Relative positioning method for jointly operating autonomous unmanned aerial vehicles

Abstract

The invention relates to a relative positioning method which enables unmanned aerial vehicles equipped with a UWB (Ultra Wide Band) module to position each other relatively in order to perform a task when they form a swarm performing a joint task (for example, carrying a common load in a swarm). More particularly, the invention relates to a method of relative positioning of swarm elements in order to achieve the desired task sharing and to prevent collisions in missions involving close-range flight where the swarm elements are required, to work together.

Description

[0001] RELATIVE POSITIONING METHOD FOR JOINTLY OPERATING

[0002] AUTONOMOUS UNMANNED AERIAL VEHICLES

[0003] Technical Field

[0004] The .invention relates to a relative positioning method which enables unmanned aerial vehicles equipped with a UWB (Ultra Wide Band) module to position each other relatively in order to perform a task when they form a swarm performing a joint task (for example, carrying a common load in a swarrn).

[0005] More particularly, the invention relates to a method for the relative positioning of swarm elements in order to achieve the desired task sharing and to prevent collisions in missions involving dose-range flight where the swarm dements are required to work together.

[0006] Prior Art

[0007] In current use, GPS receivers are used to position aircraft in the global coordinate system with a high margin of error. Due to the high margin of error, the required distance between the aircraft is higher than the GPS margin of error The desired level of technological maturity has not been reached in this field, especially in the joint cargo transport mission, due to the requirement of dose distance simultaneous flight.

[0008] Although radio waves such as Bluetooth and WiFi make it possible to measure distance with the help of the power values of the received signal, these measurements are not precise and vary according to the environment. For these reasons, relative positioning cannot be performed with these signal types on moving platforms. Electro-optical camera positioning methods are not suitable for use on moving platforms

[0009] UWB (Ultra. Wide Band) signals cause very little interference with existing radio systems. UWB signals are similar to noise because they have a low power spectral density, which reduces the probability of detection of UWB signals and ensures secure communication. Devices with UWB technology have a long battery life. It is much less affected by multipath channel effects and jamming effects.

[0010] UWB signals are pulse-form signals that allow signal time-of-flight measurements. In this way, they allow distance measurement with high accuracy and precision. Thanks to high accuracy and precision distance measurements, swarm robots can position each other with high accuracy.

[0011] CN113280815A discloses a method for positioning and managing swarming unmanned aerial vehicles.

[0012] CN 114949663 A discloses a method for fighting fire with swarming unmanned aerial vehicles.

[0013] When the known applications in the prior art are analysed, there is a need to develop a UWB-based relative positioning method that enables unmanned aerial vehicles to position each other relatively in order to perform the mission when they form a swarm performing a joint mission.

[0014] Obctives and Brief Description of the Invention

[0015] The object of the present invention is to develop a UWB (Ultra Wide Band) based relative positioning method that enables unmanned aerial vehicles, when forming a joint swarm, to position themselves relative to each other in order to perform the mission. The inventive method enables close-range flight of swarm aerial platforms and reduces the complexity of simultaneous swarm movements by means of relative distance-position measurements. Swarm members position each other in their own coordinate system instead of the global coordinate system. In this way, the control algorithms they have for the position they need, to maintain relative to each other can be simpler. When only one member of the swarm has global position information, all swarrn members can calculate the global position.

[0016] Detailed Description of the Invention

[0017] A system utilizing the method for achieving the object of the present invention is shown in the attached figures.

[0018] Figures;

[0019] Figure 1 : A view of a. swarm of four unmanned aerial vehicles carrying a. common load using the inventive method.

[0020] Figure 2: Schematic view of UWB based distance measurement between unmanned aerial vehicles

[0021] Figure 3: Schematic view of UWB-based distance measurement between unmanned aerial vehicles and load.

[0022] Figure 4: A schematic view of the elements on an unmanned aerial vehicle used in the inventive method.

[0023] Figure 5: Schematic view of the flowchart of the inventive method.

[0024] The parts in the figures are numbered individually, and the corresponding descriptions are given below.

[0025] 1. Unmanned aerial vehicle 2 Load

[0026] 3. Load connection

[0027] 4. UWB distance measurement

[0028] 5 Accompanying computer

[0029] 6. UWB unit

[0030] 7. Drive controller

[0031] 8. Inertial measurement unit

[0032] 9. Load carrying mechanism.

