METHOD FOR CONTROLLING AT LEAST ONE AUTONOMOUS WORKING DEVICE

Description

State of the art

[0001] A method for controlling, in particular for navigating, at least one autonomous working device, in particular an autonomous lawnmower, within a working area, wherein in at least one method step, working area data, in particular map data of the working area, are acquired, wherein in at least one method step, sensor data of the autonomous working device are acquired, has already been proposed.

[0002] Furthermore, reference is made to the documents US 5 793 934 A, US 2014 / 297090 A1 and DE 10 2015 119 865 A1. Disclosure of the invention

[0003] The invention is described in the appended claims.

[0004] The invention is based on a method for controlling, in particular for navigating, at least one autonomous working device, in particular an autonomous lawnmower, within a working area, wherein in at least one method step, working area data, in particular map data of the working area, are recorded, wherein in at least one method step, sensor data of the autonomous working device are recorded.

[0005] It is proposed that in at least one method step, at least one probability characteristic of possible states of the autonomous working device within the working area, which describes at least one localization uncertainty, in particular a sensing and / or movement uncertainty, of the autonomous working device, is evaluated to determine at least one movement path of the autonomous working device that at least substantially completely covers the working area.

[0006] The autonomous work device is intended, in particular, for autonomous movement, in particular over a surface. Preferably, the method for controlling, in particular for navigating, the autonomous work device is set up in an at least substantially two-dimensional work area, for example for moving on a lawn, on a floor, or the like. The work area can, in particular, be designed as a garden, a park, an apartment, in particular as a room, or as another work area that appears appropriate to a person skilled in the art. Alternatively, it is conceivable that the method for controlling, in particular for navigating, the autonomous work device is set up in a three-dimensional work area, for example for moving underwater, in the air, or the like.As an alternative to being designed as an autonomous lawnmower, it is conceivable for the autonomous working device to be designed as a vacuum robot, a lawn scarifier robot, a floor sweeping robot, a snow removal robot, a cleaning robot, a work drone, a pool cleaning robot, or any other autonomous working device that a person skilled in the art would deem appropriate. The autonomous working device is preferably intended to work the ground while moving over it. In particular, the autonomous working device is intended to mow a ground formed as a lawn, to vacuum and / or clean a ground formed as a floor, or the like. The autonomous working device preferably comprises at least one soil processing unit, in particular a cutting unit or a trimmer unit, for processing a ground.

[0007] The autonomous implement preferably comprises at least one drive unit, in particular with at least one electric motor, which is intended to drive at least one drive wheel unit of the autonomous implement for movement over a subsurface and / or to drive the soil cultivation unit. The autonomous implement preferably comprises at least one control and / or regulating unit configured to control and / or regulate the drive unit and / or the drive wheel unit. In particular, control data, in particular navigation data, are generated for the autonomous implement by means of the method. In particular, the control and / or regulating unit is configured to control and / or regulate the drive unit and / or the drive wheel unit depending on the control data. A "control and / or regulating unit" is to be understood in particular as a unit with at least one control electronics unit."Control electronics" is understood in particular to mean a unit with a processor unit and a memory unit, as well as with an operating program stored in the memory unit. Preferably, the autonomous work device, in particular at least one computing unit of the autonomous work device, and / or at least one external computing unit, for example a server, a multi-purpose computer, a laptop, or the like, is configured to carry out the method. Preferably, the control and / or regulating unit comprises the computing unit or is at least partially configured as a computing unit. "Provided" is understood in particular to mean specially equipped and / or specially configured. "Configured" is understood in particular to mean specially programmed and / or specially designed.The fact that an object is intended or configured for a specific function should be understood in particular to mean that the object fulfils and / or executes this specific function in at least one application and / or operating state.

[0008] In at least one method step, work area data, in particular map data of the work area, are preferably acquired by means of the autonomous work device and / or by means of an external device. The work area data preferably includes information about the work area, such as a size of the work area, boundaries of the work area, in particular a course of boundaries of the work area, a surrounding area of ​​the work area, obstacles, in particular positions of obstacles, in the work area, or other information about the work area that appears useful to a person skilled in the art.The workspace data can be embodied, in particular, as a SLAM (Simultaneous Localization and Mapping) map of the workspace, as a satellite image of the workspace, as a geographical map of the workspace, in particular a digital one, as a CAD model of the workspace, or as other workspace data that appears appropriate to a person skilled in the art. Preferably, at least some of the workspace data is acquired by the autonomous work device, in particular by sensors of the autonomous work device, for example, during at least one movement of the autonomous work device along the workspace, in particular during a training run monitored by a user of the autonomous work device.Alternatively or additionally, it is conceivable that at least part of the work area data is recorded by means of an external device, for example by means of a satellite, by means of a surveying tool, by means of a surveillance camera, by means of another autonomous work device or the like, and is provided in particular to the autonomous work device, in particular to the computing unit of the autonomous work device, and / or to the external computing unit.

[0009] In at least one further method step, sensor data of the autonomous work device, in particular sensor data of at least one sensor unit of the autonomous work device, are preferably acquired. In particular, at least one sensor model of the sensor unit of the autonomous work device is determined depending on the acquired sensor data. The sensor model is preferably determined outside the work area, in particular in a laboratory. The sensor unit is preferably configured to detect the surroundings of the autonomous work device, in particular obstacles in the work area, boundaries of the work area, or other landmarks in the work area that appear appropriate to a person skilled in the art. The sensor unit is preferably designed as an exteroceptive sensor unit. The sensor unit preferably comprises at least one laser scanner, in particular a 2D laser scanner.Alternatively or additionally, it is conceivable that the sensor unit comprises at least one camera, in particular a stereoscopic camera, an ultrasound scanner, a radar sensor, a lidar (light detection and distance measurement) sensor, or the like. The sensor data can be embodied as a maximum detection distance of the sensor unit, as noise in a measurement of the sensor unit, as a number of acquired measured values ​​per measuring process of the sensor unit, or the like. For example, for a sensor unit with a laser scanner, it is conceivable that the sensor unit has a maximum detection distance, in particular a scanning distance, of 8 m, a Gaussian-distributed noise in the measurement with a variance of 3 cm, and that 1850 measured values ​​are acquired per measuring process, in particular per scan, of the sensor unit.In particular, depending on the sensor data, in particular depending on the sensor model, at least a sensing uncertainty of the autonomous work device can be determined.

[0010] Preferably, a plurality of probability parameters, in particular at least one probability parameter at each possible position of the autonomous work device within the work area, is evaluated. The probability parameter is preferably designed as a, in particular Gaussian, probability distribution over possible states of the autonomous work device. Preferably, an orientation of the autonomous work device is not taken into account. In the following, the method is described by way of example as a function of a two-dimensional work area. In the two-dimensional work area, a state of the autonomous work device is preferably determined by a position ( x , y) of the autonomous work device, in particular within the work area, where ( x , y ) a coordinate pair from an area coordinate x and from another area coordinate y Preferably, the probability parameter is B In particular, the probability parameter B described by a normal distribution ( µ B , Σ B ): B ∼ N μ B Σ B , where µ B is an expected value of the probability parameter, where Σ B = σ x 2 σ xy 2 σ xy 2 σ yy 2 and where σ2< is a variance of the probability parameter. Preferably, a development of the probability parameter is simulated and modeled, in particular recursively using an Extended Kalman Filter (EKF), a Gaussian distribution, a particle filter, or the like, to show how the autonomous implement acquires information about a state of the autonomous implement through the sensor unit, in particular an exteroceptive one, and loses it through odometry drift. An "odometry drift" is to be understood in particular as an uncertainty, particularly increasing with increasing path length, in the position detection of the autonomous implement using interoceptive sensors, such as acceleration sensors, yaw rate sensors, rotary encoders, or the like. In particular, the greater the odometry drift of the autonomous implement, the greater the movement uncertainty of the autonomous implement.

[0011] The probability parameter describes, in particular, a localization uncertainty of the autonomous work device within the work area. The localization uncertainty of the autonomous work device is preferably greater, the greater the sensing uncertainty and the movement uncertainty of the autonomous work device. The sensing uncertainty depends, in particular, on the work area data and on the sensor model of the sensor unit of the autonomous work device. In particular, the sensing uncertainty is smaller the more and more precise work area data are recorded and the more precise the sensor model is. Preferably, after moving along the movement path that at least essentially completely covers the work area, the autonomous work device has been at every possible position within the work area at least once, and in particular has processed at least essentially the entire surface of the work area.A "movement path that at least substantially completely covers the work area" is to be understood in particular as a movement path that covers at least 85% of an area or volume of the work area, preferably at least 90% of the area or volume of the work area and particularly preferably at least 95% of the area or volume of the work area.

[0012] The inventive design of the method for controlling at least one autonomous work device allows for the determination of an advantageously safe, in particular at least substantially collision-free, movement path of the autonomous work device within a work area. Additional method steps, such as the installation of boundary wires, can advantageously be dispensed with. A convenient and user-friendly method for controlling the autonomous work device can advantageously be provided.

