floor dust collector, with preferably electrically driven wheels (3), wherein the device (1) is equipped with an obstacle detection system (A) consisting of optical transmitter and receiver units, wherein the device (1) further comprises a map display (K) of the environment in which the device (1) is to be used, characterized in that the map display (K) is divided into sub-areas (10) by means of a grid (9) and that for each sub-area (10) over 360° the distances (distances a) or quantities derived on the basis of distance values ​​to obstacles (H) are calculated, insofar as these are located within a predetermined distance, and that from the calculated distances (distances a) or derived quantities an algorithm is used to calculate a number (11) that characterizes the sub-area (10).wherein an average value is calculated from the sum of the distances (a) and the theoretically calculated number (11) is assigned to the corresponding sub-area (10) and stored, and that furthermore the device (1) performs a circumferential scan using its obstacle detection (A) to determine its actual position at a location within the existing map display (K) and in the same way evaluates the distances measured using the same algorithm and compares the number thus found with the multitude of theoretically calculated numbers (11) and excludes sub-areas (10) that lie outside a predefined bandwidth as possible current positions of the device (1), and furthermore, starting from the remaining possible sub-areas (10), the sub-area (10) corresponding to the actual position is determined using standard self-localization (SLAM) methods.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] The invention relates to a method for determining the position of a self-propelled device, in particular a self-propelled soil dust collector, with preferably electric motor-driven travel wheels, wherein the device is provided with an obstacle detection system consisting of optical transmitter and receiver units, wherein the device further has a map representation of the environment in which the device is to be used.

[0002] Methods of the type in question are known, for example in connection with self-propelled vacuum and / or cleaning devices for cleaning floors, and furthermore, in a further embodiment, for example in connection with a self-propelled transport device or lawnmower. Such devices are preferably equipped with distance sensors, for example to counteract a collision with objects or the like in the path of travel. These sensors preferably operate without contact, and are further preferably light or ultrasonic sensors. For this purpose, it is also known to equip the device with means for all-round distance measurement, for example in the form of an optical triangulation system arranged on a platform rotating about a vertical axis or the like.Such a system can be used to measure distances based on reflections, which can be used for spatial orientation, particularly during automatic operation for obstacle detection, and preferably for creating a map of the space to be traversed and thus for creating a map. In this regard, reference is made, for example, to DE 2008 014 912 A1. The spatial boundaries recorded as a result of the distance measurement, possibly taking into account any obstacles in the rooms, are preferably stored in the form of a map of the surroundings, in particular of the apartment consisting of several rooms, preferably in a non-volatile memory of the device, so that this map can be used for orientation during a cleaning or transport process.Furthermore, it is known to determine a favorable movement strategy for the device based on such a stored map and further stored algorithms, even when an object is detected that is preferably located in the device's travel path and detected by the sensor. This requires the most precise possible determination of the distance value to the object, such as a piece of furniture or a wall boundary.

[0003] Furthermore, methods are known in the prior art in which direction-dependent distance values ​​are stored in the form of key figures in an environmental map. Reference is made to EP 2 256 574 A1 and US 2002 / 001 55 21 A1 in this regard.

[0004] In view of the prior art described above, a technical problem of the invention is seen in further improving a method, in particular to the extent that a favorable initial determination of the position of the device in a known environment, in particular one known from a map display, can be achieved.

[0005] This problem is initially and essentially solved by the subject matter of claim 1, which is based on the fact that the map display is divided into sub-areas by means of a grid and that for each sub-area over 360° the distances or quantities derived from distance values ​​to obstacles are calculated, provided that these are located within a predetermined distance, and that from the calculated distances or derived quantities a number characterising the sub-area is calculated using an algorithm, whereby an average value is calculated from the sum of the distances and the theoretically calculated number is assigned to the assigned sub-area and stored,and that, to determine its actual position at a location within the existing map display, the device performs a 360-degree scan using its obstacle detection system. It also evaluates the distances measured using the same algorithm, compares the resulting number with the multitude of theoretically calculated numbers, and excludes sub-areas that lie outside a predetermined range as possible current positions of the device. Based on the remaining possible sub-areas, the sub-area corresponding to the actual position is determined using standard self-localization algorithms (SLAM). This assumes that the self-propelled device has a map of its current surroundings in which, for example, open areas are shown in light, while obstacles are shown in dark. Alternatively, the contrast of the color values ​​can be increased.To increase accuracy, all areas below a certain color value, e.g., white, are highlighted, and only the darker color values ​​are displayed as obstacles in the map. A grid is placed over the device's known map, preferably with a predefined grid spacing, which divides the map into a number of sub-areas. From each sub-area, the distance to the nearest object is calculated at a certain number of angles, totaling 360° around the sub-area. Thus, during the all-round distance measurement, a respective distance calculation is preferably carried out in one to two degree increments. This calculation is theoretically based on the existing map, whereby in a preferred embodiment, the respective distance is calculated, for example, using the Bresenham algorithm. The maximum distance is calculated, even if the distance to the obstacle is longer.Limited to the sensor range of the obstacle detection. The range of the obstacle detection is known for calculating the theoretical distances. A sub-area number is then calculated, preferably from the sum of the calculated distances, for example, in the form of the average value of the distances, which number is then saved and assigned to the sub-area. Such a key number is calculated and stored for each sub-area of ​​the map. To ensure that the device's computing power is not overtaxed during normal movement, e.g., to clean a floor, a preferred embodiment provides for the calculation of the distances in the map, for example, while the device is in a base station. This preparatory measure is preferably performed only once per stored map. A recalculation of the distance values ​​or sub-area numbers occurs automatically, for example, as soon as a new map has been created.

