CONDITION DETERMINATION OF OBJECTS IN AN OBJECT COMPLEX

Description

Technical area

[0001] The present description relates to a method for determining a correction value for a state value of an object in a group consisting of a plurality of objects, which method is designed to determine a correction value for its own state in a group consisting of several objects. Technical background

[0002] The condition of vehicles, such as land vehicles, watercraft, or aircraft, is particularly important when several vehicles are to move in a convoy. This requires that the individual vehicles have the most accurate condition values ​​possible, so that the relative condition to other vehicles in the convoy is known with a high degree of accuracy, either explicitly or implicitly.

[0003] Especially when multiple vehicles are moving in a convoy, a fusion of absolute state values, which refer to a global or external coordinate system or a quasi-static environment, and relative state values, which refer to non-static or non-stationary objects (i.e., to the other vehicles in the convoy), is necessary to ensure a consistent absolute state estimate of the vehicles in the convoy or a highly accurate relative state estimate of the individual vehicles. Please note that the terms "state estimate" and "state value" are used synonymously here, because a state value is essentially nothing more than a state estimate.

[0004] A state or state value can be understood as various information about a vehicle, such as position, speed, attitude, relative position to other objects, distance to other objects, relative speed, relative attitude, acceleration, etc.

[0005] EP 2 363 731 A1 describes a position determination system comprising: a position measuring means for receiving position measuring signals from a plurality of emission sources and for calculating position information of a first mobile object; a distance measuring means for measuring a distance from the first mobile object and to a second mobile object; a relationship determining means for determining the presence or absence of a relationship between the first mobile object and the second mobile object based on the relative position of the second mobile object based on the distance and for selecting position measuring signals from emission sources that are different from the emission sources of the position measuring signals received from the first mobile object;a position correction means for correcting a position of the first mobile object based on the position measurement signals received from the first mobile object and the position measurement signals selected by the relationship determination means;

[0006] CN 111 273 687 A describes a cooperative relative navigation method for multiple unmanned aircraft based on a GNSS observation quantity and a distance measurement between the vehicles. The method comprises the following steps: collecting and tracking the satellite signals by the GNSS receivers on all vehicles to obtain the observed GNSS quantity, and obtaining the relative distance information between the vehicles by a relative distance measurement module; transmitting and summarizing the navigation information of each unmanned aircraft through a communication module between the vehicles, and assisting the GNSS relative difference to estimate a baseline between the unmanned aircraft using the relative distance information between the vehicles;Taking an unmanned aircraft as a reference, detecting the position of each unmanned aircraft relative to the reference using a relative position vector, and integrating the position with the observed GNSS quantity into the relative navigation information; selecting another unmanned aircraft other than the reference and solving the position of the unmanned aircraft relative to the reference to optimize the original baseline solution using the relative navigation information with respect to the unmanned aircraft and the reference in an unmanned aircraft formation; and optimizing the baseline solutions with poorer precision in sequence using the existing baseline solutions and the measured quantity until each baseline solution in the formation converges, thereby completing the cooperative relative navigation with multiple unmanned aircraft.

[0007] FR 2 632 755 A1 describes a system for supporting the movement of mobile units in a formation. The system enables flight in a group formation. Each mobile unit is equipped with sensors and computing means that enable the measurement of flight parameters and data, including the absolute position, speed, altitude, and attitude of a mobile unit E1. Each mobile unit is also equipped with a transceiver to transmit the data to the other mobile units E2 to EN, Lj, Sj in the formation and to receive the same data corresponding to these other mobile units. It also has a computer that combines the data according to a defined procedure to calculate the relative positions of the mobile units and to derive the control commands for real-time correction of the mobile unit's trajectory.

[0008] Li, Hao; Nashashibi, Fawzi: Cooperative multi-vehicle localization using split covariance intersection filter. In: IEEE Intelligent transportation systems magazine, Vol. 5, 2013, No. 2, pp. 33–44, describes an approach for localizing vehicles in a networked vehicle environment. Each vehicle makes a position estimate and shares this estimate with neighboring vehicles. To update the position estimate of a group, an eigenvalue of a vehicle and the estimates of neighboring vehicles are used.

[0009] Fuji, Sae (et al.): Cooperative vehicle positioning via V2V communications and onboard sensors. In: 2011 IEEE vehicular technology conference (VTC Fall). IEEE, 2011, pp. 1–5, describes an approach in which multiple vehicles exchange data in a cooperative process to determine the position of surrounding vehicles. Description

[0010] It can be considered a task to increase the accuracy of a relative and absolute state determination of objects in a group of objects, or to determine an absolute state determination of objects that is consistent with regard to the recorded relative state values.

[0011] This object is achieved by the subject matter of the independent claims. Further embodiments emerge from the dependent claims and the following description.

[0012] According to the invention, a method is provided for determining a correction value for the state of an object. The object is part of a group consisting of several objects.The method comprises the following steps: detecting, by each object of the network, an absolute state of its own in an existing coordinate system; detecting, by a first object of the network, a relative state of the first object to one or more other objects of the network; transmitting, by the first object, the detected relative state and the detected absolute state of its own via a data connection to one or more of the other objects of the network; exchanging detected relative states and detected absolute states in the network of objects so that all objects have the same uniform information level; calculating a correction value by each object for all objects of the network based on the uniform information level; and applying, by an object, its own correction value to its own detected absolute state. The correction value . δx i is calculated in such a way that resulting new absolute state values x new,i , which are in accordance x new,i = x old,i + δx i from an old condition value x old,i and a correction value δx i result in relative condition values of a first object i to a second object j and thus reconstruct a relative constellation of the objects of the compound with respect to the absolute state values ​​by applying the following equation: z ⌣ ij = h x old , j + δx j − x old , i + δx i where h is a function which calculates the relative state values ​​from a difference between the absolute states of the objects i and j calculated; where Eq. 1 is set up for all relative state values ​​recorded in the network, resulting in a system of equations which is used to determine the correction values δx i is solved and all correction values ​​are determined by the system of equations δx i are linked to each other and can only be solved holistically, whereby all absolute and relative state values ​​and, if applicable, associated covariances must be transferred in order to create an equal, uniform information level for all objects of the network; whereby the first boundary condition is that a weighted sum of the correction values ​​results in zero, according to W Q − 1 ∑ i = 1 n Q i − 1 δx i = 0 where n represents the number of objects in the cluster, W Q the standardization with W Q = ∑ i = 1 n Q i − 1 and the weighting is the inverse of the covariance of the correction value Q i − 1 is chosen; whereby a second boundary condition is added to Eq. 1, which limits the weighted distances between the old and new absolute states; whereby the covariance of the correction value used in the first boundary condition Q i is solved using Eq. 3, which results from a linearized form of Eq. 1: H ij Var δx j − δx i ︸ Q i + Q j − 2 Cov Q i Q j H ij T = Var z ⌣ ij − h x old , j − x old , i where H ij = ∂ h x ∂ x x = x j − x i ; where a third boundary condition is introduced, according to which the sum of the weighted covariances of the correction values ​​represents a spherical ellipse, which ensures the optimality of the solution according to Var W Q − 1 ∑ i = 1 n Q i − 1 δx i = W Q − 1 ∑ i = 1 n Q i − 1 Var δx i ︸ Q i Q i − T = I reflects, whereby Q i is approximated with the covariance of the old state or the system of equations is solved by iterative methods.