[0013] Furthermore, it is proposed that, in at least one method step, the workspace data and the sensor data are evaluated to determine at least one localizability map that indicates a localization probability of the autonomous work device at each position within the workspace. Preferably, the workspace data are assumed to be at least substantially error-free. In particular, the surroundings of the autonomous work device, in particular at least the boundaries of the workspace, are assumed to be static. Preferably, the sensor data are evaluated to determine the sensor model, in particular in a laboratory. Preferably, the, in particular evaluated, workspace data are combined with the sensor model to determine the localizability map.

[0014] Preferably, depending on the localizability map, it can be estimated how measurements at each position within the workspace influence a localization of the autonomous work device, in particular using an approach described below. Preferably, a measurement, for example a laser scan, is simulated, in particular by means of a field comparison point cloud or a SLAM map and by means of the sensor model. In particular, the simulated measurement is repeatedly translated and / or rotated by a predetermined distance within a predetermined convergence radius. Preferably, the translated and / or rotated measurements are combined with the workspace data, in particular with a map of the workspace, in particular to determine a covariance of positions to which convergence has occurred.The smaller the resulting covariance of a position, the more features useful for localization, such as the presence of a corner point, a tree, a house wall, or the like, a detected part of the environment of the autonomous work device at that position contains. The smaller the resulting covariance of a position, the higher the probability that an actual position of the measurement can be retrieved by comparing the sensor model with the work area data. To determine the localizability map, the approach described above is applied for each position in the work area. Alternatively, it is conceivable that the determined localizability map is further processed, for example as an estimate of an update step of an extended Kalman filter.The localizability map preferably includes an expected information gain of the sensor unit, in particular an exteroceptive one, at each position within the workspace. It is conceivable that the localizability map is improved in further method steps using newly acquired information, in particular using newly acquired workspace data and / or sensor data. Advantageously, an a priori localization probability of the autonomous work device within the workspace can be determined.

[0015] It is further proposed that, in at least one method step, the localizability map is combined with at least one movement model, in particular an odometry drift, of the autonomous implement to determine the at least one probability characteristic. In particular, the localization uncertainty of the autonomous implement depends on the movement model of the autonomous implement. In particular, the localization uncertainty of the autonomous implement increases without taking the localizability map into account with increasing path length in each dimension, for example in the x-direction and in y-direction. Preferably, the localizability map, in particular the information gained by the, in particular exteroceptive, sensor unit, influences the localization uncertainty of the autonomous work device, which is dependent on the movement model. For example, it is conceivable that during a movement of the autonomous work device in y -direction the localization uncertainty of the autonomous implement in the x-direction increases due to the odometry drift of the autonomous implement, but in y-direction due to an orientation feature running in the x-direction, for example a house wall, detected by the, in particular, exteroceptive, sensor unit. Preferably, by combining the localizability map with the movement model of the autonomous work device, a localization probability of the autonomous work device along any movement path within the work area can be calculated in advance. Preferably, the totality of all probability characteristics of the autonomous work device forms a space that is defined as a set of all probability distributions across the states of the autonomous work device and which is in particular a position x Covariances space can be viewed. Advantageously, the probability parameter to be evaluated can be determined.

[0016] Furthermore, it is proposed that in at least one method step, the at least one probability parameter is discretized for evaluation, in particular by a largest eigenvalue of the at least one probability parameter. In particular, the space forming the totality of all probability parameters of the autonomous work device is discretized in such a way that only probability parameters are taken into account that are arranged at regular intervals at nodes of an imaginary grid that at least substantially completely covers the work area. Preferably, each probability parameter B by the largest eigenvalue λ ( B ) of the probability parameter. In particular, the largest eigenvalue λ ( B ) of the probability parameter is defined as λ ¯ B = max λ 1 B , λ 2 B , where λ 1.2 ( B) Eigenvalues ​​over all probability parameters of Σ B in two-dimensional ( x , y )-space. Preferably, a discretized covariance over all probability parameters is defined as Σ B diskretisiert = λ ¯ B 2 0 0 λ ¯ B 2 .

[0017] Advantageously, the probability parameter can be evaluated in a time-saving manner. Advantageously, the movement path of the autonomous work device can be determined in a user-friendly and time-saving manner.

[0018] In addition, it is proposed that in at least one method step, the at least one probability parameter, in particular the at least one discretized probability parameter, is quantized for evaluation in uncertainty levels. In particular, each probability parameter, in particular each discretized probability parameter, is quantized into one of ∥ U ∥ Intervals of uncertainty with a size δquantized. Alternatively, it is conceivable that the probability parameters, in particular the discretized probability parameters, are quantized into intervals of uncertainty of variable size, in particular intervals of uncertainty of different sizes. Preferably, ∥ U ∥ defined as U = λ max − λ min δ , where λ max a largest eigenvalue of each expected probability characteristic during operation of the autonomous work device and λ min represents a smallest eigenvalue of each expected probability characteristic during operation of the autonomous work device. Preferably, each ub ∈ U an uncertainty level of a probability parameter B, where an uncertainty level ub is defined as u b = λ ¯ B − λ min δ .

[0019] Advantageously, a further time-saving and user-friendly evaluation of the probability parameter and determination of the movement path of the autonomous work device can be made possible.

[0020] Furthermore, it is proposed that, in at least one method step for determining the movement path, at least one collision probability of the autonomous work device with at least one obstacle in the work area is taken into account, which collision probability is determined as a function of a position of the obstacle and the at least one probability parameter, in particular the localization uncertainty of the autonomous work device. In particular, the collision probability of the autonomous work device is determined at every possible position of the autonomous work device within the work area and taken into account to determine the movement path. The greater the localization uncertainty of the autonomous work device at a specific position within the work area, the larger, in particular, the probability range around the position in which the autonomous work device is presumably arranged.Preferably, if the probability range overlaps with the at least one obstacle, in particular if the obstacle is at least partially located within the probability range, a collision of the autonomous work device with the obstacle is assumed. The probability range is preferably elliptical, in particular if the autonomous work device is oriented in the x-direction and in the x-direction. y- In particular, it is conceivable that in y -direction due to a low localization uncertainty of the autonomous work device, at least essentially collision-free movement and in x-direction due to a large localization uncertainty of the autonomous work device, a possible collision with an obstacle is detected. Preferably, the movement path of the autonomous work device is adapted depending on the determined collision probabilities of the autonomous work device. For example, it is conceivable that in a movement strategy of the autonomous work device along y -direction parallel paths, a movement path is determined in which the autonomous work device, depending on a detected possible collision with an obstacle, x -Direction before completing a path, switches to another, parallel path opposite to the direction of the obstacle. Advantageously, a particularly safe, in particular at least essentially collision-free, movement path of the autonomous work device in the work area can be determined.

[0021] It is further proposed that, in at least one method step, possible partial movement paths of the autonomous work device, which at least substantially completely cover at least one obstacle-free sub-area of ​​the work area, are determined, in particular in the form of a generalized round-trip problem, taking into account the at least one probability parameter, in particular the localization uncertainty of the autonomous work device. In particular, the work area comprises a plurality of obstacle-free sub-areas, the course of which, in particular their boundaries, is / are at least partially dependent on the positions of obstacles within the work area, on the boundaries of the work area, and / or on a movement strategy of the autonomous work device, for example, traveling along parallel paths.It is conceivable that the possible partial movement paths are determined taking into account the at least one probability parameter, in particular the localization uncertainty of the autonomous implement, and in particular additionally taking into account at least one further parameter. The further parameter can be designed in particular as a desired, in particular maximum or minimum, number of rotations of the autonomous implement, as a processing pattern to be generated, in particular a mowing pattern, as a desired, in particular maximum or minimum, distance of the autonomous implement from obstacles, as a desired, in particular maximum or minimum, path length, as an energy efficiency, or as another parameter that appears appropriate to a person skilled in the art.