[0006] If, for example, the user carries the device into a room known to the device through mapping and then places it there at random, the device will begin to determine its current position on the map either automatically or after a corresponding command has been issued. The device can use its obstacle detection sensors to display a section of its surroundings, recording all obstacles at a distance depending on the sensor range. The device itself, which finds itself in an unknown position in a known environment, records all distance values ​​around it for its localization by scanning the surroundings all around, using the same angles as used to calculate the distance values ​​for the sub-areas of the map. The currently determined all-round distance values ​​are also converted into a number (key figure), e.g.The average distance values ​​are calculated and compared with a multitude of theoretically calculated values. After comparing the determined number of the device's current position with the theoretically calculated numbers from the map, the device can significantly narrow down its possible position and, in the best case, even determine it directly.

[0007] In order to minimize further computational effort when the device's position within the map is still not clearly defined, the theoretically calculated numbers to be compared that lie outside a predefined range are excluded as possible current positions of the device. Accordingly, the further calculation of the device's current position is limited to those sub-areas of the map whose theoretically calculated numbers lie within one wall width based on the number found from the distance measurement. This process is repeated if necessary until a status is reached in which only a maximum of a predefined number of possible sub-areas are given or no further restriction appears possible. The remaining possible positions are further narrowed down preferably with the help of scan matching, as further described, in particular using SLAM, until the current position is found.In individual cases, the device may also move to incorporate additional information. SLAM is a method for simultaneous localization and mapping. For more information, please refer to the scientific article "Simultaneous Localization and Mapping (SLAM)" by Hugh Durrant-Whyte, Fellow, IEEE, and Tim Bailey, Part I, published in "Robotics & Automation Magazine, IEEE," June 2006. Since the effort required for such scan matching increases with the number of possible positions, and too many positions to be considered often produce no results or an inaccurate result, it is crucial to significantly limit the number of positions to be checked. This is achieved by excluding irrelevant areas as described above.Furthermore, the more angles around the device's own position and around the sub-area of ​​the scanned map are included, the more precise the localization can be. Since scan matching also reflects the device's orientation, 360 x 360 comparisons are required to calculate the location probability, for example, with 360 measurements per orbit. The proposed pre-constraint of the device's probable location by calculating distance sums is rotation-invariant and thus requires only one comparison per position. Overall, the proposed method achieves a significant reduction in computational effort compared to state-of-the-art calculation methods.

[0008] Further features of the invention are explained below, including in the description of the figures, often in their preferred association with the subject matter of claim 1 or with features of further claims. However, they may also be important in association with only individual features of claim 1 or the respective further claim, or each may be important independently.

[0009] The proposed all-round sensor for obstacle detection, which is already known from the prior art, emits a measuring beam. This determines the distance from the point of impact of the measuring beam to the nearest obstacle, for example a wall or a piece of furniture. The measuring method preferably used here is a time-of-flight (ToF) method. A phase correlation method (PKS) is also possible in this regard. A triangulation method is more preferably considered. The measuring beam can be designed as a line (divergence angle of the beam very small, less than 50 mrad, which typically leads to a beam diameter in the range of 1 to 22 mm) or as an extended beam. In this case, a beam with a divergence of 2 to 4 mrad is preferably used, which reaches a diameter of 20 to 30 mm at a distance of 3 m.The receiving area of ​​the sensor is preferably designed as a divergent receiving lobe in the same way, whereby its divergence is larger (e.g. by a factor of 2) than that of the transmit beam in order to compensate for tolerances.