[0013] The method can be applied to determine correction values ​​for a state value for a group of objects. The group of objects corresponds to a plurality of objects. The group of objects can contain mobile and stationary objects, consist only of mobile objects, or consist only of stationary objects.

[0014] A group consisting only of mobile objects can, for example, be a group comprising several vehicles, e.g. land vehicles or aircraft.

[0015] An example of a group consisting solely of stationary objects is a group of signaling beacons positioned near a runway that emit a signal to assist vehicle navigation or aircraft approach. This approach can be used, for example, to support the alignment of a group of stationary ground-based radar units.

[0016] The group of objects represents a network of objects. Mobile objects are, for example, vehicles, such as aircraft or land vehicles, that move in a network. The goal of the network is to achieve a consistent absolute state determination of the individual objects, so that the relative accuracy derived from the absolute state determination is as accurate as possible.

[0017] Each object records its absolute state. Recording its absolute state means acquiring and using sensor values ​​and / or data values ​​(e.g., data values ​​received via a data connection or read from memory, where the latter may correspond to a default value or initialization value) to determine its absolute state. The source of the values ​​retrieved and used, as well as any processing steps applied to these values, are irrelevant. The values ​​used to record its absolute state can be measured values ​​(e.g., sensor values), values ​​received via a data connection, or arbitrary values, such as an initialization value.The values ​​can be available and used as raw data, or they can first undergo one or more processing steps prior to being used to determine the absolute state. This absolute state is recorded in an external coordinate system, for example, using navigation satellites and a global coordinate system. The absolute state can contain information such as position, speed, attitude, etc.

[0018] The relative state is recorded from one, several, or all objects relative to another, several other, or all other objects in the group of objects. It can be performed using optical, acoustic, or electromagnetic detection units, e.g., an optical camera, a thermal imaging camera, an ultrasonic sensor, a radar system, a lidar system, etc. The relative state indicates the state of one object relative to another object. Generally speaking, the relative state is determined based on measured values ​​and / or received data.

[0019] The relative state and the absolute state are preferably transmitted between the objects in the network via a wireless connection. For this purpose, multiple exclusive connections can be established between two objects. However, it is also conceivable for the objects in the network to provide data to the other objects in a broadcast mode (one to several or one to all).

[0020] Each object executes a system of equations to determine the correction value. Using this system of equations, each object determines the state correction value for every other object in the network. However, only its own correction value is applied to correct its own absolute state value.

[0021] The correction value is a value that is applied to the absolute own state, for example by adding the correction value to the absolute own state, or multiplying the absolute own state by a factor, or replacing the absolute own state by a new absolute state.

[0022] Alternatively, it is conceivable that an object or a central unit, which does not have to belong to the network but has a data connection to the network and each individual member of the network, assumes a master role. All values ​​(recorded absolute state values, relative state values ​​recorded by the objects, etc.) are transmitted to this central unit. The central unit solves the system of equations and transmits the individual correction values ​​to the individual objects via a data connection.

[0023] If a mobile object has not left its assigned movement corridor, taking into account uncertainties in the state values, and specified minimum distances to other objects have not been exceeded, a control system can perform its function based solely on the corrected state values. In this case, there is no need to use derived relative state values.

[0024] The approach described here thus constitutes a method for the consistent absolute state determination of multiple objects (and further, thereby, for the derived relative state determination with respect to the other objects) in a network consisting of a plurality of objects, which are designed to determine a correction value for their own absolute state within a network of objects. Provided that all vehicles remain within an individual movement corridor, taking into account the uncertainty of their own absolute state values, the method enables control of the vehicles moving in formation based solely on their own absolute state values, since these absolute state values ​​of the members of the network are consistent with one another, and therefore a relative state is redundant.

[0025] According to one embodiment, in addition to the detected relative states and the detected absolute states, a covariance associated with each state value is also exchanged.

[0026] Alternatively, the covariance associated with a state value can be estimated by the receiving members (the objects of the network or the central unit) so that it does not need to be transmitted.

[0027] According to a further embodiment, the correction values ​​are calculated such that the absolute state error of the composite of objects is minimized.

[0028] According to a further embodiment, the correction value is used in a filter of at least one object to correct a state value of this object.

[0029] The state value is a mapping or representation of the state of an object. The state value can contain multiple individual values, so that the totality of the individual values ​​represents a multitude of individual pieces of information and describes the entire relevant state of an object. The correction value is therefore used locally within an object to correct that object's own state value.

[0030] In particular, all members of the network correct their state values ​​using the correction values ​​simultaneously to avoid inconsistencies until all members have made the correction.

[0031] The filter used here could, for example, be a Kalman filter. Other suitable filters may be used.

[0032] Simultaneous correction of the state values ​​means that the correction values ​​are corrected by all members within a specified period of time or the correction is made within a deterministic number of working cycles.

[0033] According to a further embodiment, a state value of the absolute state and / or the relative state contains information about one or more of the following parameters of an object: position, speed, attitude, acceleration, angular rate, relative position, relative distance to one or more other objects, relative speed to one or more other objects, relative attitude to one or more other objects.

[0034] For example, an optical camera can be used to determine the relative position to other objects by determining and using azimuth and elevation.