[0022] Preferably, the sub-areas of the work area, in particular the possible partial movement paths of the autonomous work device covering the sub-areas, are determined by means of a Boustrophedon approach, in particular by means of a Boustrophedon coverage path planning algorithm, taking into account the probability parameters.Alternatively, it is conceivable that the sub-areas of the workspace, in particular the possible partial movement paths of the autonomous work device covering the sub-areas, are determined by means of a random pattern approach, a grid round-trip problem approach (grid TSP approach), a neural network approach, a grid local energy approach, a contour line approach, a convex scalable parallel calculation approach (convex spp approach) or another approach that appears reasonable to a person skilled in the art, taking into account the probability parameters.Preferably, at least the boundaries of the work area, in particular in the form of polygons, a Morse function, and a path width are used as input variables for determining the obstacle-free sub-areas, in particular the possible partial movement paths of the autonomous work device covering the sub-areas. Using the Boustrophedon approach, a pattern of parallel paths is preferably generated that covers the at least substantially entire work area and that, in particular, takes obstacles into account by dividing the work area into the obstacle-free sub-areas. Alternatively, particularly depending on an approach designed differently from the Boustrophedon approach, the generation of other patterns, for example random patterns, spiral paths, branched tree patterns, or the like, is conceivable.Preferably, an obstacle-free sub-area can be at least substantially completely covered, in particular traversed, by the autonomous work device using two different modes of movement. In particular, a first mode of movement of the autonomous work device is designed as movement along the parallel paths. In particular, a second mode of movement of the autonomous work device is designed as movement along the obstacles. Taking the two modes of movement into account, there are preferably four different options, in particular partial movement paths, for covering an at least substantially completely obstacle-free sub-area. In particular, each of the four partial movement paths has a starting point at which the autonomous work device begins to traverse the pattern and an exit point at which the partial movement path ends.For example, for an at least essentially rectangular obstacle-free sub-area, it is conceivable that a starting point of a first possible partial movement path is at an upper left corner point (. tl ) of the sub-area, a starting point of a second possible partial movement path at a lower left corner point ( bl ) of the sub-area, a starting point of a third possible partial movement path at an upper right corner point ( tr ) of the sub-area and a starting point of a fourth possible partial movement path at a lower right corner point ( br ) of the sub-area.

[0023] Preferably, the partial movement paths are determined in the form of a generalized round-trip problem, where the generalized round-trip problem is defined by a graph G , where G = V E w , where V are nodes, E are edges, and w are edge weights. Preferably, the nodes V are divided into pairwise disjoint sets V = ∪ i ∈ A V i , where A a set of obstacle-free subareas with size n is and where V i = ∪ sp ∈ tl tr bl br v i , sp , where a node v i , sp is to be understood as a solution for an obstacle-free sub-area i starting from a starting point sp. In particular, the generalized round-trip problem is defined as a problem of determining, in particular computing, a least-cost round-trip that contains exactly one node from each disjoint set V i Preferably, a set of nodes V is determined, in particular calculated, which comprises a node for each solution of all obstacle-free sub-regions of the work area, starting points and probability parameters, in particular uncertainty levels. In particular, a strategy is determined which is designed to determine a node vi,sp,u which is to be understood in particular as a solution of an obstacle-free sub-area istarting from a starting point sp with an uncertainty level u. In particular, to solve the knot vi,sp,uan expected development of the probability parameter along the movement path is tracked, and all movements of the autonomous work device that could lead to collisions are avoided by prematurely switching to an adjacent parallel path. In particular, depending on the strategy, in particular the movement strategy, uncovered, in particular untraversed, regions can remain within an obstacle-free sub-area. Preferably, the uncovered regions are regarded as independent obstacle-free sub-areas. In particular, taking into account the at least one probability parameter, in particular the localization uncertainty of the autonomous work device, possible partial movement paths of the autonomous work device that at least essentially completely cover the uncovered regions are determined, in particular in the form of a local generalized round-trip problem.Preferably, an uncovered region, in particular an area of ​​the uncovered region, is weighed against a distance to be covered up to the uncovered region, in particular by means of a parameter. β For example, it is conceivable that a detour of 50 m is avoided in order to cover an uncovered region with a comparatively small area.

[0024] Preferably, a set of nodes V i for each obstacle-free sub-area, in particular calculated, by iterating over all obstacle-free sub-areas, starting points and uncertainty levels, in particular using the following formula: V i = v i , sp , u , where ∀ i ∈ A , where sp ∈ {tl, tr, bl, br} and where u ∈ U. The nodes are preferably connected to form V. In particular, each node includes an induced exit probability parameter B exits ( vi,sp,u ), a path length d ( vi,sp,u ) and a remaining uncovered region o ( vi,sp,u ), in particular for determining, in particular calculating, node costs c(vi,sp,u ), in particular using the following formula: c v i , sp , u = d v i , sp , u + β × o v i , sp , u .

[0025] Preferably, depending on the node costs c ( vi,sp,u ) the edge weights w determined, in particular calculated. Advantageously, a user-friendly method can be provided for a time-saving determination of possible movement paths within the obstacle-free sub-areas, taking into account, in particular evaluating, the probability parameters.

[0026] Furthermore, it is proposed that, in at least one method step, possible transition paths of the autonomous work device between a plurality of sub-regions of the work area are determined, taking into account the at least one probability parameter, in particular the localization uncertainty of the autonomous work device. In particular, possible transition paths of the autonomous work device between adjacent sub-regions of the work area are determined, taking into account the at least one probability parameter. In particular, the edges E and the edge weights w are determined, in particular calculated. Preferably, the edge weights include target node costs and transition costs. Preferably, the edge weights are w determined according to the following formula: w v i v j = c v i v j + c v j , where ∀ i, j ∈ V , where vi = v i , sp , u , where i ∈ A , where sp ∈ { tl, tr, bl, br}, where u ∈ U and where c ( vi , vj ) the transition costs of a transition between an obstacle-free sub-area i and another barrier-free area j The transition takes place in particular between B exits ( vi ), the starting point of the further sub-area and the uncertainty levels of vj Preferably, a transition path between two sub-areas is determined that represents a compromise between a final localization uncertainty and path length. In particular, a traveled path length is compared with an accumulated localization uncertainty using a compromise parameter α, in particular by using a wavefront algorithm. Preferably, a transition path is determined, in particular calculated, for all (exit probability parameter, starting point) pairs, which is dependent on α and the localization uncertainty of the autonomous work device at a target node vj an uncertainty level ( B exits ( vi ), ( vi,sp,uinduced )) induced, where u induced ∈ U. In particular, all edge weights are determined, in particular calculated, in such a way that edges at which good transition paths can be determined are formed as edges with finite edge costs, in particular according to the following formula: w ( v i 1, sp 1 , u b1 , v i 2, sp 2 ,u b2 ) = c v i 2 , sp 2 , u b 2 + c v i 1 , sp 1 , u b 1 v i 2 , sp 2 , u b 2 wenn u b 2 = u induziert . ∞ sonst

[0027] Preferably, each node ( n - 1) × 4 × ∥ U ∥ edges. In particular, there are 4 × ∥U ∥ nodes per obstacle-free sub-area. For the at least essentially complete working area, this results in 16 × ∥ U ∥ 2< × (n - 1) × n edges, where n is a number of obstacle-free subregions in the workspace. Preferably, only a single transition between each (exit probability parameter, starting point) pair is considered. In particular, considering a single transition between each (exit probability parameter, starting point) pair results in a number of finite edges of 16 × ∥ U ∥ × ( n - 1) × n Advantageously, a user-friendly method can be provided for a time-saving determination of possible transition paths between a plurality of sub-areas, taking into account, in particular evaluating, the probability parameters.

[0028] Furthermore, it is proposed that, in at least one method step, the movement path of the autonomous work device is determined for a one-time, at least substantially complete coverage of each sub-area of ​​the work area as a function of the possible partial movement paths and transition paths of the autonomous work device, in particular in the form of a generalized round-trip problem. In particular, a coherent movement path of the autonomous work device is determined for a one-time, at least substantially complete coverage of each sub-area of ​​the work area as a function of the possible partial movement paths and transition paths of the autonomous work device, in particular in the form of a generalized round-trip problem. Preferably, a single partial movement path for each sub-area of ​​the work area is connected to form a coherent movement path by means of the transition paths.Preferably, a solution to the generalized round-trip problem is initialized using a greedy algorithm. In particular, the greedy algorithm is created by iteratively adding a nearest node from each sub-area until a complete round is found. An initial solution determined using the greedy algorithm is preferably improved, in particular using local search operators for the generalized round-trip problem, in particular until a completion criterion is met. Preferably, the generalized round-trip problem is solved with a priority regarding determining a movement path that includes a minimal number of collisions of the autonomous work device.Alternatively, it is conceivable that the generalized round-trip problem is solved with a priority regarding the determination of a movement path that has a minimal expected mean localization uncertainty, with a priority regarding the determination of a movement path that has a minimal path length, with a priority regarding the determination of a movement path that has aesthetically pleasing, in particular symmetrical, processing patterns, in particular mowing patterns, or with a combination of different priorities, in particular a combination of the aforementioned priorities. Advantageously, a coherent, safe, in particular low-collision, movement path of the autonomous work device can be determined, covering at least the substantially entire work area.

[0029] Furthermore, the invention is based on an autonomous working device, in particular the aforementioned autonomous working device, in particular an autonomous lawnmower, with at least one navigation unit for navigation within a work area, with at least one sensor unit for recording work area data and with at least one computing unit.

[0030] It is proposed that the computing unit be configured to evaluate at least one probability characteristic of possible states of the autonomous work device within the work area, which describes at least one localization uncertainty, in particular a sensing and / or movement uncertainty, of the autonomous work device, in particular to determine at least one movement path of the autonomous work device that at least substantially completely covers the work area. In particular, the computing unit is configured to carry out the method described above. Preferably, the computing unit is configured to provide determined navigation data to the navigation unit for navigation of the autonomous work device within the work area. Preferably, the control and / or regulating unit of the autonomous work device comprises the navigation unit.Advantageously, an autonomous work device can be provided which is configured for a user-friendly, in particular independent, determination of a movement path within the work area.