[0010] The non-contact sensor for measuring the distance to an object is equipped with at least one optical light source that generates the measuring beam. The light source emits light in the visible or invisible wavelength range, e.g., red in the range of 650 nm or green in the range of 532 nm. Light in the infrared range with a wavelength of more than 700 nm is preferred, and beam sources (laser diodes or LEDs) with wavelengths of, for example, 785 nm, 850 nm, or 980 nm are more preferred. The light source can be a lamp or LED, but laser diodes are preferred.

[0011] Furthermore, the sensor is preferably equipped with at least one optical receiving element that is sensitive at least in the selected wavelength range of the transmitting element, more preferably embodied, for example, as at least one photodiode, photoresistor, CCD chip, or CMOS chip. The optical receiver can be designed as a single receiving element for detecting a single incoming light signal or as a multi-cell array or a multi-cell row for the simultaneous or sequential detection of multiple incoming light signals. Extended sensors are also known in this regard, for example, linear or planar PSD elements.

[0012] Optical elements can optionally be assigned to the light source and the light-sensitive element (preferably for forming a transmitted beam / transmission cone or received beam / receive cone), without these being explicitly listed below. Optical elements can be lenses (converging lenses, diverging lenses, also designed as Fresnel lenses, aspherical or spherical), apertures (circular, slit-shaped, of any shape, designed as separate parts or integrated into the sensor housing), prisms, mirrors (flat, concave, convex, freeform), light guides, or even free-form elements. Optical elements also include protective or cover screens in the optical beam path that serve to mechanically protect the optical device.Preferably, optical filters are provided associated with the light source and / or the light-sensitive element, which filters are as transparent as possible for the intended optical wavelength, but as little transparent as possible for others. Optical filters can be provided as separate elements or integrated into other optical elements (e.g. coloring or coating of lenses, protective windows, etc.). Other possible types of coloring, coating, vapor deposition, etc. are also possible. If mirrors are used, surface mirrors are preferred because they have lower optical losses. In this case, it is also advisable to design the mirror as an injection-molded plastic part and to provide the actual mirror surface as a surface coating with a reflective layer. The use of metallic layers, e.g. silver, gold or copper, more preferably aluminum, is recommended here.When using aluminum, a corrosion protection layer is preferably provided as an additional coating (passivation).

[0013] In a preferred embodiment, the device performs a movement before the calculation or after a first calculation in order to then compare the number then found with the multitude of theoretically calculated numbers once again. This can lead to a further exclusion of sub-areas that lie outside a predetermined bandwidth, so that in particular the computational effort can be reduced by means of conventional determination for self-localization. In this regard, it is further proposed that a movement of the device be carried out within the remaining sub-areas corresponding to the possible position of the device, more preferably until the remaining area comprising sub-areas representing possible positions is restricted to such an extent that the subsequent possible computational effort for conventional self-localization is kept to a minimum.

[0014] It is also proposed that the device move to a wall before or after the initial calculation and perform the comparison there. This further simplifies position determination. By moving the device to a wall, preferably to the nearest wall shown on the map, many positions of the device, and therefore irrelevant sub-areas, are already excluded from any subsequent calculation. Moving closer to the wall can also specify a recalculation of the position, which leads to a reduction in computational effort.

[0015] To further reduce computing power, a further development proposes that not all distance values ​​be calculated, but only maximum and minimum distance values. Thus, not all recorded values ​​are calculated around 360°, but only the extreme values ​​are given priority, which significantly reduces the number of values ​​to be considered and the required computing power.

[0016] The invention is explained in more detail below with reference to the accompanying drawing, which represents only one exemplary embodiment. It shows: Fig. 1 a self-propelled device in the form of a soil dust collector in perspective view; Fig. 2 a schematic floor plan representation of a space to be travelled by the device; Fig. 3 a schematic map representation of a room according to Fig. 2; Fig. 4 Area IV in Fig. 3 after rasterising the map into sub-areas and subsequently calculating the distance to obstacles in the map from each sub-area; Fig. 5 in schematic, opposite Fig. 4 enlarged view of sub-areas of the map with key figures assigned to each sub-area; Fig. 6 one of the Fig. 2 corresponding representation of the room floor plan with the device placed on the floor of the room; concerning the situation during an all-round distance measurement for position determination; Fig. 7 one of the Fig. 4 corresponding representation, whereby after the all-round distance measurement according to Fig. 6 non-position-relevant sub-areas are excluded from further position determination; Fig. 8 one of the Fig. 7 corresponding representation, after moving the device to a new position, for further restriction of the position-relevant sub-areas.