[0035] According to the invention, a correction value is calculated such that resulting new absolute state values ​​reconstruct the detected relative state values ​​of a first object i to a second object j and thus a relative geometry of the objects of the composite with respect to the absolute state values ​​by applying the following equation: z ⌣ ij = h x old , j + δx j − x old , i + δx i where δx i Correction values ​​are calculated in such a way that the resulting new absolute states x new,i = x old,i + δx i ) the recorded relative condition values from object i to object j, and thus the relative geometry of the composite.

[0036] According to the invention, the boundary condition is that the sum of the correction values ​​is zero.

[0037] The sum of the correction values ​​can be a simple sum of the correction values. Alternatively, the sum can be weighted by the inverse of the covariance of the correction values.

[0038] According to the invention, a method is provided for determining a correction value of an absolute state of an object in a group consisting of several objects.The method comprises the following steps: detecting, by each object of the network, an absolute state of its own in an existing coordinate system; detecting, by a first object of the network, a relative state of the first object to one or more other objects of the network; transmitting, by the first object, the detected relative state and the detected absolute state of its own via a data connection to a central unit, so that the central unit contains all detected absolute and relative states of the objects of the network; calculating a correction value for the state of each object of the network by the central unit based on the absolute and relative states of all objects of the network; transmitting the correction values ​​to the respective object of the network; applying, by each object of the network, its own correction value to its own detected absolute state. The correction value . δx i is calculated in such a way that resulting new absolute state values x new,i , which are in accordance x new,i = x oLd,i + δx i from an old condition value x old,i and a correction value δx i result in relative condition values of a first object i to a second object j and thus reconstruct a relative constellation of the objects of the compound with respect to the absolute state values ​​by applying the following equation: z ⌣ ij = h x old , j + δx j − x old , i + δx i where h is a function which calculates the relative state values ​​from a difference between the absolute states of the objects i and j calculated; where Eq. 1 is set up for all relative state values ​​recorded in the network, resulting in a system of equations which is used to determine the correction values δx i is solved and all correction values ​​are determined by the system of equations δx i are linked to each other and can only be solved holistically, whereby all absolute and relative state values ​​and, if applicable, associated covariances must be transferred in order to provide the central unit with the same uniform information level; whereby the first boundary condition is that a weighted sum of the correction values ​​results in zero, according to W Q − 1 ∑ i = 1 n Q i − 1 δx i = 0 where n represents the number of objects in the cluster, W Q the standardization with W Q = ∑ i = 1 n Q i − 1 and the weighting is the inverse of the covariance of the correction value Q i − 1 is chosen; whereby a second boundary condition is added to Eq. 1, which limits the weighted distances between the old and new absolute states; whereby the covariance of the correction value used in the first boundary condition Q i is solved using Eq. 3, which results from a linearized form of Eq. 1: H ij Var δx j − δx i ︸ Q i + Q j − 2 Cov Q i Q j H ij T = Var z ⌣ ij − h x old , j − x old , i where H ij = ∂ h x ∂ x x = x j − x i ; where a third boundary condition is introduced, according to which the sum of the weighted covariances of the correction values ​​represents a spherical ellipse, which ensures the optimality of the solution according to Var W Q − 1 ∑ i = 1 n Q i − 1 δx i = W Q − 1 ∑ i = 1 n Q i − 1 Var δx i ︸ Q i Q i − T = I reflects, whereby Q i is approximated with the covariance of the old state or the system of equations is solved by iterative methods.

[0039] The method according to this aspect corresponds to the method described above, with the difference that the correction value is not determined by the individual members of the network, but by a central unit that assumes a master role. The central unit can be a member of the network itself, or it can only have a data connection to the network, but is not itself a member of the network.

[0040] In general, the central unit performs the same steps to determine the correction value as described above with reference to the first variant of the procedure. The difference is that in this variant of the procedure, the calculation is performed centrally and at a single location rather than decentrally, and the members of the network are then provided with the corresponding correction values.

[0041] The features described above with reference to the first variant of the procedure also apply analogously to this second variant of the procedure, even without the said features being repeated here.

[0042] According to a further aspect, a vehicle is provided. The vehicle has a detection unit, a first filter unit, a second filter unit, and a monitoring unit.The detection unit is designed to detect absolute and relative state values ​​of the vehicle, wherein the detection unit has one or more sensors for determining an absolute and a relative state, and wherein the detection unit has a data interface which is designed to transmit its own relative and absolute state values ​​and to receive relative and absolute state values ​​from objects in a network. The first filter unit is designed to process the absolute and relative state values ​​detected by the detection unit in order to determine a corrected absolute state value and a corrected relative state value of the vehicle with respect to other objects in the network using the method according to one of the embodiments described above.When determining the corrected absolute and relative state values, a correction value is used, which was determined with the aid of a correction value calculation unit based on absolute state values ​​and relative state values ​​to other objects. The second filter unit is configured to process the absolute and relative state values ​​acquired by the acquisition unit in order to determine a relative state value to other vehicles as an output value. The monitoring unit is configured to monitor a status of the first filter unit and the second filter unit.The monitoring unit is configured to forward the corrected absolute state value and the corrected relative state value from the first filter unit if a normal operating state has been determined for the first filter unit, otherwise to forward the output value from the second filter unit for the relative state value if a faulty operating state has been determined for the first filter unit.

[0043] It is also conceivable that the monitoring unit forwards the output value of the second filter unit instead of the corrected state values ​​of the first filter unit if an estimated relative accuracy of the second filter unit is better than the estimated relative accuracy of the first filter unit.

[0044] The vehicle receives state values ​​from mobile objects (vehicles, e.g. land vehicles or aircraft) or from stationary objects which together with the vehicle form a network, so that the vehicle knows the relative and absolute state values ​​of other objects in order to incorporate these state values ​​into the calculation of the correction value.

[0045] According to one embodiment, the monitoring unit is designed to classify the status of the first filter unit as a faulty operating state if no data connection is established via the data interface of the detection unit.

[0046] The output value of the first filter unit is only used, i.e., forwarded by the monitoring unit, if the first filter unit is functioning correctly and also receives relative state values ​​and absolute state values ​​from other objects. Otherwise, the relative state values ​​are forwarded by the second filter unit, whereby the second filter unit is designed such that data collected by its own platform is sufficient to determine the relative state values, i.e., relative state values ​​and absolute state values ​​from other platforms (objects) do not need to be used to determine a relative state value.