[0031] Furthermore, the invention is based on a system with at least one working device according to the invention and with at least one server unit, in particular a cloud server.

[0032] It is proposed that the autonomous work device have at least one, in particular wireless, communication unit for receiving control data, in particular navigation data, determined from the server unit by means of a method according to one of claims 1 to 9. The server unit is preferably arranged spatially separate from the autonomous work device, for example in a data center, in a server farm, in an office building or the like. Alternatively, it is conceivable that the server unit is arranged within the work area, for example integrated into a base station, in particular a charging station, for the autonomous work device. The communication unit can be designed in particular as a WLAN module, as a mobile radio module, in particular an LTE module, as a Bluetooth module, as a radio module or as another communication unit that appears appropriate to a person skilled in the art.Alternatively, it is conceivable for the communication unit to be wired, for example comprising a data transmission cable, data transmission contacts, or the like. Preferably, the communication unit is connected to the control and / or regulating unit, in particular to the navigation unit, by signal transmission technology, in particular via a signal transmission element, for providing received control data, in particular navigation data. Preferably, the server unit comprises at least one communication unit, in particular designed at least substantially analogously to the communication unit of the autonomous working device, for transmitting the determined control data, in particular navigation data, to the autonomous working device, in particular to the communication unit of the autonomous working device. In particular, the server unit is configured to transmit the determined control data, in particular navigation data, via an internet.Alternatively or additionally, it is conceivable that the server unit is configured to provide the determined control data to an external communication unit, in particular arranged in the base station for the autonomous work device, which is configured in particular to provide the received control data to the autonomous work device, in particular to the communication unit of the autonomous work device. Advantageously, determining the control data by the autonomous work device can be dispensed with, and energy consumption of the autonomous work device can be kept low. Advantageously, an autonomous work device with a user-friendly, long operating time can be provided.

[0033] It is further proposed that the server unit be configured to optimize at least one movement path of the autonomous work device determined by the computing unit of the autonomous work device, in particular by means of a method according to one of claims 1 to 9, and to provide it to the autonomous work device. In particular, the computing unit can be configured to terminate an improvement in the solution of the generalized round-trip problem for determining the movement path after a predetermined number of iterations and / or after a predetermined period of time.In particular, the computing unit can be configured to terminate the improvement of the solution to the generalized round-trip problem for determining the movement path after a maximum of 16,000 iterations and / or after a period of at most 60 seconds, preferably after a maximum of 8,000 iterations and / or after a period of at most 30 seconds, and particularly preferably after a maximum of 4,000 iterations and / or after a period of at most 15 seconds. Preferably, the server unit is configured to further improve the solution determined by the computing unit and in particular provided to the server unit by the autonomous work device, in particular by means of further iterations. Preferably, the server unit is configured to determine all partial movement paths and transition paths, in particular all nodes, and to provide them to the autonomous work device for storage, in particular in a lookup table.Preferably, the computing unit of the autonomous work device is configured to determine the movement path of the autonomous work device for a one-time, at least substantially complete coverage of each sub-area of ​​the work area, depending on the partial movement paths and transition paths of the autonomous work device determined and provided by the server unit, in particular in the form of a generalized round-trip problem. This advantageously enables a user-friendly, time-saving determination of an optimized movement path.

[0034] The method according to the invention, the autonomous working device according to the invention, and / or the system according to the invention are not intended to be limited to the application and embodiment described above. In particular, the method according to the invention, the autonomous working device according to the invention, and / or the system according to the invention can have a number of individual elements, components, units, and method steps that differs from the number stated herein in order to fulfill a functionality described herein. Furthermore, in the value ranges specified in this disclosure, values ​​within the stated limits are also to be considered disclosed and can be used arbitrarily. Drawings

[0035] Further advantages will become apparent from the following description of the drawings. The drawings illustrate an exemplary embodiment of the invention. The drawings, the description, and the claims contain numerous features in combination. Those skilled in the art will also expediently consider the features individually and combine them into useful further combinations.

[0036] They show: Fig. 1 shows a system according to the invention with an autonomous working device according to the invention and with a server unit in a perspective view, Fig. 2 shows a flow diagram of a method according to the invention in a schematic view, Fig. 3 shows a localizability map in a schematic view, Fig. 4 shows a section of the localizability map from Fig. 3in a schematic representation, Fig. 5 a sub-area of ​​a working area in a schematic representation, Fig. 6 a discretization via probability parameters in a symbolic schematic representation, Fig. 7 a section of a working area in a schematic representation, Fig. 8 the working area in a schematic representation and Fig. 9 the working area from Fig. 8 in another schematic representation. Description of the embodiment

[0037] Figure 1shows a system 64 with at least one autonomous working device 12, in particular an autonomous lawnmower, and with a server unit 66, in particular a cloud server, in a perspective view. The autonomous working device 12 preferably comprises at least one navigation unit 58 for navigation within a work area 14, at least one sensor unit 60 for recording work area data, and at least one computing unit 62. The computing unit 62 is preferably configured to evaluate at least one probability characteristic of possible states of the autonomous working device 12 within the work area 14, which describes at least one localization uncertainty, in particular a sensing and / or movement uncertainty, of the autonomous working device 12, in particular to determine at least one movement path 22 of the autonomous working device 12 that at least substantially completely covers the work area 14.The autonomous working device 12 preferably has at least one, in particular wireless, communication unit 68 for receiving from the server unit 66 by means of a, in particular in . Figure 2 illustrated, method 10, in particular for controlling, in particular for navigating the at least one autonomous work device 12, control data, in particular navigation data.

[0038] The autonomous working device 12 is intended, in particular, for autonomous movement, in particular over a subsurface. Preferably, the method 10 for controlling, in particular for navigating, the autonomous working device 12 is configured in an at least substantially two-dimensional work area 14, for example, for movement on a lawn, on a floor, or the like. The work area 14 can, in particular, be configured as a garden, a park, an apartment, in particular as a room, or as another work area that appears appropriate to a person skilled in the art. In the present exemplary embodiment, the work area 14 is configured, for example, as a garden. Alternatively, it is conceivable that the method 10 for controlling, in particular for navigating, the autonomous working device 12 is configured in a three-dimensional work area 14, for example, for movement underwater, in the air, or the like.As an alternative to being designed as an autonomous lawnmower, it is conceivable for the autonomous working device 12 to be designed as a vacuum robot, a lawn scarifier robot, a floor sweeping robot, a snow removal robot, a cleaning robot, a work drone, a pool cleaning robot, or any other autonomous working device that would appear appropriate to a person skilled in the art. The autonomous working device 12 is preferably designed to work the ground while moving over it. In particular, the autonomous working device 12 is designed to mow a surface designed as a lawn, to vacuum and / or clean a surface designed as a floor, or the like. The autonomous working device 12 preferably comprises at least one soil processing unit, in particular a cutting unit or a trimmer unit (not shown further here), for working a surface.

[0039] The autonomous working device 12 preferably comprises at least one drive unit 70, in particular with at least one electric motor, which is intended to drive at least one drive wheel unit 72 of the autonomous working device 12 for movement over a subsurface and / or to drive the soil cultivation unit. The autonomous working device 12 preferably comprises at least one control and / or regulating unit 74, which is configured to control and / or regulate the drive unit 70 and / or the drive wheel unit 72. In particular, control data, in particular navigation data, are generated for the autonomous working device 12 by means of the method 10. The computing unit 62 is preferably configured to provide determined navigation data to the navigation unit 58 for navigation of the autonomous working device 12 within the work area 14.The control and / or regulating unit 74 of the autonomous working device 12 preferably comprises the navigation unit 58. In particular, the control and / or regulating unit 74, in particular the navigation unit 58, is configured to control and / or regulate the drive unit 70 and / or the drive wheel unit 72 as a function of the control data. The autonomous working device 12, in particular the computing unit 62 of the autonomous working device 12, and / or at least one external computing unit, for example a server, a multi-purpose computer, a laptop or the like, in particular the server unit 66, is preferably configured to carry out the method 10. The control and / or regulating unit 74 preferably comprises the computing unit 62 or is at least partially designed as a computing unit 62.