[0017] It is presented and described first with reference to Fig. 1, a device 1 in the form of a vacuum and / or sweeping device, further in the form of a self-propelled household vacuum robot. This has a chassis which, on the underside facing the floor 2 to be cleaned, carries electric motor-driven travel wheels 3 and preferably a brush that protrudes beyond the lower edge of the chassis floor and is also electric motor-driven. The chassis is covered by a device hood 4, with the device having a circular floor plan. Regarding the design of the device 1 as a vacuum and / or sweeping device, reference is made, for example, to DE 102 42 257 A1 cited above.

[0018] Furthermore, although not shown, the device 1 can have a suction port in addition to or as an alternative to the brush. In this case, an electrically operated suction fan motor is also arranged in the device 1.

[0019] The electrical supply of the individual electrical components of the device 1, such as the electric motor of the travel wheels 3, the electric drive of the brush, if applicable, the suction fan and, in addition, the further electronics provided in the device 1 for controlling the same, is provided via a rechargeable battery (not shown).

[0020] The device 1 is further provided with a distance / obstacle detection device A in the form of a contactless sensor 5, comprising a light beam emitting device and a light beam receiving device. This sensor 5 is arranged on top of the device cover 4 of the device 1 and is rotatable about a vertical axis d, which simultaneously represents the central vertical axis of the device 1. The sensor 5 preferably consists of a triangulation system, by means of which an all-round distance measurement (over 360° around the axis d, arrow r in Fig. 1) can be carried out.

[0021] With the help of the sensor 5, obstacle detection is initially achieved, this as a result of rotating scanning in a preferably horizontal scanning plane, i.e. in a plane running parallel to the floor 2, so that the device 1 can move on the floor 2 or in its surroundings without collision. Furthermore, as is further preferred, an all-round distance measurement of the surroundings is possible via the sensor 5, wherein the distance values ​​determined here to obstacles and walls in the surroundings are preferably used to create a map representation K of the entire area, which map representation K is saved and stored in the device 1.

[0022] In Fig. 2 schematically shows the floor plan of a room R to be driven through by the device 1, for example for floor cleaning. According to the all-round distance / obstacle detection A, a map display K according to Fig. 3 is created, whereby obstacles such as room boundaries 6 and projections 7 as well as sections of furniture 8 located in the scanning plane are reflected as obstacles in the map display K. In the map display K, free areas are preferably shown in light, while obstacles H are shown in dark.

[0023] Preferably, before a first free travel of the device 1 in the space R, the map display K is first divided by means of a grid 9 into equally sized sub-areas 10, preferably into square sub-areas with an edge length of 20 to 60 mm, more preferably 40 mm. The sub-areas 10 are then preferably initially assigned numbers, e.g., the number 255 (light) if the sub-area 10 is free, the number 0 (dark) if the sub-area is occupied by an obstacle H, and further with numbers between 0 and 255 for various gray levels.

[0024] Starting from each sub-area 10, in particular free sub-area 10, a map section K is first created. A defined, whereby the respective edge length is preferably adapted to the detection range of the obstacle detection A. In order for each sub-area 10 to be rotationally invariant, i.e. independent of the current orientation of the device 1, the corners of the square map sections K A darkened or hidden, so that a rotation-invariant map section K KA with the assigned sub-area 10 as the means (cf. Fig. 4).

[0025] Starting from each sub-area 10, the distance a to the nearest obstacle H within the map circular section K is measured in a given number of angles, preferably 360 angles, a total of 360° around the sub-area 10. KAcalculated, preferably using the Bresenham algorithm. The maximum distance a is accordingly limited by the map area K KA equal to the radius of the circular segment.

[0026] The average value is calculated from the sum of the distances a (preferably 360 distance measurements around the sub-area 10) and the theoretically calculated number 11 is assigned to the assigned sub-area 10 and saved. This distance calculation is carried out for each sub-area 10 of the map display K ( Fig. 5 shows schematically the allocation of several sub-areas 10 with associated theoretically calculated numbers 11).

[0027] If the device 1, which has the map display K containing the sub-areas 10 assigned with key numbers, is placed on the floor 2 of the room R, for example by being carried, the device 1 must first orient itself and accordingly first determine its current position within the map display K. This is done according to the proposed method according to an all-round distance measurement using the device-side obstacle detection A while the device is at a standstill. Using the obstacle detection A, the respective distance to obstacles (e.g. room boundaries 6 and / or furniture 8) is determined all around, i.e. over 360° around the axis of rotation d in the same number of angles as in the theoretical all-round distance measurement, this more preferably also taking into account the preferably specified maximum range of the distance sensor.