[0047] According to one embodiment, the vehicle is an aircraft.

[0048] The aircraft may be manned or unmanned, in particular an aircraft used to carry out any type of mission in conjunction with several other aircraft. Short description of the characters

[0049] The following examples are described in more detail with reference to the accompanying drawings. The illustrations are schematic and not to scale. Like reference numerals refer to like or similar elements. They show: Fig. 1 a schematic representation of the steps of a method according to an embodiment. Fig. 2 a schematic representation of the functional modules of a vehicle according to another embodiment. Fig. 3 a schematic representation of the steps for calculating a correction value for a vehicle state in a vehicle network according to a further embodiment. Detailed description of implementation examples

[0050] Fig. 1 shows schematically the steps of a method for determining a correction value of a state of an object in a group consisting of several objects.

[0051] In a first step 11, each object in the network records its own absolute state value in an existing coordinate system. In a further step 12, a first object in the network records the relative state of this first object to one or more other objects in the network. In a further step 13, the recorded relative state and its own absolute state are transmitted by the first object via a data connection to one or more of the other objects in the network. In a further step 14, recorded relative states are exchanged in the network of objects so that all objects have the same uniform information level. In a further step 15, each object calculates a correction value for all objects in the network based on the same uniform information level.In a further step 16, an object applies its own correction value to its own recorded absolute state.

[0052] Thus, the method provides a correction value for the absolute state value itself, achieving a consistent absolute state estimate across the entire cluster. Furthermore, the method minimizes the error of the absolute state values ​​across the cluster.

[0053] The state of an object can refer in particular to the speed and position of an object, such as a land vehicle, a watercraft, or an aircraft. The absolute state contains information about the position, speed, attitude, etc., while the relative state contains information about the relative position, relative distance, relative speed, and relative attitude to other objects, etc.

[0054] Especially when several vehicles are moving in a convoy, it is helpful to determine the absolute state of the vehicles with the highest possible relative accuracy, from which the relative states can then be derived. The recorded relative state values ​​are processed with the absolute state values ​​to obtain a correction value, which is then applied using a filter or a predictor correction method to correct the absolute state value.

[0055] The absolute state can be estimated using one or more navigation methods in a filter. For example, satellite navigation methods such as GPS, GLONASS, Beidou, and Galileo can be used, which can be summarized under the general term global navigation satellite system (GNSS). Other navigation methods that can be used here include inertial navigation or environment-correlating navigation methods such as Terrain Referenced Navigation (TRN). The inertial navigation system outputs a state of the vehicle and is typically fused with GNSS navigation methods, whereby the inertial navigation system provides the state values ​​with a higher update rate and less noise than a GNSS method can provide. The position, velocity, and attitude solution of an inertial navigation system exhibits a so-called drift, i.e.The output solution degrades over time, which is inherent in the inertial navigation method, and therefore requires correction. GNSS navigation methods also contain measurement errors that may require correction.

[0056] To determine a suitable correction value, at least one object in the object network records its relative state with respect to other objects in the network and transmits this relative state, along with its own absolute state and the associated covariances for the recorded absolute and relative state values, to some or all of the other objects in the network via a data connection. In this context, a timestamp can also be used for the individual state values, which also requires global time synchronization between the members of the network, for example using high-precision clocks that are initially or continuously synchronized. There is a recurring exchange of measured values ​​between the members of the network. The totality of the absolute and relative state values ​​and the associated covariances is then used to determine a correction value for each object in the network.It is necessary for all objects in the network to determine the correction value(s) based on the same, uniform information. The correction value associated with each object is applied using a filter or a predictor correction method to correct its own state value.

[0057] The approach described here enables the absolute state determinations of the objects in the network to be consistent with each other, so that the relative states can be reconstructed up to an error residual and thus the determined states of the objects reflect the physical, actual relative geometry as accurately as possible, whereby the term geometry used here is not limited to the position domain, but also includes speed, attitude, etc. If both access to GNSS and the data connection between the objects in the network are available, which simultaneously provide absolute state values ​​and relative state values, this offers the best absolute and relative accuracy of the state values.If no drift-free absolute navigation method is available, but the data connection between the objects exists, at least the absolute states can still be determined consistently, so that the relative accuracy with regard to the determined absolute states is maintained, although the absolute state values ​​may have an error that can even increase over time, for example, due to the drift of an inertial navigation system. However, if at least one member of the network has access to a drift-free absolute navigation method, e.g., TRN, this information is implicitly included in the absolute state determination due to the filter and is thus exchanged with the other members of the network via the data connection and thus flows into the calculation of the correction value, so that the error of the absolute state values ​​of the entire network remains limited.If the data link fails but absolute navigation methods, such as GNSS and / or TRN, are available for the members of the cluster, the geometry of the cluster can be maintained with respect to the absolute state values, whereby the magnitude of the relative error is smaller than that of the absolute error, since the absolute navigation methods of the cluster members experience similar, local error contributions that compensate each other - this fact can be further exploited, for example by using the same navigation satellites for GNSS.If both the data link and GNSS data are unavailable due to disruption or operational reasons, and no alternative absolute navigation methods are available to the members, the absolute and derived relative state determination based on the inertial navigation system will initially be maintained, but will degrade over time - if necessary, the monitoring unit will switch to the second filter if the degradation of the first filter has progressed so far that the relative accuracy of the second filter exceeds that of the first filter.

[0058] The method described here is based on the idea that consistency between the determined absolute state values ​​of the members of the cluster is sought, so that these absolute state values ​​reconstruct the recorded relative state values ​​of the cluster. Input values ​​for the flight control system are normally the absolute state values, as long as the intended (flight) corridor is not departed from, taking into account the uncertainty of the absolute states, and the minimum distances to the relative states derived from the absolute states are not exceeded. Otherwise, the flight control system regulates based on a determined relative state value derived from the absolute state values ​​or determined directly by the second filter unit.

[0059] Fig. 2 shows the functional structure of an object 100, here in the form of an aircraft. It should be noted that the reference to an aircraft does not mean that the description is limited to an aircraft. Rather, the aircraft is used merely as an example, and the principles described here apply equally to any type of vehicle, including watercraft and land vehicles, as well as a combination thereof. An aircraft can be understood as a manned or unmanned aircraft. The aircraft can also be a guided missile, a rocket, or a ballistic object.