[0033] The inventive method comprises;

[0034] Unmanned aerial vehicles (1) initially generate position estimates for each other (101),

[0035] Early take-off of one of the unmanned aerial vehicles (1) and comparison of its movements with the measurements made with the UWB unit (6) and the inertial measurement unit (8) (102), Improving the position estimates of unmanned, aerial vehicles (1) with error reduction methods as a result of consistency analysis

[0036] ( 103),

[0037] Sharing of position estimates among all unmanned aerial vehicles (I)

[0038] (104),

[0039] Other unmanned aerial vehicles (1) take action and compare their movements with the measurements made by the UWB unit (6) and inertial measurement unit (8) on board and update their position estimates (105),

[0040] Calculating UWB distance measurement (4) and position estimates of unmanned aerial vehicles (1) on a accompanying computer (5) (106),

[0041] When the value of the residual value matrix between the UWB distance measurement (4) and the position estimates falls to a specified threshold value, the positioning phase of the unmanned aerial vehicles (1) ends and the driving controller (7) is switched on for position maintenance (107),

[0042] Unmanned aerial vehicles (1 ) autonomously travelling to the location of the load (2) and approaching the load (2) (108),

[0043] Each unmanned aerial vehicle (1) also performs UWB distance measurement (4) with the load (2) to finalize the position of the load

[0044] (2) (109),

[0045] Docking of the load (2) and the unmanned aerial vehicle (1) by means of the load carrying mechanism (9) via the load connection

[0046] (3) (110),

[0047] Unmanned aerial vehicles (1 ) determine and maintain their position in such a way that the weight can be carried in a balanced manner by means of driving controllers (7) for periodic position estimation and position maintenance (111).

[0048] In the method, the unmanned aerial vehicles (1) are initially allowed, to generate position estimates (101 } for each other. The initial estimate values can be randomized or, if the initial pl acement of the unmanned aerial vehicles ( 1 ) is known, these values can be used. After estimation, the system tries to improve the estimate by calculating how inaccurate this estimate is compared to the measured distance values, so the initial estimates are not critical, they only affect how fast the algorithm can reach the correct result.

[0049] Then, one of the unmanned aerial vehicles (1) takes off earlier than the other vehicles. This moving unmanned aerial vehicle (1) continuously records its movement through the inertial measurement unit (8). Here, the inertial measurement unit (8) calculates the acceleration and rotation (direction) information of the unmanned aerial vehicle ( I) and the direction and distance in which it is displaced relative to the initial coordinate. Them it measures the UWB distance measurement (4) between the other unmanned aerial vehicles (1 ) through the UWB unit (6) on itself and the UWB units (6) on the other unmanned aerial vehicles (1), and compares the data from the inertial measurement unit (8) with the measurements from the UWB unit (6) on the accompanying computer (5). (102)

[0050] Then, as a result of the consistency analysis, the position estimates of the unmanned aerial vehicles (1) are improved with error reduction methods (103). in the first stage, this process is performed only on the moving (first take-off) unmanned aerial vehicle (1). The moving unmanned aerial vehicle (1) transmits the position of each vehicle and its own position to all members via UWB messaging (104). Them when ah vehicles take off, each vehicle will compare its measurements with the UWB unit (6) and the inertial measurement unit (8) on their own, and make position estimation updates (105) and share them with the other vehicles. All these operations will be calculated on the accompanying computer (5) (106).

[0051] As a result of the consistency analysis, the improvement of the position estimates of the unmanned aerial vehicles (1) with error reduction methods (103) and the update of the position estimates of the other unmanned aerial vehicles (105) after the other unmanned aerial vehicles (1) take off are obtained, through an algorithm based on the application of the Gaussian Mixture Model (GMM) method to the Extended Kalman Filter (EKF). The details of the algorithm that provides precise position estimation are given below.

[0052] The Extended Kalman Filter (EKF) equations in the developed algorithm structure are given below. It is assumed, that the position (x position, y position) of each device used in this algorithm is calculated by using GPS together with other sensors such as inertial measurement unit (8). In the EKF algorithm, unlike the Kalman Filter structure, the observation function (formula IV) is used. The observation function is designed, to take the observations made by the measurements as corrections to the filter state vector (xfe) . The observation vector (H) used in the filter is constructed as specified by formula V to provide the relationship between the received measurement and &e state vector (formula I) xk: the value of tec predicted position on the x -- axis, yk; tiie uaiue of tiie predicted position on the y ■--- axis,

[0053] (formula II)

[0054] (V of (process noise) (formula III) P ’

[0055] (formula IV)

[0056] (h(x): Or’servcdion

[0057] / correction function for tire member ivhose location is estimated),

[0058] (xoositton: ownxposition t / iat con be measured by tee device (with / Afti)),

[0059] (yp()Sit!rte owny position that can be measured by the device (with

[0060] (formula V)

[0061] (formula VI) / Vote: Expected measurement (formula VII )

[0062] (formula. VIII) b (standard deviation of measurement noise ~5cm)

[0063] (formula IX)

[0064] (formula X) (distance measurement), fitter stare vector according to distance measurements

[0065] In order to real! se the equations given above, that is, in order to generate the position of many stationary devices whose position is unknown from distance measurements with the EKF algorithm, the unmanned aerial vehicle (1) that will perform die position estimation must perform a position change. In order to increase the speed of the positioning process before the unmanned aerial vehicle (1 ) starts Its movement, eight different EKF filters are created based on the Gaussian Mixture Model (GMM) structure, which accepts the state vector consisting of position values at eight different starting points (up, down, right, left, and diagonal points between these 4 points) around the autonomous vehicle. The state vector (xk) of each filter generated for any load at steady state consists of 8 equally spaced points on the unit circle centred, on the location of the autonomous vehicle. Among the normal likelihood values generated using the distance measurement correction value and. measurement covariance obtained for these eight filters, the most appropriate one is evaluated and the positioning result for is produced for the stationary unmanned aerial vehicle (1). In the EKF algorithm in the GMM structure, when the trace of the state covariance matrix of the selected filter ( trace(Pfe|k)) is sufficiently low, the position value found is expected to have sufficient accuracy (values less than 1 meter). For a location estimation with this precision, the POMDP (Partially observable Markov decision process) algorithm determines the movement that will optimally reduce the state covariance matrix traces created by the moving autonomous vehicle for other autonomous vehicles.