[0028] The distance values ​​to obstacles determined in this way are calculated using the same algorithm, preferably the Bresenham algorithm. Finally, the average value is calculated from the sum of the distances, and the resulting value is compared with the multitude of theoretically calculated values ​​11 in the map representation K.

[0029] Partial areas 10 whose number 11 does not match the number found as a result of the all-round distance measurement using obstacle detection A (possibly taking into account a predefined bandwidth) are excluded from further positioning unless a partial area 10 is clearly identified. This is shown schematically in Fig. 7 by the hatched sections 10. The non-hatched sections 10 represent possible positions of the device 1, with further Fig. 7 where x represents the current position of device 1.

[0030] If the remaining range (restricted sub-ranges 10) of possible positions of the device 1 is within a predetermined bandwidth, the actual position of the device 1 is determined by means of a conventional determination for self-localization, preferably using the SLAM method.

[0031] To further narrow down the possible device positions and thus to further minimize computing power in the possible subsequent SLAM process, the device 1 can be positioned according to the schematic representation in Fig. 8 within the range of sub-areas 10 of possible device positions in order to carry out another all-round distance measurement in a second device position x1 using the obstacle detection A, after which the determined distance values ​​are calculated to an average value and compared with the numbers 11 of the remaining sub-areas 10. In the illustrated embodiment, this leads to a further restriction of the possible device positions (bright sub-areas 10 in Fig. 8).

[0032] In the preferred subsequent SLAM method for the preferred exact determination of the device position, scattered particles that are outside the possible device position (outside the bright partial areas 10 in the Fig. 7 or Fig. 8) are hidden for further calculation. Alternatively, in the non-relevant sub-areas 10 (hatched sub-areas 10 in the Fig. 7 or Fig. 8) no particles scattered.

[0033] In order to further reduce the computing power both in determining the partial area numbers 11 and in the all-round distance measurement using the obstacle detection A and in the subsequent comparison of the numbers, not all all-round distance values ​​(e.g. 360 distance values) are processed, but only extremes, so on, for example, only maximum distance values ​​a max and minimum distance values ​​a min (compare Fig. 4). List of reference symbols: 1 device 2 floor 3 Bicycle 4 Device cover 5 Sensor 6 Room boundaries 7 lead 8 furniture 9 grids 10 sub-area 11 number a distance a max maximum distance a min minimum distance d axis of rotation r arrow x Device position x1 second device position A obstacle detection H Obstacle K Map display K A Map section K KA Map circular section R Room

Claims

[1] Method for determining the position of a self-propelled device (1), in particular a self-propelled ground dust collector, with preferably electric motor-driven travel wheels (3), wherein the device (1) is provided with an obstacle detection system (A) consisting of optical transmitter and receiver units, wherein the device (1) further has a map representation (K) of the environment in which the device (1) is to be used, characterized bythat the map display (K) is divided into sub-areas (10) by means of a grid (9) and that for each sub-area (10) the distances (distances a) or quantities derived from distance values ​​to obstacles (H) are calculated over 360°, provided that these are located within a predetermined distance, and that from the calculated distances (distances a) or derived quantities, a number (11) characterizing the sub-area (10) is calculated using an algorithm, whereby an average value is calculated from the sum of the distances (a) and the number (11) thus theoretically calculated is assigned to the assigned sub-area (10) and stored,and that the device (1) further carries out an all-round scan by means of its obstacle detection (A) to determine its actual position at a location within the existing map display (K) and in the same way evaluates the distances measured thereby using the same algorithm and compares the number thus found with the plurality of theoretically calculated numbers (11) and excludes sub-areas (10) that lie outside a predetermined bandwidth as possible current positions of the device (1), wherein further starting from the remaining possible sub-areas (10) the sub-area (10) corresponding to the actual position is determined by means of conventional determination for self-localization (SLAM). [2] Method according to claim 1, characterized bythat the device (1) carries out a movement before the calculation or after a first calculation in order to then compare the number then found with the plurality of theoretically calculated numbers (11) once again. [3] Method according to one of the preceding claims, characterized by that the device (1) moves to a wall before the calculation or after a first calculation and carries out the comparison there. [4] Method according to one of the preceding claims, characterized by that not all distance values ​​(distances a) are calculated but only maximum distance values ​​(maximum distance a max ) and minimum distance values ​​(minimum distance a min ).

Citation Information

Patent Citations

  • Self-propelled floor dust collector

    DE102008014912A1

  • Self-propelled floor dust collection device, as well as a combination of such a collection device and a base station

    DE10242257A1

  • Autonomous mobile robot, self-position estimation method, environment map generation method, environment map generating device, and environment map data structure

    EP2256574A1

  • Apparatus and method for recognizing self-position in robot system

    US20020015521A1