[0060] A stationary object can be slightly modified compared to the one in Fig. 2 shown structure. Thus, a stationary object can dispense with a second filter unit and otherwise have a functionally identical structure.

[0061] The aircraft 100 has a detection unit 110, which is designed to detect a state of the aircraft and provide corresponding state values, as well as a data interface 114 for exchanging absolute and relative state values, as well as covariances of the absolute and relative state values. In addition, the aircraft 100 has a first filter unit 120 and a second filter unit 130, as well as a monitoring unit 140.

[0062] The acquisition unit 110 has one or more units 112 for absolute navigation methods, which provide or update an absolute state value. In a network of multiple mobile objects, not all mobile objects need to have a unit 112 for absolute navigation methods. For the unit 112, a GNSS navigation method and / or TRN can be used, among other things, to provide an absolute state value, and an inertial navigation system, which updates the state values ​​based on the measured accelerations and angular rates of the vehicle.

[0063] Furthermore, the detection unit has one or more sensors 116, which determine a relative state to other objects in the network. In In a network of several mobile objects, not all mobile objects need to have a sensor 116. The sensor 116 can have optical, acoustic or electromagnetic sensors, e.g. an optical camera and / or a thermal imaging camera and / or an ultrasonic sensor and / or lidar and / or radar and / or time-of-flight measurement based on the data connection. In addition, the detection unit 110 has a data interface 114, via which it can send the self-detected absolute state values ​​and / or relative state values ​​concerning the relative state to other objects to other objects and receive them from these other objects. The detection unit for the relative state, i.e. the one or more sensors 116, is connected to the data interface 114 in order to exchange data, as indicated by the arrow in Fig. 2 displayed. Relative state values ​​of the sensors 116 are made available to the computing unit 122 via this connection via the data interface 114, so that the computing unit 122 knows the relative state values ​​of its own platform.

[0064] The absolute state values ​​are determined with reference to an external coordinate system and are output in this external coordinate system or in another convertible external coordinate system. The recorded relative state values, however, refer to another network member.

[0065] The first filter unit 120 has a computing unit 122 for calculating correction values, a filter 124, e.g., a Kalman filter, with an error model 125 of the inertial navigation system, and a computing unit 126 for deriving the relative states. The first filter unit 120 receives absolute state values, relative state values ​​to other objects, as well as the associated covariances of the absolute and relative state values, the absolute state values ​​received from other objects via the data connection, including the covariance, and the relative state values, including the covariance. If the covariance is not present or is transmitted, it can also be estimated using suitable models or methods.

[0066] The second filter unit 130 is a fallback option or a backup system for the first filter unit 120. The output values ​​of the second filter unit 130 are used if the first filter unit 120 has failed or if the first filter unit 120 does not receive all the values ​​necessary for its function (this condition can also be referred to as a fault condition of the first filter unit 120). The second filter unit 130 is designed to operate solely with the data that can be determined in the aircraft 100 itself. Thus, it is not necessary for the second filter unit to receive, for example, GNSS values ​​or values ​​received via the data interface 114; instead, it is sufficient to receive only relative status values ​​supplied by sensors on board the aircraft 100.

[0067] The second filter unit 130 has a first filter 132 with a dynamic model 133 of the other aircraft and a second filter 134 with a motion model 135. The first filter 132 and the second filter 134 use relative state values ​​and absolute state values ​​from an inertial navigation system to autonomously estimate the relative states to the other aircraft in the network. For initialization and during nominal operation, the second filter unit can also access additional data, e.g., data received via the data connection or data from the first filter unit.

[0068] In an alternative embodiment, the absolute state values ​​can be omitted for the second filter unit.

[0069] The monitoring unit 140 monitors the state of the acquisition unit 110 and / or the first filter unit 120. When the components of the acquisition unit 110 acquire or receive data (the absolute navigation methods, but at least the inertial platform, optionally the sensors for the relative state values, and the data interface receives values ​​from other objects in the object network) and when the first filter unit 120 performs its functions, a state value from the first filter unit is used for the aircraft 100. However, if no data is received from other objects via the data interface 114 and the estimated relative accuracy of the first filter unit falls below that of the second filter unit, or the first filter unit 120 is not functional, the monitoring unit 140 switches to the second filter unit 130 for the relative state values ​​and outputs a corresponding status signal 144.However, in the event of a data connection unavailability, absolute state values ​​are still provided that are consistent with the absolute state values ​​of the other members at the time the failure occurred. This consistency (relative accuracy between the absolute states of the members) degrades over time to a certain limit if all members continue to have at least one absolute, drift-free navigation method available; otherwise, the consistency degrades indefinitely.

[0070] The monitoring unit 140 provides an absolute state value 142 of its own state, an associated absolute state covariance, a relative state estimate 146, and an associated relative state covariance, as well as a status signal 144.

[0071] The first filter unit 120 provides absolute state values ​​with high relative accuracy, as well as relative state values ​​derived from the absolute state values. The correction value calculation unit 122 receives relative state values, including associated covariances, and absolute state values, including associated covariances, of the host vehicle, as well as relative and absolute state values, including associated covariances, of the other vehicles, which are transmitted via the data connection, in order to calculate correction values ​​for the entire network. The host correction value is applied in the first filter unit so that the absolute state values ​​of the members reflect the actual geometry (not limited to position states) of the network as faithfully as possible, and the absolute state error of the entire network is minimized.

[0072] The filter 124 is preferably a Kalman filter and uses an error model 125 of the inertial navigation system, thus ensuring relatively long-term stability of the absolute state values ​​and the associated implicit relative accuracy in the absence of absolute and / or relative measurements and / or the data connection, provided that a consistent absolute state estimate of the members has been ensured beforehand. The filter 124 carries out the correction values ​​supplied by the computing unit 122, which take into account the absolute and relative state values, including the respective covariances of the overall network, so that the output values ​​of the filter 124 for the various network members are synchronized and consistent with each other, since these output values ​​are based on the same, uniform input values.

[0073] In the event of a fault, the second filter unit 130 operates autonomously and independently of the first filter unit 120. The second filter unit 130 is designed to operate autonomously and, in the event of a fault, uses states recorded on board the aircraft without external data, so that the second filter unit 130 can determine a relative state value even if no absolute navigation methods, such as a GNSS navigation method, are present and no data is received via the data interface. In one example, the second filter unit 130 operates on data it receives from the inertial navigation system and the sensors for relative state determination. For initialization and in the nominal case (everything is functional), external data sources such as GNSS and the data connection can also be used.

