Computer-implemented method of creating an environment map for running a mobile agent

Abstract

The invention relates to a computer-implemented method for operating a mobile agent (1) in an environment based on a pose of the mobile agent (1), wherein for operating the mobile agent (1) a localization of the mobile agent (1) is performed with the following steps: - detecting (S1) sensor data about walls (2) and / or objects (3) located in the environment, wherein the sensor data describes an alignment and a distance of the walls (2) and / or objects (3) in an agent coordinate system (A) fixed relative to the agent; - determining (S11-S18) a pose of the mobile agent (1) in the environment by means of a SLAM algorithm, taking into account a Manhattan orientation.

Description

Technical Field

[0001] This invention relates to the field of autonomously controlled mobile agents within a movement area. In particular, this invention relates to a method for locating mobile agents in an environment. Background Technology

[0002] Mobile agents are known for many applications. Here, a mobile agent is required to move independently in an environment equipped with boundaries (walls) and obstacles (objects). Such mobile agents can include, for example, robots, vehicles, autonomous cleaning robots, autonomous lawnmowers, and the like.

[0003] The following capabilities form the basis for trajectory planning of motion paths within an environment defined by sensor data: determining the accurate pose (position and orientation) of the mobile agent in that environment, and determining the environmental map associated with the accurate pose.

[0004] The conventional approach to creating environmental maps involves using environmental sensing devices on a mobile agent. These devices may include, for example, lidar systems, cameras, inertial sensors, distance sensors, and the like. To create the environmental map, the mobile agent typically moves in a controlled or automatic manner and evaluates sensor data from the environmental sensing devices to create the map.

[0005] Traditional methods for creating environmental maps involve applying a so-called SLAM algorithm, which processes data recorded by environmental detection sensors to determine the pose of a mobile agent and, consequently, the corresponding environmental map. Summary of the Invention

[0006] According to the present invention, a method for locating a mobile intelligent agent within an environment is provided, as well as a device for locating a mobile intelligent agent, a device for controlling a mobile intelligent agent in an environment, and a mobile intelligent agent are provided.

[0007] Other construction options are also described below.

[0008] According to the first aspect, a method is provided for operating a mobile agent based on the pose of the mobile agent in an environment, wherein the localization of the mobile agent is performed using the following steps:

[0009] - Detect sensor data about walls and / or objects in the environment, wherein the sensor data describes the alignment and distance of the walls and / or objects in a fixed agent coordinate system relative to the agent;

[0010] - Taking Manhattan orientation into account, the pose of a mobile agent in the environment is determined using the SLAM algorithm.

[0011] A fundamental drawback of SLAM algorithms used for locating mobile agents—that is, SLAM algorithms used to determine the absolute attitude of mobile agents within an environment—lies in the cumulative drift of the detected sensor data. In particular, due to coupling, detection errors regarding the determined orientation cause significant drift in the coupled position of the mobile agent. While orientation drift can be compensated for by direct measurement alignment, such direct measurement alignment is typically only possible via compass measurement, which is unavailable in many situations, especially in buildings or in environments heavily loaded by stray magnetic fields.

[0012] The above approach combines SLAM algorithms with the Manhattan World hypothesis. The Manhattan World concept posits that many artificial environments possess regular structures, particularly with walls that are parallel or perpendicular to each other, and that the walls typically extend in straight lines over long stretches and are flat. For example, the Manhattan World concept is known in the publication "Accurate On-Line 3D Occupancy Grids Using Manhattan World Constraints" by B. Beasley et al. (2012, IEEE / RSJ International Conference on Intelligent Robots and Systems, Vilamoura, pp. 5283-5290).

[0013] Furthermore, the SLAM algorithm can be configured to minimize an error function related to the measurement error of the sensor data, which is related to Manhattan orientation.

[0014] According to one embodiment, the SLAM algorithm may correspond to a graph-based SLAM algorithm based on a SLAM graph having nodes and edges, wherein the SLAM graph has attitude nodes that respectively describe the attitudes of the mobile agent determined from sensor data, and has transformation edges between each pair of attitude nodes, wherein the transformation edges describe the attitude changes determined from sensor data between the attitudes associated with each pair of attitude nodes, wherein Manhattan nodes and at least one Manhattan edge between the Manhattan node and one of the attitude nodes are added according to Manhattan orientation.