[0066] In summary, after obtaining die UWB distance measurement (4) over the mutual UWB signal, the moving unmanned aerial vehicle (1) initiates 8 EKF filters at 8 equally spaced points on the unit circle around itself based on the GMM for other unmanned aerial vehicles (1) that have not yet started moving. According to the POMDP decision process, which is used for the most optimised, cumulative reduction of the uncertainty values (covariance matrix traces) of each prediction, it performs manoeuvre / position changes that may seem random to an outside observer. When the uncertainty values (covariance matrix traces) of the other stationary unmanned aerial vehicles (1) reach the desired levels (values lower than 1 meter) thanks to the continuous UWB distance measurement (4) corrections taken from the beginning of the filter, the precise position estimation process of the other stationary unmanned aerial vehicles (1) is performed by the moving unmanned aerial vehicle (1).

[0067] At the end of the positioning phase, the driving controller (7) is switched on for position protection (107) and the vehicles continue their movements under the control of the driving controller (7). With this movement, the unmanned aerial vehicles (1) autonomously go to the point where the load (2) is located and approach the load (2) (108). In the meantime, each unmanned aerial vehicle ( 1) makes UWB distance measurement (4) with the load (2) to determine the position of the load (2) (109), and then the load (2) and the unmanned aerial vehicles (1) are docked (110) by means of the load carrying mechanism (9) through the load connection (3), and the unmanned aerial vehicles (1) determine and maintain their position ( 1 1 1) in such a way that the weight can be carried in a balanced manner thanks to the driving controllers (7) for periodic position estimation and position maintenance

Claims

CLAIMS1 A relative positioning method that allows unmanned aerial vehicles with UWB (Ultra. Wide Band) modules to position each other relatively in order to perform the mission when they form a swarm performing a joint mission, it comprises,Unmanned aerial vehicles (1) initially generate position estimates for each other (101),Early take-off of one of the unmanned aerial vehicles (1) and comparison of its movements with the measurements made with the UWB unit (6) and the inertial measuremem unit (8) ( 102),Improving the position estimates of unmanned aerial vehicles (1) with error reduction methods as a result of consistency analysis (103),Sharing of position estimates among all unmanned aerial vehicles (1) (104), Other unmanned aerial vehicles ( 1) take action and compare their movements with the measurements made by the UWB unit (6) and inertial measurement unit (8) on board and update their position estimates (105),Calculating UWB distance measurement (4) and position estimates of unmanned aerial vehicles (1) on a accompanying computer (5) (106),When the value of the residual value matrix between the UWB distance measurement (4) and the position estimates falls to a specified threshold value, the positioning phase of the unmanned aerial vehicles (I) ends and the driving controller (7) is switched on for position maintenance (107),Unmanned aerial vehicles (1 ) autonomously travelling to the location of the load (2) and approaching the load (2) (108),Each unmanned aerial vehicle ( 1) also performs UWB distance measurement (4) with the load (2) to finalize the position of the load (2) (109),Docking of the load (2) and the unmanned aerial vehicle (1) by means of the load carrying mechanism (9) via the load connection (3) (110),Unmanned aerial vehicles (1) determine and maintain their position in such a way that the weight can be carried in a balanced manner by means of driving controllers (7) for periodic position estimation and position maintenance (111) and characterized by it comprises, applying the Gaussian Mixture Model (GMM) method to the Extended Kalman Filter (EKF) when optimizing the location estimation (103) and other unmanned aerial vehicles (1 ) to update the position estimates of other vehicles after takeoff (105).

2. A relative positioning method according to claim 1 , characterized by, using formula (IV) in the extended Kalman filter, as the observation function.(formula IV)(b(x): Observation / correction function for the member whose tocation is estimated).(xnositiyn: xposition that can be measured by the device (with / Ad(yposition:own v position that can be measured by the device (with / M3. A relative positioning method according to claim 2, characterized by using the function In Formula V to relate the observation vector (H) and the received measurement to the state vector(formula V)Observation / Correction vector),4. A relative positioning method according to claim 3, characterized by, eight separate EKF filters are generated to initialise a state vector consisting ofposition, values at eight different starting points to increase the speed of the positioning process before the unmanned aerial vehicle (UAV) (!) starts its movement.

5. A relative positioning method according to claim 4; characterized by, the state vector (xfe) of each filter generated for any payload at steady state consists of 8 equally spaced points on a unit circle centred on the location of the autonomous vehicle, and the most appropriate one among the normal probability values generated using the distance measurement correction value and the measurement covariance obtained for these eight filters is evaluated to generate a positioning result for the stationary unmanned aerial vehicle (1).