[0074] If no absolute navigation methods are available or only the inertial navigation platform is available and / or no data is received via the data interface, the absolute and thus relative implicit and / or explicit accuracy of the state value provided by the first filter unit may decrease. In this case, the monitoring unit uses the relative state value provided by the second filter unit.

[0075] The arithmetic unit 122 for calculating the correction values ​​works as follows: The correction values δx i should be calculated in such a way that the resulting new absolute states ( x new,i = x old,i + δx i ) the recorded relative condition values of object i with respect to object j and thus ultimately reconstruct the relative geometry of the compound: z ⌣ ij = h x old , j + δx j − x old , i + δx i

[0076] The so-called measurement equation h is a function which calculates the expected recorded relative state values ​​from the difference between the absolute states i and j and possibly other variables.

[0077] Equation (1) is established for all relative state values ​​recorded in the network, resulting in a system of equations that must be solved for the correction values. All correction values ​​are linked to each other by this system of equations and can only be solved as a whole. Therefore, all absolute and relative state values, and any associated covariances, must be transferred to create a uniform, consistent level of information for all members or to make it available to the central unit.

[0078] The system of equations (1)may not be solvable at first, since the relative geometry of the compound members and thus the relative state values ​​in the absolute state space can be shifted arbitrarily, which can be achieved by inserting an arbitrary offset Δx for all correction values: z ⌣ ij = h x old , j + δx j + Δ x − x old , i + δx i + Δ x = h x old , j + δx j + Δ x − Δ x ︸ = 0 − x old , i + δx i

[0079] However, since the recorded relative state values ​​have no absolute reference point, they cannot influence the (weighted) center of gravity of the composite—the weighted center of gravity must therefore remain unchanged. This boundary condition is expressed mathematically by the weighted sum of the correction values ​​being 0: W Q − 1 ∑ i = 1 n Q i − 1 δx i = 0 where n represents the number of network members, W Q the standardization with W Q = ∑ i = 1 n Q i − 1 and the weighting is the inverse of the covariance of the correction value Q i − 1 is chosen. Choosing the inverse of the covariance of the correction value as the weighting is statistically optimal with respect to the weighted center of gravity of the cluster. It ensures that the cluster members with an accurate absolute state (small covariance) are assigned a smaller correction value than the cluster members with an inaccurate absolute state (large covariance) – thus, the accurate state hardly needs to be corrected.

[0080] Since the boundary condition (2) ("weighted center of gravity remains unchanged") is defined only with respect to the weighted center of gravity of the composite, the resulting relative state values ​​can still be rotated around the weighted center of gravity in the absolute state space. To eliminate this degree of freedom, a further boundary condition is added to the system of equations (1)added which limits the weighted distances between the old and new absolute states - for example, by placing virtual springs between the old and new states with a certain neutral position to allow for some mobility and a certain spring hardness to reflect the weighting.

[0081] The boundary condition (2) introduced covariance of the correction value Q i is initially unknown and must again be solved using a system of equations resulting from the linearized form of equation (1): H ij Var δx j − δx i ︸ Q i + Q j − 2 Cov Q i Q j H ij T = Var z ⌣ ij − h x old , j − x old , i where H ij = ∂ h x ∂ x x = x j − x i .

[0082] The system of equations (3)may not initially be uniquely solvable, therefore an analogous boundary condition is introduced according to boundary condition (2). This boundary condition means that the sum of the weighted covariances of the correction values ​​represents a spherical error ellipse, which reflects the optimality of the solution: Var W Q − 1 ∑ i = 1 n Q i − 1 δx i = W Q − 1 ∑ i = 1 n Q i − 1 Var δx i ︸ Q i Q i − T = I

[0083] There Q i is not known, Q i be approximated with the covariance of the old state or alternatively the system of equations can be solved by iterative methods.

[0084] Generally speaking, the correction value is calculated in the arithmetic unit 122 and passed to the filter 124. The filter 124 determines the corrected absolute state of the object. Using the object's own absolute state and the absolute states of the other members of a cluster, the arithmetic unit 126 determines the relative state of its own object to the other members.

[0085] Fig. 3 illustrates the operation of the arithmetic unit 122 for calculating the correction value from Fig. 2 .

[0086] In Fig. 3 Three aircraft 100 are shown, numbered 1, 2, 3. These aircraft each move along an actual trajectory 4.

[0087] The illustration on the far left shows the movement of aircraft 1, 2, and 3 along trajectory 4. The recorded absolute state 5 is based on an inertial navigation system. It can be seen that this absolute state 5 deviates further and further from the actual trajectory at a later point in time (aircraft shown in dashed lines) (the so-called drift). The drift of the individual aircraft can vary. For example, the absolute state of aircraft 2 drifts to the left and the absolute state of aircraft 1 and 3 drifts to the right, with the absolute state of aircraft 3 drifting more to the right than that of aircraft 1. Furthermore, the recorded relative states between aircraft 1, 2, and 3 are indicated by lines 6.The recorded relative states including the associated covariance as well as the recorded absolute state values ​​including the associated covariance are exchanged between the aircraft via a data connection 105.

[0088] The second diagram from the left shows that the expected relative state values ​​7 are determined based on the recorded absolute state values ​​5. The correction values ​​8 are calculated by all members for all members of the network, as shown in the third diagram from the left. The members apply their own correction value so that the determined absolute state values ​​reflect the recorded relative states and so that the absolute error of the network is minimized. This restores the actual relative constellation of the members with regard to the absolute state values. Alternatively, a central unit calculates the correction values ​​of all members and transmits the correction values ​​to the corresponding members, who apply the correction value to their own absolute state.

[0089] The illustration on the right shows that the calculated correction value is used in filter 124 to correct the state value. Alternatively, a predictor-correction operating principle can be used, i.e., the correction value is applied directly. From the previous absolute state values, shown with dashed lines, new absolute state values ​​are derived for each of the aircraft 1, 2, and 3.