[0015] Furthermore, the SLAM algorithm can use an error function determined from the SLAM graph, taking into account the error probabilities associated with nodes and edges, to determine the corrected pose associated with nodes in the SLAM graph.

[0016] In principle, this method sets forth an improvement upon traditional SLAM algorithms for 2D environments. Traditional SLAM algorithms can be constructed as graph-based SLAM algorithms, and a graph is created as follows: the graph has pose nodes and transformation edges between each pair of pose nodes. A pose node represents the pose detected by the mobile agent through an environment detection sensor, and a transformation edge represents the measured pose change between two different poses of the mobile agent. The pose is described with respect to a reference coordinate system, which is fixed in position relative to the environment.

[0017] Due to noise or other measurement errors, it is impossible to precisely decompose these graphs such that the pose described by the pose nodes of the mobile agent matches the measured motion described by the mobile agent through the transformed edges. Instead, the determined pose and motion are described using error probabilities, and the error function, which is summed over all transformed edges, is minimized to determine the exact pose of the mobile agent. In particular, minimization is performed using conventional optimization methods, such as Gauss-Newton or Levenberg-Marquardt.

[0018] Based on the Manhattan world concept, additional nodes and edges can be added to the graph. For each determination of a valid Manhattan orientation in the detection step, Manhattan nodes and Manhattan edges between one of the pose nodes and the involved Manhattan nodes can be inserted into the SLAM graph. Then, the subsequent optimization algorithm for minimizing the error uses additional boundary conditions, pre-given by the Manhattan orientation, to determine not only the pose of each pose node. The reference coordinate system can then be aligned to the Manhattan orientation.

[0019] The method described above enables the detection of Manhattan orientation in each detection cycle in order to determine Manhattan orientation as an angle in the range from -45° to +45°.

[0020] In particular, the method described above does not require reference to previous detection cycles and can also be applied to environments that only partially correspond to the Manhattan world. By explicitly mapping the Manhattan orientation, it is possible to align the reference coordinate system to the Manhattan orientation, thus providing an environment map aligned with the reference coordinate system. By continuously re-evaluating sensor data, the method used to find pairs with previous detection cycles can be discarded.

[0021] In addition, a set of points can be determined from sensor data, which describes the coordinates of the detected walls and / or objects, wherein the points in the set are combined into local cells that are adjacent to each other, wherein Manhattan normalization is applied to cells that have a cell orientation with a straight line or a 90° structure, wherein the Manhattan orientation is determined based on the dominant orientation of the straight line structure in the local cell.

[0022] In particular, in order to determine the dominant orientation of these units, the similarity between each unit orientation of a unit and all other unit orientations can be determined, and a similarity metric (e.g., as L2 spacing) can be determined for each unit, in particular as the sum of the similarities of the corresponding unit with respect to the remaining units, where the Manhattan orientation corresponds to the unit orientation of one of these units selected according to the similarity metric, in particular the similarity metric of that unit is the largest.

[0023] Furthermore, similarity can be determined based on the Manhattan-normalized orientation difference between each pair of unit orientations in the unit orientation, particularly based on an exponential function of the negative Manhattan-normalized difference.

[0024] It can be configured such that the determined Manhattan orientation is only considered if the orientation of more than a predetermined number of these cells is within the difference between the Manhattan-normalized orientation and the determined Manhattan orientation.

[0025] In one implementation, an environmental map can be determined using the current pose of the mobile agent, which then operates based on this map. For example, the environmental map and the mobile agent's pose can be used for trajectory planning in a trajectory planning problem.

[0026] According to other aspects, a device, particularly a control unit, is provided for operating a mobile intelligent agent based on the posture of the mobile intelligent agent in an environment, wherein the device is configured to perform localization of the mobile intelligent agent using the following steps:

[0027] - Detect sensor data about walls and / or objects in the environment, wherein the sensor data describes the alignment and distance of the walls and / or objects in a fixed agent coordinate system relative to the agent;

[0028] - Taking Manhattan orientation into account, the pose of a mobile agent in the environment is determined using the SLAM algorithm.

[0029] According to other aspects, a mobile intelligent agent is provided, which has the above-mentioned equipment, an environmental detection sensing device for providing sensor data, and a motion actuation device configured to move the mobile intelligent agent. Attached Figure Description

[0030] The implementation details are then described in more detail with reference to the accompanying drawings.