[0040] The server unit 66 is preferably arranged spatially separate from the autonomous work device 12, for example in a data center, in a server farm, in an office building or the like (not shown in more detail here). Alternatively, it is conceivable that the server unit 66 is arranged within the work area 14, for example, integrated into a base station, in particular a charging station, for the autonomous work device 12. The communication unit 68 can be designed in particular as a WLAN module, as a mobile radio module, in particular an LTE module, as a Bluetooth module, as a radio module or as another communication unit that appears appropriate to a person skilled in the art. Alternatively, it is conceivable that the communication unit 68 is designed to be wired, for example comprising a data transmission cable, data transmission contacts or the like.Preferably, the communication unit 68 is connected by signal transmission technology, in particular via a signal transmission element 76, to the control and / or regulating unit 74, in particular to the navigation unit 58, for providing received control data, in particular navigation data. Preferably, the server unit 66 comprises at least one communication unit 78, in particular configured at least substantially analogously to the communication unit 68 of the autonomous work device 12, for transmitting the determined control data, in particular navigation data, to the autonomous work device 12, in particular to the communication unit 68 of the autonomous work device 12. In particular, the server unit 66 is configured to transmit the determined control data, in particular navigation data, via the Internet.Alternatively or additionally, it is conceivable that the server unit 66 is configured to provide the determined control data to an external communication unit, in particular arranged in the base station for the autonomous working device 12, which is configured in particular to provide the received control data to the autonomous working device 12, in particular to the communication unit 68 of the autonomous working device 12.

[0041] Preferably, the server unit 66 is configured to optimize at least one movement path 22 of the autonomous working device 12 determined by the computing unit 62 of the autonomous working device 12, in particular by means of a, in particular in Figure 2illustrated, method 10, and the autonomous work device 12. In particular, the computing unit 62 can be configured to terminate an improvement of a solution to a generalized round-trip problem for determining the movement path 22 after a predetermined number of iterations and / or after a predetermined period of time. In particular, the computing unit 62 can be configured to terminate the improvement of the solution to the generalized round-trip problem for determining the movement path 22 after a maximum of 16,000 iterations and / or after a period of time of at most 60 seconds, preferably after a maximum of 8,000 iterations and / or after a period of time of at most 30 seconds, and particularly preferably after a maximum of 4,000 iterations and / or after a period of time of at most 15 seconds.Preferably, the server unit 66 is configured to further improve the solution determined by the computing unit 62 and, in particular, provided to the server unit 66 by the autonomous work device 12, in particular by means of further iterations. Preferably, the server unit 66 is configured to determine all partial movement paths 44, 46 and transition paths 54, in particular all nodes, and to provide them to the autonomous work device 12 for storage, in particular in a lookup table.Preferably, the computing unit 62 of the autonomous working device 12 is configured to determine the movement path 22 of the autonomous working device 12 for a one-time, at least substantially complete coverage of each sub-area 48, 50 of the work area 14 as a function of the partial movement paths 44, 46 and transition paths 54 of the autonomous working device 12 determined and provided by the server unit 66, in particular in the form of a generalized round-trip problem.

[0042] Figure 2shows a flow diagram of the method 10 for controlling, in particular for navigating, at least one autonomous working device 12, in particular an autonomous lawnmower, within a working area 14, wherein in at least one method step 16 work area data, in particular map data of the work area, are recorded, wherein in at least one method step 18 sensor data of the autonomous working device 12 are recorded, in a schematic representation.Preferably, in at least one method step 20, at least one probability characteristic of possible states of the autonomous working device 12 within the working area 14, which describes at least one localization uncertainty, in particular a sensing and / or movement uncertainty, of the autonomous working device 12, is evaluated, in particular to determine at least one movement path 22 of the autonomous working device 12 that at least substantially completely covers the working area 14.

[0043] In method step 16, work area data, in particular map data of the work area 14, are preferably acquired by means of the autonomous work device 12 and / or by means of an external device. The work area data preferably includes information about the work area 14, such as a size of the work area 14, boundaries 80 of the work area 14, in particular a course of boundaries 80 of the work area 14, an environment of the work area 14, obstacles 36, 38, 40, in particular positions of obstacles 36, 38, 40, in the work area 14, or other information about the work area 14 that appears useful to a person skilled in the art.The work area data can be embodied, in particular, as a SLAM map of the work area 14, as a satellite image of the work area 14, as a geographical map, in particular a digital map, of the work area 14, as a CAD model of the work area 14, or as other work area data that appears appropriate to a person skilled in the art. Preferably, at least some of the work area data is acquired by means of the autonomous work device 12, in particular by means of sensors, in particular the sensor unit 60, of the autonomous work device 12, for example, during at least one movement of the autonomous work device 12 along the work area 14, in particular during a training run monitored by a user of the autonomous work device 12.Alternatively or additionally, it is conceivable that at least part of the work area data is captured by means of an external device, for example by means of a satellite, by means of a surveying tool, by means of a surveillance camera, by means of another autonomous work device or the like, and is provided in particular to the autonomous work device 12, in particular to the computing unit 62 of the autonomous work device 12, and / or to the external computing unit, in particular to the server unit 66.

[0044] In method step 18, sensor data of the autonomous work device 12, in particular sensor data of the sensor unit 60 of the autonomous work device 12, are preferably acquired. In particular, at least one sensor model of the sensor unit 60 of the autonomous work device 12 is determined depending on the acquired sensor data. The sensor model is preferably determined outside the work area 12, in particular in a laboratory. The sensor unit 60 is preferably configured to detect an environment of the autonomous work device 12, in particular obstacles 36, 38, 40 in the work area 14, boundaries 80 of the work area 14, or other landmarks in the work area 14 that appear appropriate to a person skilled in the art. The sensor unit 60 is preferably designed as an exteroceptive sensor unit. The sensor unit 60 preferably comprises at least one laser scanner, in particular a 2D laser scanner.Alternatively or additionally, it is conceivable that the sensor unit 60 comprises at least one camera, in particular a stereoscopic camera, an ultrasound scanner, a radar sensor, a lidar sensor, or the like. The sensor data can be embodied as a maximum detection distance of the sensor unit 60, as noise in a measurement of the sensor unit 60, as a number of acquired measured values ​​per measuring process of the sensor unit 60, or the like. For example, for the sensor unit 60 with the laser scanner, it is conceivable that the sensor unit 60 has a maximum detection distance, in particular a scanning distance, of 8 m, a Gaussian-distributed noise in the measurement with a variance of 3 cm, and that 1850 measured values ​​are acquired per measuring process, in particular per scan, of the sensor unit 60.In particular, depending on the sensor data, in particular depending on the sensor model, at least one sensing uncertainty of the autonomous working device 12 can be determined.

[0045] Preferably, in particular in method step 20, a plurality of probability parameters, in particular at least one probability parameter at each possible position of the autonomous work device 12 within the work area 14, is evaluated. The probability parameter is preferably designed as a, in particular Gaussian, probability distribution over possible states of the autonomous work device 12. Preferably, an orientation of the autonomous work device 12 is disregarded. In the following, the method 10 is described by way of example as a function of the two-dimensional work area 14. In the two-dimensional work area 14, a state of the autonomous work device 12 is preferably determined by a position ( x , y ) of the autonomous work device 12, in particular within the work area 14, wherein ( x , y ) a coordinate pair from an area coordinatex and from another area coordinate y Preferably, the probability parameter is B In particular, the probability parameter B described by a normal distribution ( µ B , Σ B ): B ∼ N μ B Σ B , where µ B is an expected value of the probability parameter, where ∑ B = σ x 2 σ xy 2 σ xy 2 σ yy 2 and where σ2< is a variance of the probability parameter. Preferably, a development of the probability parameter is simulated and modeled, in particular recursively using an extended Kalman filter, a Gaussian distribution, a particle filter, or the like, to show how the autonomous work device 12 acquires information about a state of the autonomous work device 12 through the, in particular exteroceptive, sensor unit 60 and loses information through an odometry drift. In particular, the greater the odometry drift of the autonomous work device 12, the greater the movement uncertainty of the autonomous work device 12.

[0046] The probability parameter describes, in particular, a localization uncertainty of the autonomous work device 12 within the work area 14. Preferably, the localization uncertainty of the autonomous work device 12 is greater the greater the sensing uncertainty and the movement uncertainty of the autonomous work device 12. The sensing uncertainty depends, in particular, on the work area data and on the sensor model of the sensor unit 60 of the autonomous work device 12. In particular, the sensing uncertainty is smaller the more and more precise the work area data is recorded and the more precise the sensor model is.Preferably, after moving along the movement path 22 covering the work area 14 at least substantially completely, the autonomous working device 12 has been at least once at every possible position within the work area 14, in particular has processed at least substantially the entire surface of the work area 14.

[0047] Preferably, in at least one further method step 24, the workspace data and the sensor data are evaluated to determine at least one localizability map 26, which indicates a localization probability of the autonomous work device 12 at each position within the workspace 14. Preferably, the workspace data are assumed to be at least substantially error-free. In particular, the surroundings of the autonomous work device 12, in particular at least the boundaries 80 of the workspace 14, are assumed to be static. Preferably, the sensor data are evaluated to determine the sensor model, in particular in a laboratory. Preferably, the, in particular evaluated, workspace data are combined with the sensor model to determine the localizability map 26.