[0090] To calculate the correction values, the recorded relative state values ​​are compared with the expected relative states according to the absolute state values ​​of the filters. The correction values ​​correspond to the difference between a recorded and an expected relative state value, which are transformed from the absolute state space to the measurement space of the relative states using the observation matrix H, expressed by the formula Hδx -< = z̃ - h ( x̂ -< ) . The absolute state values ​​of a moving object are insufficient to determine the expected relative state values ​​with respect to moving objects. The term "moving objects" is defined as those whose state changes over time, or rather, changes non-deterministically. An example of a deterministic change would be the Earth's rotation, which, due to its deterministic nature, can be compensated for by the object itself.

[0091] The state values ​​of the moving objects are necessary to deal with the fact that the absolute states and the relative states are defined in fundamentally different coordinate systems (the absolute state values ​​are defined with respect to an absolute coordinate system, while the relative states refer to a moving object). By exchanging the absolute state values ​​between network members, the expected relative states can be calculated based on the difference between the absolute state values.

[0092] Now, based on the comparison between the recorded relative state values ​​and the expected relative state values, which are determined from the difference between the absolute state values, a correction value can be theoretically determined. However, these theoretical resulting correction values ​​cannot yet be unambiguously assigned to the network members because the superposition of multiple correction values ​​can occur in different ways. This ambiguity can be resolved by introducing the boundary condition, as outlined above, that the weighted sum of all correction values ​​is 0. The weighting is performed according to the inverse of the covariance of the correction values, as described above.

[0093] The arithmetic unit 122 for calculating the correction values ​​operates in such a way that the absolute states of the cluster members resulting from the application of the correction values ​​are determined consistently, i.e., the relative geometry of the cluster is correctly reconstructed with respect to the absolute state values. If no absolute, drift-free navigation methods are available in the entire cluster, the determined absolute state values ​​of the cluster drift as a whole because the individual solutions of the first filter unit of the cluster members are synchronized due to the correction value calculation, which is performed using the same uniform information level for all members. This drift is due to the inertial navigation systems.Statistically, the accuracy of the absolute state values ​​of the entire network, or in other words, the accuracy of the network's weighted center of gravity with respect to the absolute state values, is lower than the accuracy of the absolute state values ​​of any individual object in the network. As soon as a single aircraft in the network has an absolute, drift-free navigation method available, the determined absolute state values ​​of all members are stable over the long term due to synchronization using the correction values. Network members with sensor failures are supplied with state values ​​by other network members, namely via the data link between the network members.

[0094] The following summary provides an overview of the interaction of the objects in the network and the functions of an individual network member: A network member transmits the determined absolute state values ​​and the associated covariance as well as the recorded relative state values ​​and the associated covariance to other network members.

[0095] The computing unit 122 of the network members determines the observation matrix H of the recorded relative state values ​​of the entire network ( H ij = ∂ h x ∂ x ), determines the covariance of the correction values ​​in the state space of the entire network (Q), determines the measurement residual in the measurement space between the recorded relative state and the expected relative state with respect to the absolute state values ​​of the entire network ( δz -< = z̃ - h ( x̂ -< )) , determines the correction values ​​in the state space of the entire network ( Hδx -< = δz -< ) using the boundary condition of the weighted center of gravity of the composite ( W Q − 1 ∑ i = 1 n Q i − 1 δx i = 0 ) and determines the correction values ​​in the state space of the entire composite, additionally taking into account the second boundary condition of the springs.

[0096] The filter 124 uses the part of the correction values ​​and the correction value covariance that corresponds to its own composite member for correcting its own (estimated) state values.

[0097] The following describes relative explicit and relative implicit state estimation based on absolute state values ​​and their implications for a control system. In explicit relative state estimation, the relative state is actually calculated based on the difference between the absolute state values, whereas in implicit relative state estimation, the relative state is not actually calculated. Accordingly, a control system regulates / controls based on the relative state values ​​in the case of explicit relative state estimation, while in the case of implicit relative state estimation, the absolute state estimation is used.

[0098] The own absolute state determination with the same validity date as the state determination of the other network members is explicitly subtracted from the absolute state determinations of the other network members to obtain a relative state determination. However, the available absolute state determinations of the other network members, and thus these explicit relative state determinations, are older than the own most recent absolute state determination due to the latency of the processing components (especially the data connection) used to obtain the absolute state determination from the other network members. Alternatively, the received data can be propagated to obtain the same validity date as for the own absolute state determination, but this results in error propagation through propagation.

[0099] The second option provides that the relative state determinations are not explicitly determined by the network members. The resulting implicit relative state accuracy can only be determined in post-processing or in a simulation. As long as each network member does not leave its flight corridor (taking into account the uncertainty of the state determination), the network members do not need to explicitly know the relative state to each other, which is why no flight control laws based on relative state determinations are necessary in this case. However, this assumes that the absolute state determination of the network members is consistent, which is ensured by the first filter unit 120. List of reference symbols

[0100] 1Vehicle 2Vehicle 3Vehicle 4Actual trajectory 5Detected absolute state 6Detected relative state 7Expected relative state 8Correction value 10Procedure 11-16Procedure steps 100Vehicle 105Data connection 110Detection unit 112Unit for absolute navigation procedures 114Data interface 116Sensor (relative state determination, optical sensors, camera, lidar) 120First filter unit 122Calculation unit for correction value calculation 124Filter 125Error model for sensors 126Calculation unit for determining a relative state value 130Second filter unit 132Filter 133Dynamic model of the other vehicles 134Filter 135Motion model 140Monitoring unit 142Own absolute state determination 144Status output 146Relative state determination