[0031] Figure 1 A schematic diagram of a mobile intelligent agent within the environment to be detected is shown;

[0032] Figure 2 A flowchart illustrating a method for determining the posture of a mobile agent within an environment is shown;

[0033] Figure 3 The following diagram illustrates this: the graph is created within the scope of the SLAM algorithm to determine the pose of the mobile agent;

[0034] Figure 4 A flowchart illustrating the determination of Manhattan orientation is shown;

[0035] Figure 5a A schematic diagram illustrating the detection of an environment map using a traditional SLAM algorithm is shown, and

[0036] Figure 5b An illustration shows the detected environment map based on the SLAM algorithm, taking into account the Manhattan world concept. Detailed Implementation

[0037] Figure 1 A schematic diagram of a mobile agent 1 is shown in an environment defined using walls 2 and objects 3. This environment corresponds to a typical environment, such as that commonly found in indoor rooms, factory workshops, or urban areas, and has walls and / or object edges that extend parallel to each other and / or perpendicular to each other.

[0038] The mobile agent 1 has a control unit 11 for controlling the mobile agent 1. Furthermore, the mobile agent 1 has an environmental detection sensor 12 for detecting the orientation and spacing of the surrounding walls and objects 2 and 3 with respect to the agent's coordinate system A. Additionally, the mobile agent 1 has a motion actuator 13 for moving the mobile agent 1 within the environment according to the function to be performed.

[0039] The environmental sensing device 12 may include at least one of the following sensor devices: a lidar device, a radar device, a laser scanning device, a camera, an inertial sensing device, a distance sensing device, and the like. Based on preprocessing depending on the sensor device used, sensor data is processed in the control unit 11 so that the alignment and distance of walls or objects 2 and 3 can be determined and provided with respect to the agent coordinate system A. To determine the environmental map, the attitude of the mobile agent 1 relative to a fixed reference coordinate system K is determined. Then, the attitude of the mobile agent 1 is obtained from the positional offset between the agent coordinate system A and the reference coordinate system K, and from the rotational difference between the agent coordinate system and the reference coordinate system.

[0040] From the pose of the mobile agent 1 in the reference coordinate system, the corrected positions of the wall 2 and the object 3 in the agent coordinate system can then be determined based on sensor data.

[0041] To create an environment map illustrating the positions of wall 2 and object 3 in the reference coordinate system, and to determine the current pose of mobile agent 1, the control unit 11 implements a method combining... Figure 2 The flowchart illustrates the method in more detail. This method can be implemented as a software and / or hardware algorithm in a control device. The method is described below with reference to a two-dimensional environment.

[0042] In step S1, sensor data is detected in the detection step using the environmental detection sensing device 12. This sensor data describes the orientation (distance and orientation relative to the proxy coordinate system) of one or more walls and objects 2, 3 from the perspective of the mobile agent 1. The sensor data point-by-point describes the measured coordinates of the distance to the wall 2 and obstacle / object 3 in the corresponding orientation, allowing the relative position of the involved wall or object 3 with respect to the mobile agent 1 to be determined from the alignment angle and distance (relative to the proxy coordinate system). The sensor data can be determined, for example, using lidar detection and distance measurement.

[0043] In step S2, a reference coordinate system is selected or updated. The reference coordinate system is selected based on a parameterizable standard. Therefore, the reference coordinate system is updated when movement of the mobile agent has occurred. By selecting a reference coordinate system, sensor data can be associated with a fixed position. Preferably, the reference coordinate system is updated only when movement of the mobile agent has occurred.

[0044] Initially, the reference coordinate system can be defined using the starting pose of mobile agent 1. When the pose change of mobile agent 1 relative to the current, last valid reference coordinate system exceeds a threshold (e.g., 2m or 60 degrees), a new reference coordinate system is selected. The updated reference coordinate system is then placed within the current pose of mobile agent 1.

[0045] Subsequently, a SLAM graph is created or expanded, as exemplified in [the following text is missing from the original] Figure 3 As shown in the diagram, a SLAM graph is created to determine the optimal pose for each pose node through optimization.