[0048] Preferably, depending on the localizability map 26, it can be estimated how measurements at each position within the workspace 14 influence a localization of the autonomous work device 12, in particular using an approach described below. Preferably, a measurement, for example a laser scan, is simulated, in particular by means of a field comparison point cloud or a SLAM map and by means of the sensor model. In particular, the simulated measurement is repeatedly translated and / or rotated by a predetermined distance within a predetermined convergence radius. Preferably, the translated and / or rotated measurements are combined with the workspace data, in particular with a map of the workspace 14, in particular to determine a covariance of positions to which convergence has occurred.The smaller the resulting covariance of a position, the more features useful for localization, such as the presence of a corner point, a tree, a house wall, or the like, a detected part of the surroundings of the autonomous work device 12 at the position contains. The smaller the resulting covariance of a position, the higher the probability that an actual position of the measurement can be retrieved by comparing the sensor model with the work area data. To determine the localizability map 26, the approach described above is applied for each position of the work area 14. Alternatively, it is conceivable that the determined localizability map 26 is further processed, for example as an estimate of an update step of an extended Kalman filter.Preferably, the localizability map 26 comprises an expected information gain of the, in particular exteroceptive, sensor unit 60 at each position within the work area 14. It is conceivable that the localizability map 26 is improved in further method steps by means of newly acquired information, in particular by means of newly acquired work area data and / or sensor data.

[0049] Figure 3shows the localizability map 26 in a schematic representation. The localizability map 26 is divided, in particular, into different areas 82, 84, 86, 88, 90, in which the autonomous working device 12 has different localization probabilities. In a first area 82 and in a fifth area 90, the autonomous working device 12 has, in particular, a greater localization probability than in a second area 84, in a third area 86, and in a fourth area 88. In particular, the localization probabilities of the autonomous working device 12 in the first area 82 and in the fifth area 90 at least largely have a sum of eigenvalues ​​of at most 10 cm. In particular, the localization probabilities of the autonomous working device 12 in the second area 84, in the third area 86, and in the fourth area 88 at least largely have a sum of eigenvalues ​​of at least 100 cm.

[0050] Figure 4 shows one, especially in Figure 3 marked, section 92 from the localizability map 26 from Fig. 3 in a schematic representation. The localization probabilities of the autonomous work device 12 at different positions within the work area 14 are represented in particular by means of error ellipses 94. In particular, the smaller the localization probability of the autonomous work device 12 at a position, the larger the error ellipse 94 at that position. In the present exemplary embodiment, the largest error ellipses 94 have, for example, a maximum diameter of at most 1 m.

[0051] Preferably, in at least one further method step 28, the localizability map 26 is combined with at least one movement model, in particular an odometry drift, of the autonomous working device 12 to determine the at least one probability characteristic. In particular, the localization uncertainty of the autonomous working device 12 depends on the movement model of the autonomous working device 12. In particular, the localization uncertainty of the autonomous working device 12 increases without taking the localizability map 26 into account with increasing path length in each dimension, for example in the x-direction 98 and in y -direction 96. Preferably, the localizability map 26, in particular the information gained by the, in particular exteroceptive, sensor unit 60, influences the localization uncertainty of the autonomous work device 12, which depends on the movement model.

[0052] Figure 5shows a partial area 48 of the work area 14 in a schematic representation. For example, it is conceivable that during a movement of the autonomous work device 12 in y -Direction 96 the localization uncertainty of the autonomous work device 12 in x -direction 98 due to the odometry drift of the autonomous working device 12, but in y -Direction 96 remains constant due to an orientation feature 100, for example a house wall, running in the x-direction 98 and detected by the sensor unit 60, in particular an exteroceptive one. The change in the localization uncertainty of the autonomous work device 12 is represented in particular by the error ellipses 94.

[0053] Preferably, by combining the localizability map 26 with the movement model of the autonomous work device 12, a localization probability of the autonomous work device 12 along any movement path 22 within the work area 14 can be calculated in advance. Preferably, the totality of all probability characteristics of the autonomous work device 12 forms a space that is defined as a set of all probability distributions across the states of the autonomous work device 12 and, in particular, as a position x Covariances -room can be viewed.

[0054] Preferably, in at least one further method step 30, the at least one probability parameter is discretized for evaluation, in particular by a largest eigenvalue of the at least one probability parameter. In particular, the space forming the totality of all probability parameters of the autonomous work device 12 is discretized in such a way that only probability parameters are taken into account that are arranged at regular intervals at nodes of an imaginary grid that at least substantially completely covers the work area 14. Preferably, each probability parameter B by the largest eigenvalue λ ( B ) of the probability parameter. In particular, the largest eigenvalue λ ( B ) of the probability parameter is defined as λ ( B ) = max( λ 1 ( B ), λ 2 ( B)), where λ 1.2 ( B ) Eigenvalues ​​over all probability parameters of Σ B in two-dimensional ( x , y )-space. Preferably, a discretized covariance over all probability parameters is defined as ∑ B diskretisiert = λ ¯ B 2 0 0 λ ¯ B 2 .

[0055] Figure 6 shows the discretization over probability parameters Σ B in a symbolic schematic representation. The covariance over all probability parameters Σ B is symbolized in particular by an ellipse 102. The discretized covariance over all probability parameters Σ Bdiscretized is symbolized in particular by another ellipse 104. A first eigenvalue λ 1 ( B ) of a Gaussian distribution with covariance Σ B over the probability parameters is symbolized by a first arrow 106. A second eigenvalue λ 2 ( B) of the Gaussian distribution with covariance Σ B over the probability parameters is symbolized by a second arrow 108. A discretization is carried out in particular via the largest eigenvalue λ ( B ), which is symbolized in particular by a third arrow 110. Preferably, the discretization over the largest eigenvalue λ ( B ) a Gaussian probability distribution with covariance matrix Σ Bdiscretized .

[0056] Preferably, in at least one further method step 32, the at least one probability parameter, in particular the at least one discretized probability parameter, is quantized for evaluation in uncertainty levels. In particular, each probability parameter, in particular each discretized probability parameter, is quantized into one of ∥ U ∥ Intervals of uncertainty with a size δquantized. Alternatively, it is conceivable that the probability parameters, in particular the discretized probability parameters, are quantized into intervals of uncertainty of variable size, in particular intervals of uncertainty of different sizes. Preferably, ∥ U ∥ defined as U = λ max − λ min δ , where λ max a largest eigenvalue of each expected probability characteristic during operation of the autonomous work device 12 and λ min represents a smallest eigenvalue of each expected probability characteristic during operation of the autonomous work device 12. Preferably, each ub ∈ U an uncertainty level of a probability parameter B, where an uncertainty level ub is defined as u b = λ ¯ B − λ min δ .

[0057] According to the invention, in at least one method step 34 for determining the movement path 22, at least one collision probability of the autonomous working device 12 with at least one obstacle 36, 38, 40 in the work area 14 is taken into account, which collision probability is determined as a function of a position of the obstacle 36, 38, 40 and the at least one probability characteristic, in particular the localization uncertainty of the autonomous working device 12.

[0058] According to the invention, the collision probability of the autonomous work device 12 is determined at every possible position of the autonomous work device 12 within the work area 14 and taken into account to determine the movement path 22. The greater the localization uncertainty of the autonomous work device 12 at a specific position within the work area 14, the larger, in particular, a probability range around the position in which the autonomous work device 12 is presumably arranged. Preferably, if the probability range overlaps with the at least one obstacle 36, 38, 40, in particular if the obstacle 36, 38, 40 is at least partially arranged in the probability range, a collision of the autonomous work device 12 with the obstacle 36, 38, 40 is assumed. Preferably, the probability range is elliptical, in particular if the autonomous work device 12 is in the x-direction 98 and iny -direction has 96 different localization uncertainties. The probability range is in Figure 5 in particular at least essentially represented by the error ellipses 94. In particular, it is conceivable that in y -direction 96 due to a low localization uncertainty of the autonomous working device 12, an at least substantially collision-free movement is determined and in the x-direction 98 due to a high localization uncertainty of the autonomous working device 12, a possible collision with an obstacle 36 is determined (cf. Figure 5 ). Preferably, the movement path 22 of the autonomous working device 12 is adapted depending on the determined collision probabilities of the autonomous working device 12. For example, it is conceivable that in a movement strategy of the autonomous working device 12 along y-A movement path 22 is determined from parallel paths 112, 114 running in the direction 96, in which the autonomous working device 12, depending on a determined possible collision with an obstacle 36, is x -Direction 98 changes to another parallel path 114 opposite to the direction of the obstacle 36 before completing a path 112.

[0059] Preferably, in at least one further method step 42, taking into account the at least one probability parameter, in particular the localization uncertainty of the autonomous work device 12, possible partial movement paths 44, 46 of the autonomous work device 12, which at least substantially completely cover at least one obstacle-free sub-area 48, 50 of the work area 14, are determined, in particular in the form of a generalized round-trip problem. In particular, the work area 14 comprises a plurality of obstacle-free sub-areas 48, 50, the course of which, in particular the boundaries thereof, is / are at least partially dependent on the positions of obstacles 36, 38, 40 within the work area 14, on the boundaries 80 of the work area 14, and / or on a movement strategy of the autonomous work device 12, for example, traveling along parallel paths 112, 114.It is conceivable that the possible partial movement paths 44, 46 are determined taking into account the at least one probability parameter, in particular the localization uncertainty of the autonomous working device 12, and in particular additionally taking into account at least one further parameter. The further parameter is preferably specified and / or determined in at least one further method step 154. The further parameter can be designed in particular as a desired, in particular maximum or minimum, number of rotations of the autonomous working device 12, as a processing pattern to be generated, in particular a mowing pattern, as a desired, in particular maximum or minimum, distance of the autonomous working device 12 from obstacles 36, 38, 40, as a desired, in particular maximum or minimum, path length, as an energy efficiency, or as another parameter that appears appropriate to a person skilled in the art.