Claims

1. Method for determining a correction value of a state of an object in an association consisting of a plurality of objects, wherein the method comprises the following steps: each object in the association recording (11) its own absolute state in an existing coordinate system; a first object in the association recording (12) a relative state of the first object in relation to one or more other objects in the association; the first object transmitting (13) the recorded relative state and its own recorded absolute state via a data connection to one or more of the other objects in the association; exchanging (14) recorded relative states and recorded absolute states in the association of objects so that all objects have the same uniform information status; each object calculating (15) a correction value for all objects in the association based on the uniform information status; an object applying (16) its own correction value to its own recorded absolute state; characterized in that the correction value δxi is calculated in such a way that new absolute state values xnew,i resulting therefrom, which are obtained from an old state value xold,i and a correction value δxi according to Xnew,i = Xold,i + δxi, reconstruct relative state values Žij of a first object i in relation to a second object j and therefore a relative configuration of the objects in the association in relation to the absolute state values by applying the following equation: z ⌣ ij = h x old , j + δx j − x old , i + δx i where h is a function which calculates the relative state values Žij from a difference between the absolute states of objects i and j; wherein Eq. 1 is established for all the relative state values recorded in the association, thus resulting in an equation system which is solved in order to determine the correction values δxi, and all correction values δxi are coupled to one another by the equation system and can only be solved as a whole, wherein all absolute and relative state values and, where applicable, associated covariances must be transferred in order to create an identical uniform information status for all objects in the association; wherein the first boundary condition is that a weighted sum of the correction values is zero according to W Q − 1 ∑ i = 1 n Q i − 1 δx i = 0 where n represents the number of objects in the association, WQ represents the normalization with W Q = ∑ i = 1 n Q i − 1 and the inverse of the covariance of the correction value Q i − 1 is chosen as the weighting; wherein a second boundary condition is added to Eq. 1 and limits the weighted distances between the old and new absolute states; wherein the covariance of the correction value Qi that is used in the first boundary condition is solved using Eq.3 which is obtained from a linearized form of Eq. 1: H ij Var δx j − δx i ︸ Q i + Q j − 2 Cov Q i Q j H ij T = Var z ⌣ ij − h x old , j − x old , i where H ij = ∂ h x ∂ x x = x j − x i ; wherein a third boundary condition is introduced, according to which the sum of the weighted covariances of the correction values represents a spherical ellipse which reflects the optimality of the solution according to Var W Q − 1 ∑ i = 1 n Q i − 1 δx i = W Q − 1 ∑ i = 1 n Q i − 1 Var δx i ︸ Q i Q i − T = I ; where Qi is approximated with the covariance of the old state or the equation system is solved by iterative methods.

2. Method according to Claim 1, wherein a covariance associated with each state value is also exchanged along with the recorded relative states and the recorded absolute states.

3. Method according to Claim 1 or 2, wherein an absolute error value of the association of objects is minimized for calculating the correction value.

4. Method according to one of the preceding claims, wherein the correction value is used in a filter of at least one object in order to correct a state value of this object.

5. Method according to one of the preceding claims, wherein a state value of the absolute state and / or the relative state contains information relating to one or more of the following parameters of an object: position, speed, bearing, acceleration, rotational speed, relative position in relation to one or more other objects, relative distance in relation to one or more other objects, relative speed in relation to one or more other objects, relative bearing in relation to one or more other objects.

6. Method for determining a correction value of a state of an object in an association consisting of a plurality of objects, wherein the method comprises the following steps: each object in the association recording (11) its own absolute state in an existing coordinate system; a first object in the association recording (12) a relative state of the first object in relation to one or more other objects in the association; the first object transmitting (13) the recorded relative state and its own recorded absolute state via a data connection to a central unit so that the central unit receives all recorded absolute and relative states of the objects in the association; the central unit calculating (15) a correction value for each object in the association based on the absolute and relative states of all objects in the association; transmitting the correction values to the respective object in the association; each object in the association applying (16) its own correction value to its own recorded absolute state; characterized in that the correction value δxi is calculated in such a way that new absolute state values xnew,i resulting therefrom, which are obtained from an old state value xold,i and a correction value δxi according to xnew,i = Xold,i + δxi, reconstruct relative state values Žij of a first object i in relation to a second object j and therefore a relative configuration of the objects in the association in relation to the absolute state values by applying the following equation: z ⌣ ij = h x old , j + δx j − x old , i + δx i where h is a function which calculates the relative state values Žij from a difference between the absolute states of objects i and j; wherein Eq. 1 is established for all the relative state values recorded in the association, thus resulting in an equation system which is solved in order to determine the correction values δxi, and all correction values δxi are coupled to one another by the equation system and can only be solved as a whole, wherein all absolute and relative state values and, where applicable, associated covariances must be transferred in order to provide the central unit with an identical uniform information status; wherein the first boundary condition is that a weighted sum of the correction values is zero according to W Q − 1 ∑ i = 1 n Q i − 1 δx i = 0 where n represents the number of objects in the association, WQ represents the normalization with W Q = ∑ i = 1 n Q i − 1 and the inverse of the covariance of the correction value Q i − 1 is chosen as the weighting; wherein a second boundary condition is added to Eq. 1 and limits the weighted distances between the old and new absolute states; wherein the covariance of the correction value Qi that is used in the first boundary condition is solved using Eq.3 which is obtained from a linearized form of Eq. 1: H ij Var δx j − δx i ︸ Q i + Q j − 2 Cov Q i Q j H ij T = Var z ⌣ ij − h x old , j − x old , i where H ij = ∂ h x ∂ x x = x j − x i ; wherein a third boundary condition is introduced, according to which the sum of the weighted covariances of the correction values represents a spherical ellipse which reflects the optimality of the solution according to Var W Q − 1 ∑ i = 1 n Q i − 1 δx i = W Q − 1 ∑ i = 1 n Q i − 1 Var δx i ︸ Q i Q i − T = I ; where Qi is approximated with the covariance of the old state or the equation system is solved by iterative methods.

7. Vehicle (100), in particular an aircraft, having: a recording unit (110) which is designed to record absolute and relative state values of the vehicle, wherein the recording unit (110) has one or more sensors (112, 116) for determining an absolute and a relative state, and wherein the recording unit has a data interface (114) which is designed to transmit its own relative and absolute state values and to receive relative and absolute state values from objects in an association; a first filter unit (120) which is designed to process the absolute and relative state values recorded by the recording unit in order to determine a corrected absolute state value and a corrected relative state value of the vehicle in relation to other objects in the association using the method according to one of Claims 1 to 5; a second filter unit (130) which is designed to process the absolute and relative state values recorded by the recording unit in order to determine a relative state value in relation to other vehicles as an output value; a monitoring unit (140) which is designed to monitor a status of the first filter unit (120) and the second filter unit (130); wherein the monitoring unit is designed to forward the corrected absolute state value and the corrected relative state value from the first filter unit if a normal operational state has been identified for the first filter unit (120), or otherwise to forward the relative state value from the second filter unit if an incorrect operational state has been established for the first filter unit (120) or if an estimated relative precision of the second filter unit is better than an estimated relative precision of the first filter unit.

8. Vehicle (100) according to Claim 7, wherein the monitoring unit (140) is designed to categorize the status of the first filter unit as an incorrect operational state if no data connection is set up via the data interface (114) of the recording unit (110) .