[0046] Therefore, in step S3, after selecting a reference coordinate system, an attitude node P (ellipse) is added to the SLAM graph. The attitude node P represents the measured attitude of the mobile agent 1 relative to the selected reference coordinate system. The measured attitude is determined by the attitude of the mobile agent 1 in the agent coordinate system and the transformation between the agent coordinate system and the reference coordinate system (the position and orientation of the origin of the agent coordinate system). The transformation can be, for example, through a rotation matrix.

[0047]

[0048] and position offset The pose of mobile agent 1 is defined by the following vector: (That is, 3 values), among which .

[0049] Besides the attitude node P, the SLAM graph has transformation edges TK. Transformation edges TK (blocks) are added in step S4 and connect the two attitude nodes i and j respectively, defining the attitude. and And can be obtained through vectors To illustrate, the following applies to rotation R and translation T:

[0050] ,in

[0051] ,in

[0052] or and

[0053] A transformation edge TK is added to the SLAM graph. The transformation edge TK can be obtained from distance measurements and the identification of the already initiated attitude. The transformation edge TK shows the measured relative transformation between two attitude nodes P, and in particular, it shows the measured relative transformation in the SLAM graph between the current attitude node P and one or more previously reached attitude nodes.

[0054] Furthermore, when identifying the previous angefahrenen position of the mobile agent 1, a loop edge S (a loop-closure edge) can be inserted between two corresponding attitude nodes P, with the same position and / or orientation associated with the two corresponding attitude nodes P. The loop edge S can be described as a transformation corresponding to a position change of 0 and a rotation change of 0°.

[0055] Such loops can be identified using a multi-level approach. In the first step, nodes are identified as potential candidates for loop edges based on their estimated pose, for example, when the spacing between nodes does not exceed a threshold (e.g., 1 meter). In the second step, the optimal transformation among the potential candidate nodes is determined using a scan matching method. In the third step, the mutual consistency of all potential candidate nodes identified so far is checked, for example, using spectral clustering. Only when the check is successful is the potential candidate node inserted into the graph as a loop edge. Exemplary implementations of this method are described in detail, for example, in Edwin B. Olson's (2008) "Robust and Efficient Robotic Mapping" (Ph. D. Dissertation, Massachusetts Institute of Technology, Cambridge, MA, USA).

[0056] In step S5, it is checked whether Manhattan orientation can be detected. To do this, the following steps are implemented: Figure 4 The flowchart then illustrates the method in more detail. Based on the probe of the Manhattan orientation, Manhattan nodes MK and Manhattan edges M are added to the SLAM graph.

[0057] In step S6, optimization of the SLAM graph is performed, as described, for example, in Gristetti, G. et al., “A Tutorial on Graph-Based SLAM” (IEEE Intelligent Transportation Systems Magazine, 2.4 (2010), pp. 31–43). Performing optimization is necessary because the sensor data contains noise or measurement errors, making it impossible to find a definitive solution for the SLAM graph. Instead, an error function is minimized, which corresponds to a summation function over all edges, where, for example, for each edge i, j, the error function is defined as… ,in , The poses corresponding to pose nodes i and j, and The motion measurement corresponds to the transformation edge ij. The total error function corresponds to...

[0058] ,

[0059] in Corresponding to measurement The inverse of the covariance of the association.

[0060] Traditional optimization algorithms, such as Gauss-Newton or Levonberg-Marquardt, are used to minimize the error function. The result of the optimization method is the corrected pose of the mobile agent 1 for the pose nodes of the graph. Therefore, the current pose of the mobile agent 1 in the reference coordinate system can be determined from the last added pose node P, and an environment map can be created based on sensor data if necessary. Now, based on the environment map and the current pose of the mobile agent 1, trajectory planning can be performed in a known manner.

[0061] exist Figure 4 The middle section is a flowchart used to illustrate the integration of the Manhattan world concept into the previously described method for locating the mobile agent 1 in the reference coordinate system K and the location of the environment map.

[0062] In step S11, it is checked whether Manhattan orientation exists. For this purpose, the sensor data is grouped into point clouds of the detection locations of walls 2 or objects 3 in the cells. For example, these cells may have a pre-given size of, for example, 1m × 1m or the like. For all such cells containing at least three detection points, the mean and covariance matrix of these points are calculated. This is based on methods known per se and is described, for example, as the basis for so-called NDT-based scan matching methods in Biber, W. et al., “The normal distributions transform: A new approach to laser scan matching” (Proceedings 2003, IEEE / RSJ International Conference on Intelligent Robots and Systems (IROS 2003), Vol. 3, 2003).