[0060] In particular, the work area 14 is divided into the obstacle-free sub-areas 48, 50 in at least one further method step 116. Preferably, a task of the autonomous work device 12, in particular determined and / or defined in at least one further method step 118, and a movement strategy of the autonomous work device 12, in particular determined and / or defined in at least one further method step 120, are used as input parameters for dividing the work area 14 into the obstacle-free sub-areas 48, 50. In the present exemplary embodiment, the task of the autonomous work device 12 is embodied, for example, as covering, in particular traveling, the at least substantially entire work area 14. In the present exemplary embodiment, the movement strategy of the autonomous work device 12 is embodied, for example, as moving along parallel paths 112, 114.

[0061] Preferably, the sub-areas 48, 50 of the working area 14, in particular the possible partial movement paths 44, 46 of the autonomous work device 12 covering the sub-areas 48, 50, are determined by means of a Boustrophedon approach, in particular by means of a Boustrophedon coverage path planning algorithm, taking into account the probability parameters. Alternatively, it is conceivable that the sub-areas 48, 50 of the work area 14, in particular the possible partial movement paths 44, 46 of the autonomous work device 12 covering the sub-areas 48, 50, are determined by means of a random pattern approach, by means of a grid round-trip problem approach, by means of a neural network approach, by means of a grid local energy approach, by means of an isoline approach, by means of a convex scalable parallel calculation approach or by means of another approach that appears reasonable to a person skilled in the art, taking into account the probability parameters.Preferably, at least the boundaries 80 of the work area 14, in particular in the form of polygons, a Morse function, and a path width are used as input variables for determining the obstacle-free sub-areas 48, 50, in particular the possible partial movement paths 44, 46 of the autonomous work device 12 covering the sub-areas 48, 50. Using the Boustrophedon approach, a pattern of parallel paths 112, 114 is preferably generated that covers at least substantially the entire work area 14 and that, in particular, takes obstacles 36, 38, 40 into account by dividing the work area 14 into the obstacle-free sub-areas 48, 50. Alternatively, particularly depending on an approach different from the Boustrophedon approach, it is conceivable to generate other patterns, for example random patterns, spiral paths, branched tree patterns, or the like.Preferably, an obstacle-free sub-area 48, 50 can be at least substantially completely covered, in particular traversed, by the autonomous work device 12 using two different modes of movement of the autonomous work device 12. In particular, a first mode of movement of the autonomous work device 12 is designed as movement along the parallel paths 112, 114. In particular, a second mode of movement of the autonomous work device 12 is designed as movement along the obstacles 36, 38, 40. Taking the two modes of movement into account, there are preferably four different possibilities, in particular partial movement paths 44, 46, for covering an at least substantially completely obstacle-free sub-area 48, 50.In particular, each of the four partial movement paths 44, 46 has a starting point 122, 124 at which the autonomous work device 12 begins to travel the pattern, and an exit point 126, 128 at which the partial movement path 44, 46 ends. For example, for an at least substantially rectangular obstacle-free sub-area 48, 50, it is conceivable that a starting point 124 of a first possible partial movement path 46 is at an upper left corner point (. tl ) 130 of the sub-area 50, a starting point 122 of a second possible partial movement path 44 at a lower left corner point ( bl ) 132 of sub-area 48, a starting point of a third possible partial movement path at an upper right corner point ( tr ) of the sub-area and a starting point of a fourth possible partial movement path at a lower right corner point ( br ) of the sub-area.

[0062] Figure 7shows a section of the work area 14 in a schematic representation. Shown are an obstacle-free sub-area 48, a further obstacle-free sub-area 50 and two additional obstacle-free sub-areas 134, 136. The obstacle-free sub-area 48 comprises in particular a lower left corner point 132, an upper left corner point 138, an upper right corner point 140 and a lower right corner point 142. The further obstacle-free sub-area 50 comprises in particular an upper left corner point 130, a lower left corner point 144, an upper right corner point 146 and a lower right corner point 148. By way of example, in the obstacle-free sub-area 48, a possible partial movement path 44 of the autonomous work device 12 is shown with a starting point 122 at the lower left corner point 132 of the sub-area 48 and with an exit point 126 at the lower right corner point 142 of the sub-area 48.By way of example, in the further obstacle-free sub-area 50, a possible partial movement path 46 of the autonomous work device 12 is shown with a starting point 124 at the upper left corner point 130 of the further sub-area 50 and with an exit point 128 at the upper right corner point 146 of the further sub-area 50. The first mode of movement of the autonomous work device 12 is shown in particular by fourth arrows 150. The second mode of movement of the autonomous work device 12 is shown in particular by fifth arrows 152. For the sake of clarity, no corner points and partial movement paths of the additional sub-areas 134, 136 are shown.

[0063] Preferably, the partial movement paths 44, 46 are determined in the form of a generalized round-trip problem, wherein the generalized round-trip problem is defined by a graph G, where G = V E w , where V are nodes, E are edges, and w are edge weights. Preferably, the nodes V are divided into pairwise disjoint sets V = U i ∈ A V i , where A a set of obstacle-free sub-areas 48, 50 with size n is and where V i = ∪ sp ∈ tl tr bl br v i , sp , where a node vi,sp is to be understood as a solution for an obstacle-free sub-area i starting from a starting point sp. In particular, the generalized round-trip problem is defined as a problem of determining, in particular computing, a least-cost round-trip that contains exactly one node from each disjoint set V iPreferably, a set of nodes V is determined, in particular calculated, which comprises a node for each solution of all obstacle-free sub-regions 48, 50 of the working area 14, starting points 122, 124 and probability parameters, in particular uncertainty levels. In particular, a strategy is determined which is designed to determine a node vi,sp,u which is to be understood in particular as a solution of an obstacle-free sub-area i starting from a starting point sp with an uncertainty level u. In particular, to solve the knot vi,sp,uan expected development of the probability characteristic along the movement path 22 is tracked, and all movements of the autonomous working device 12 that could lead to collisions are avoided by prematurely switching to an adjacent parallel path 112, 114. In particular, depending on a strategy, in particular the movement strategy, uncovered, in particular untraversed, regions can remain within an obstacle-free sub-area 48, 50. Preferably, the uncovered regions are regarded as independent obstacle-free sub-areas. In particular, taking into account the at least one probability characteristic, in particular the localization uncertainty of the autonomous working device 12, possible partial movement paths of the autonomous working device 12 that at least substantially completely cover the uncovered regions are determined, in particular in the form of a local generalized round-trip problem.Preferably, an uncovered region, in particular an area of ​​the uncovered region, is weighed against a distance to be covered up to the uncovered region, in particular by means of a parameter. β For example, it is conceivable that a detour of 50 m is avoided in order to cover an uncovered region with a comparatively small area.

[0064] Preferably, a set of nodes V i for each obstacle-free sub-area 48, 50 determined, in particular calculated, by iterating over all obstacle-free sub-areas 48, 50, starting points 122, 124 and uncertainty levels, in particular using the following formula: V i = v i , sp , u , where ∀ i ∈ A , where sp ∈ {tl, tr, bl, br} and where u ∈ U. The nodes are preferably connected to form V. In particular, each node includes an induced exit probability parameter B exits ( vi,sp,u), a path length d ( vi,sp,u ) and a remaining uncovered region o ( vi,sp,u ) , in particular to determine, in particular calculate, node costs c ( vi,sp,u ), in particular using the following formula: c v i , sp , u = d v i , sp , u + β × o v i , sp , u .

[0065] Preferably, depending on the node costs c ( vi,sp,u ) the edge weights w are determined, in particular calculated.

[0066] Preferably, in at least one further method step 52, possible transition paths 54 of the autonomous work device 12 between a plurality of sub-areas 48, 50 of the work area 14 are determined, taking into account the at least one probability parameter, in particular the localization uncertainty of the autonomous work device 12. In particular, possible transition paths 54 of the autonomous work device 12 between adjacent sub-areas 48, 50 of the work area 14 are determined, taking into account the at least one probability parameter. In particular, to determine the transition paths 54, the edges E and the edge weights w determined, in particular calculated. Preferably, the edge weights include target node costs and transition costs. Preferably, the edge weights w are determined according to the following formula: w v i v j = c v i v j + c v j , where ∀ i, j ∈ V, where vi = v i,sp, u , where i ∈ A , where sp ∈ {tl, tr, bl, br}, where u ∈ U and where c ( vi , vj ) the transition costs of a transition between an obstacle-free sub-area i and another barrier-free area j The transition takes place in particular between B exits ( vi ), the starting point 124 of the further sub-area 50 and the uncertainty levels of vj Preferably, a transition path 54 between two sub-areas 48, 50 is determined, which represents a compromise between a final localization uncertainty and path length. In particular, a traveled path length is compared with an accumulated localization uncertainty by means of a compromise parameter α, in particular by using a wavefront algorithm. Preferably, a transition path 54 is determined, in particular calculated, for all (exit probability parameter, starting point) pairs, which is dependent on α and the localization uncertainty of the autonomous work device 12 at a target node vj an uncertainty level ( B exits ( vi ), ( v j ,p, u induced )) induced, where u induced ∈ U. In particular, all edge weights are determined, in particular calculated, in such a way that edges at which good transition paths 54 can be determined are formed as edges with finite edge costs, in particular according to the following formula: w ( v i 1, sp 1, u b1 , v i 2, sp 2, u b2 ) = c v i 2 , sp 2 , u b 2 + c v i 1 , sp 1 , u b 1 v i 2 , sp 2 , u b 2 wenn u b 2 = u induziert ∞ sonst . Preferably, each node ( n - 1) × 4 × ∥ U∥ edges. In particular, there are 4 × ∥ U ∥ Nodes per obstacle-free sub-area 48, 50. For the at least essentially complete working area 14, in particular 16 × ∥ U ∥ 2< × ( n - 1) × n edges, where n a number of obstacle-free sub-regions 48, 50 in the working area 14. Preferably, only a single transition between each (exit probability parameter, starting point) pair is considered. In particular, considering a single transition between each (exit probability parameter, starting point) pair results in a number of finite edges of 16 × ∥ U ∥ × ( n - 1) × n .

[0067] Figure 8shows the work area 14 in a schematic representation. In particular, the movement path 22 of the autonomous work device 12, which at least substantially completely covers the work area 14, is shown with a global starting point 156. The movement path 22 includes the transition paths 54 between the sub-areas 48, 50 of the work area 14. The transition paths 54 between the sub-areas 48, 50 of the work area 14 are shown in dashed lines for clarity.

[0068] Preferably, in at least one further method step 56, the movement path 22 of the autonomous work device 12 is determined for a one-time, at least substantially complete coverage of each sub-area 48, 50 of the work area 14 as a function of the possible partial movement paths 44, 46 and transition paths 54 of the autonomous work device 12, in particular in the form of a generalized round-trip problem. In particular, a coherent movement path 22 of the autonomous work device 12 is determined for a one-time, at least substantially complete coverage of each sub-area 48, 50 of the work area 14 as a function of the possible partial movement paths 44, 46 and transition paths 54 of the autonomous work device 12, in particular in the form of a generalized round-trip problem.Preferably, a single partial movement path 44, 46 for each sub-area 48, 50 of the workspace 14 is connected by means of the transition paths 54 to form a connected movement path 22. Preferably, a solution to the generalized round-trip problem is initialized using a greedy algorithm. In particular, the greedy algorithm is created by iteratively adding a nearest node from each sub-area 48, 50 until a complete round is found. An initial solution determined using the greedy algorithm is preferably improved, in particular using local search operators for the generalized round-trip problem, in particular until a completion criterion is met. Preferably, the generalized round-trip problem is solved with a priority regarding determining a movement path 22 that includes a minimal number of collisions of the autonomous work device 12.Alternatively, it is conceivable that the generalized round-trip problem is solved with a priority with regard to determining a movement path 22 which has a minimum expected mean localization uncertainty, with regard to determining a movement path 22 which has a minimum path length, with regard to determining a movement path 22 which has aesthetically, in particular symmetrically, pleasing processing patterns, in particular mowing patterns, or with a combination of different priorities, in particular a combination of the aforementioned priorities.

[0069] Figure 9 shows the workspace 14 from Fig. 8 in another schematic representation. The development of the probability parameter along the movement path 22 is represented by error ellipses 94.

[0070] With regard to further method steps of the method 10 for controlling, in particular for navigating, the autonomous work device 12, reference may be made to the preceding description of the autonomous work device 12 and / or the system 64, since this description is to be read analogously to the method 10 and thus all features with regard to the autonomous work device 12 and / or with regard to the system 64 are also deemed to be disclosed with regard to the method 10 for controlling, in particular for navigating, the autonomous work device 12.

Claims

1. Method for controlling, in particular navigating, at least one autonomous working device, in particular an autonomous lawn mower, within a working region (14), wherein in at least one method step (16) working region data, in particular map data of the working region, are captured, wherein in at least one method step (18) sensor data of the autonomous working device are captured, wherein in at least one method step (20) at least one probability characteristic of possible states of the autonomous working device within the working region (14) that describes at least a localization uncertainty, in particular a sensing and / or movement uncertainty, of the autonomous working device is evaluated for determining at least one path of movement (22) of the autonomous working device that covers the working region (14) at least substantially completely, characterized in that in at least one method step (34) at least one collision probability of the autonomous working device with at least one obstacle (36, 38, 40) in the working region (14) that is determined in dependence on a position of the obstacle (36, 38, 40) and the at least one probability characteristic, in particular the localization uncertainty, of the autonomous working device, is taken into consideration for the determination of the path of movement (22), wherein the collision probability of the autonomous working device is determined at each possible position of the autonomous working device within the working region (14) and is taken into consideration for the determination of the path of movement (22).

2. Method according to Claim 1, characterized in that in at least one method step (24) the working region data and the sensor data are evaluated for a determination of at least one localizability map (26), which indicates a localization probability of the autonomous working device at each position within the working region (14).

3. Method according to Claim 2, characterized in that in at least one method step (28) the localizability map (26) is combined with at least one movement model, in particular an odometry drift, of the autonomous working device for a determination of the at least one probability characteristic.

4. Method according to one of the preceding claims, characterized in that in at least one method step (30) the at least one probability characteristic is discretized for an evaluation, wherein the space forming the entirety of all the probability characteristics of the autonomous working device is discretized in such a way that only probability characteristics that are arranged at regular intervals at nodes of an imaginary grid covering the working region (14) at least substantially completely are taken into consideration.

5. Method according to Claim 4, characterized in that in at least one method step (32) the at least one probability characteristic, in particular the at least one discretized probability characteristic, is quantized at the uncertainty level for an evaluation.

6. Method according to one of the preceding claims, characterized in that in at least one method step (42) possible partial paths of movement (44, 46) of the autonomous working device that cover at least one obstacle-free subregion (48, 50) of the working region (14) at least substantially completely, are determined, in particular in the form of a generalized round-trip problem, while taking into consideration the at least one probability characteristic, in particular the localization uncertainty, of the autonomous working device.

7. Method according to Claim 6, characterized in that in at least one method step (52) possible transitional paths (54) of the autonomous working device between a plurality of subregions (48, 50) of the working region (14) are determined while taking into consideration the at least one probability characteristic, in particular the localization uncertainty, of the autonomous working device.

8. Method according to Claim 7, characterized in that in at least one method step (56) the path of movement (22) of the autonomous working device is determined, in particular in the form of a generalized round-trip problem, for a one-time at least substantially complete coverage of each subregion (48, 50) of the working region (14) in dependence on the possible partial paths of movement (44, 46) and transitional paths (54) of the autonomous working device.

9. Autonomous working device, in particular autonomous lawn mower, with at least one navigation unit (58) for navigating within a working region (14), with at least one sensor device (60) for capturing working region data and with at least one computing unit (62), wherein the computing unit (62) is configured to evaluate at least one probability characteristic of possible states of the autonomous working device within the working region (14) that describes at least one localization uncertainty, in particular a sensing and / or movement uncertainty, of the autonomous working device for a determination of at least one path of movement (22) of the autonomous working device that covers the working region (14) at least substantially completely, characterized in that the computing unit (62) is configured so that at least one collision probability of the autonomous working device with at least one obstacle (36, 38, 40) in the working region (14) that is determined in dependence on a position of the obstacle (36, 38, 40) and the at least one probability characteristic, in particular the localization uncertainty, of the autonomous working device is taken into consideration for the determination of the path of movement (22), wherein the collision probability of the autonomous working device is determined at each possible position of the autonomous working device within the working region (14) and is taken into consideration for the determination of the path of movement (22).

10. System with at least one autonomous working device according to Claim 9 and with at least one server unit (66), in particular a cloud server, characterized in that the autonomous working device has at least one, in particular wireless, communication unit (68) for receiving from the server unit (66) control data, in particular navigation data, determined by means of a method according to one of Claims 1 to 9.

11. System according to Claim 10, characterized in that the server unit (66) is configured so as to optimize, in particular by means of a method according to one of Claims 1 to 9, at least one path of movement (22) of the autonomous working device that is determined by the computing unit (62) of the autonomous working device and to provide it to the autonomous working device.