[0063] In step S12, cells with angular structures are identified. To do this, the eigenvalues ​​of the covariance matrix are determined. Only cells whose ratio between the largest and smallest eigenvalues ​​is greater than a pre-defined threshold (e.g., 100) are retained as selected cells.

[0064] In step S13, it is checked whether the number of these selected units is less than a pre-given threshold. If so (alternatively: yes), the method terminates without determining the Manhattan orientation and continues by jumping to step S5. Figure 2 Alternatively (or: no), continue with the method in step S14.

[0065] In step S14, the orientation of the unit is calculated from its corresponding eigenvector (Ex, Ey), which is calculated as a = atan2(Ey, Ex) based on the largest eigenvector.

[0066] Next, in step S15, the orientation is normalized to an intermediate range from -45° to +45°. This normalization is commonly referred to as Manhattan normalization and is performed by adding or subtracting 90° or a multiple thereof.

[0067] In step S16, the Manhattan orientation is calculated as the dominant orientation from the group of calculated cell orientations of the selected cells. This can be done using various methods.

[0068] For example, according to one method, the similarity 's' between each unit orientation and other orientations can be calculated. The similarity between the orientations of the selected units is calculated separately, for example, according to... The angular spacing d between the Manhattan-normalized orientations of the cells is calculated, where σ can be a chosen parameter. Then, a similarity value for a cell is determined by summing all similarities s.

[0069] The unit orientation with the highest similarity value is specified as the Manhattan orientation for that detection cycle.

[0070] In step S17, if a predetermined number or more of the cell orientations are within a defined Manhattan-normalized distance from the calculated Manhattan orientation, the calculated Manhattan orientation is considered valid. The Manhattan orientation represents a preferred orientation of the environment, which is fixedly related to a reference coordinate system.

[0071] In step S18, the Manhattan node corresponding to the calculated Manhattan orientation is added to the SLAM graph. Then, Manhattan edges are added to the detected Manhattan orientation probes for a predetermined minimum number of times, and the Manhattan edges connect the corresponding attitude nodes to the newly added Manhattan nodes, the corresponding attitude nodes corresponding to the probes.

[0072] The Manhattan error function is associated with the Manhattan edge, and this Manhattan error function is defined as follows:

[0073] ,

[0074] in Corresponding to the orientation of the attitude node, Corresponding to the detected Manhattan orientation, Corresponding to the state of the Manhattan node, and Manhattan_normalize corresponds to the Manhattan normalization function, which normalizes the orientation to a range between -45° and 45°.

[0075] Each Manhattan edge is equipped with a robust kernel with an appropriate bandwidth (e.g., 0.1). The robust kernel allows for accurate estimation of the correct Manhattan angle when several edges with measurements containing errors are included in the graph. This technique corresponds to existing techniques and is described, for example, in "Robust map optimization using dynamic covariance scaling" by Agarwal, P., Tipaldi, GD, Spinello, L., Stachniss, C., and Burgard, W. (May 2013, IEEE International Conference on Robotics and Automation, pp. 62-69).

[0076] Next, the method uses Figure 2 Continue with step S6 of the method.

[0077] exist Figure 5a The image shows a schematic diagram of environmental map detection using a traditional SLAM algorithm. In contrast, Figure 5b An illustration shows the detected environment map based on the SLAM algorithm, taking into account the Manhattan world concept.

[0078] Expanding the SLAM graph by traversing Manhattan nodes and edges adds another set of variables to the optimization problem. Furthermore, an additional error term is generated for each Manhattan edge, which must be considered in the optimization problem.

[0079] The method described above enables the integration of the Manhattan world concept into a graph-based SLAM algorithm. This integration eliminates the need to refer to previously detected sensor data; instead, it supplements the SLAM graph with information derived from the understanding of the Manhattan world concept.

Claims

1. A computer-implemented method for operating a mobile agent (1) based on the pose of the mobile agent (1) in an environment, wherein, in order to operate the mobile agent (1), the following steps are used to perform localization of the mobile agent (1): - Detect sensor data about walls (2) and / or objects (3) located in the environment, wherein the sensor data describes the alignment and distance of the walls (2) and / or objects (3) in a fixed agent coordinate system (A) relative to the agent; - Considering Manhattan orientation, the pose of the mobile agent (1) in the environment is determined using a SLAM algorithm, wherein for each determination of a valid Manhattan orientation in the detection step, a Manhattan node and a Manhattan edge between one of the pose nodes and the Manhattan node involved are inserted into the SLAM graph, so that the subsequent optimization algorithm for minimizing error uses additional boundary conditions, which are pre-given by the Manhattan orientation, to determine the pose of each pose node, wherein the SLAM algorithm corresponds to a graph-based SLAM algorithm based on a SLAM graph with nodes and transformation edges, wherein the pose nodes respectively describe the pose of the mobile agent determined from the sensor data, and wherein there are transformation edges between each pair of pose nodes, wherein the transformation edges describe the pose change determined from the sensor data between the poses associated with each pair of pose nodes, wherein, according to the Manhattan orientation, a Manhattan node and at least one Manhattan edge between the Manhattan node and one of the pose nodes are added, wherein, The SLAM algorithm uses an error function determined from the SLAM graph, which takes into account the error probabilities associated with the nodes and edges, to determine the corrected pose associated with the nodes in the SLAM graph.

2. The method according to claim 1, wherein, The SLAM algorithm is configured to minimize an error function related to the measurement error of the sensor data, which is related to the Manhattan orientation.

3. The method according to claim 1, wherein, A set of points is determined from the sensor data, the set of points indicating the coordinates of the detected wall (2) and / or object (3), wherein the points in the set of points are combined into local units that are adjacent to each other, wherein Manhattan normalization is applied to the units having a unit orientation with a straight line or a 90° structure, wherein the Manhattan orientation is determined based on the dominant orientation of the straight line structure in the local unit.

4. The method according to claim 3, wherein, To determine the dominant orientation of the unit, for each unit orientation of a unit, a similarity to all other unit orientations is determined, and a similarity metric is determined for each unit, wherein the Manhattan orientation corresponds to the unit orientation of a unit selected from the units based on the similarity metric.

5. The method according to claim 4, wherein, The similarity is determined based on the orientation difference between each pair of cell orientations in the cell orientation after Manhattan normalization.

6. The method according to any one of claims 3 to 5, wherein, The determined Manhattan orientation is considered only if the orientation of more than a predetermined number of the cells is within the difference between the Manhattan-normalized orientation and the determined Manhattan orientation.

7. The method according to claim 1 or 2, wherein, The mobile agent (1) is operated according to the determined pose, and / or an environment map is determined with the aid of the current pose of the mobile agent (1), and the mobile agent (1) is operated according to the environment map.

8. The method according to claim 4, wherein, For each unit, the similarity metric is determined as the sum of the similarities of the corresponding unit with respect to the remaining units.

9. The method according to claim 4, wherein, The Manhattan orientation corresponds to the cell orientation of the cell that has the highest similarity metric among the cells.

10. The method according to claim 5, wherein, The similarity is determined by an exponential function of the negative difference after Manhattan normalization.

11. A device for operating a mobile agent (1) based on the posture of the mobile agent (1) in an environment, wherein the device is configured to perform localization of the mobile agent (1) according to the method of claim 1 using the following steps: - Detect sensor data about walls (2) and / or objects (3) located in the environment, wherein the sensor data describes the alignment and distance of the walls (2) and / or objects (3) in a fixed agent coordinate system (A) relative to the agent; - Taking Manhattan orientation into account, the pose of the mobile agent (1) in the environment is determined by means of a SLAM algorithm, wherein for each determination of a valid Manhattan orientation in the detection step, a Manhattan node and a Manhattan edge between one of the pose nodes and the Manhattan node involved are inserted into the SLAM graph, so that the subsequent optimization algorithm for minimizing the error then uses additional boundary conditions, which are given in advance by the Manhattan orientation, to determine the pose of each pose node.

12. The device according to claim 11, wherein the device is a control unit (11).

13. A mobile intelligent agent (1) having the device according to claim 11, an environmental detection sensing device (12) for providing sensor data, and a motion actuation device (13) configured to move the mobile intelligent agent (1).

14. A computer program product having instructions, the computer program product being configured to implement the method according to any one of claims 1 to 10 when the instructions are executed on a computing unit.

15. A machine-readable storage medium having a computer program product according to claim 14 stored on the machine-readable storage medium.

Images