Method and device for drone landing

Abstract

A system and method for determining a landing zone for a drone, comprising obtaining images representing at least a search perimeter for a landing zone; obtaining 3D information representing at least said search perimeter; selecting a potential landing zone within said search perimeter by analysing the obtained images and 3D information; obtaining data relating to the moisture content and / or hardness of the ground in the potential landing zone, acquired by means of at least one contact sensor; and determining a measure of suitability of the ground in the potential landing zone for landing the drone, based on the data relating to the moisture content and / or hardness of the ground, and optionally the images and / or 3D information.

Description

Description Title of the invention: Method and device for drone landing Technical Field

[0001] This presentation concerns drones, commonly called UAVs for "unmanned aerial vehicle" in English, and more specifically drone landing systems. Previous technique

[0002] Automatic drone landing, particularly in areas with little or no prior knowledge of the terrain—in other words, in areas where topographic or satellite imagery is insufficient or nonexistent—is a crucial step in drone piloting. Indeed, landing is a critical phase that, if not properly executed, can lead to serious damage. This can be especially true during flights conducted following a natural disaster or in conflict zones.

[0003] In addition, it is sometimes necessary to plan, preferably automatically, an emergency landing, for example when a malfunction is identified.

[0004] In order to allow for safer landings, numerous solutions have been proposed.

[0005] Existing methods rely primarily on the analysis and fusion of two-dimensional images and three-dimensional (or more) data of terrain to be analyzed for potential landing. The images and data are provided by cameras and LiDAR-type devices, the latter being mounted on the drone. These methods generally involve segmenting the images to classify the pixels, merging these segmented images with LiDAR data to obtain a 3D model of the terrain, potentially identifying areas unsuitable for landing, and then selecting one or more suitable landing areas based on the 3D model and predefined criteria.The predefined criteria are classically related to the slope of the ground, the presence of obstacles, the roughness of the ground and / or a distance to be covered in view of a landing.

[0006] These known methods, however, cannot guarantee that a drone will land safely. In particular, it is not uncommon for images and 3D models to provide only surface information about the soil type of the analyzed terrain, which is insufficient to prevent, for example, the drone getting stuck in the mud. Generally speaking, there is no way to estimate the potential dynamics of the terrain.

[0007] US 2023 / 0315127 describes a method for controlling the landing of an aerial vehicle from images and scans.

[0008] CN 113 401336 describes a drone support for drone landing on various surfaces. [0009JII There is therefore a need for improved systems and methods enabling drone landing, preferably autonomously, while preventing or at least further limiting damage to the drone. Description of the invention

[0010] This presentation concerns a method for determining a landing zone for a drone, comprising the following steps: a) obtaining images representing at least one search perimeter of a landing zone; b) obtaining 3D information representing at least said search perimeter; c) selection of a potential landing zone within said search perimeter by analysis of the images and 3D information obtained; the method being characterized in that it further comprises the following steps: d) obtaining data relating to soil moisture and / or hardness in the potential landing area acquired by means of at least one contact sensor; e) determining a measure of soil suitability in the potential landing area for drone landing, the suitability measure being determined from the data relating to soil moisture and / or hardness, and possibly from images and / or 3D information.

[0011] The invention can advantageously allow the estimation of soil density before the drone lands and thus anticipate a possible getting stuck of the drone, a possible landslide, or any other soil instability that could lead to any damage or the impossibility of a subsequent takeoff of the drone.

[0012] In general, adding contact sensor(s) to the drone ultimately allows for a wider range of measurements, providing a better characterization of the environment that would be impossible with remote data acquisition using only cameras and radar devices such as Lidar. In particular, some variations may be invisible to cameras and Lidar.

[0013] Thus, unlike known methods that rely solely on measurements to identify whether the area is a city, forest, field, mountain, etc., and to determine geometric constraints such as slopes or roughness, the invention makes it possible to assess soil density or soil moisture. This allows, for example, the evaluation of risks of instability or landslides, particularly in the case of landing zones covered with snow or fine sand, where images from a camera and information from a LiDAR could lead to the conclusion that these surfaces are suitable for landing.

[0014] The method according to the invention may further include a step of releasing said contact sensor.

[0015] Preferably, the method according to the invention includes a step of detecting, prior to release, any moving object(s) in the potential landing zone, such as an animal or a vehicle, the release step being conditional upon the absence of any moving object in said potential landing zone. Preferably, the detection step is performed using images acquired with a thermal camera. In particular embodiments, the detection step is performed using images acquired with an RGB camera. The detection step can also be performed using images acquired with both an RGB camera and a thermal camera. These embodiments are not limiting.

[0016] In some embodiments, the detection covers a larger area than the potential landing zone and may include an estimation of the trajectories and / or speeds of the detected moving object(s). This advantageously allows for anticipating the entry of an object into the potential landing zone.

[0017] If a moving object is detected, the process can be paused for a predetermined duration, or until the object crosses the potential landing zone. Regardless of the area covered by the detection, estimating the trajectories and speeds of the moving object(s) can be used to assess the waiting time required for the object(s) to exit the potential landing zone.

[0018] In some embodiments, the presence of an object in the area may lead to the selection of a new potential landing zone or even the determination of an instruction to move the drone to a new search perimeter.

[0019] These steps advantageously prevent any collision between the contact sensor and a moving object during release.

[0020] In some embodiments, the process involves bringing the drone closer to the potential landing area to allow for a safe drop.

[0021] In other specific embodiments, the method involves bringing the contact sensor, and possibly the drone, closer to the landing zone, so as to allow the sensor to make contact with the ground in said contact zone, without actually landing or releasing the drone. Detecting the presence of moving objects can also help prevent collisions between the drone and a moving object during such a landing. rapprochement.

[0022] Images can be acquired using at least one camera, preferably at least one RGB camera and / or one infrared camera. Images are composed of pixels and represent two-dimensional information.

[0023] 3D information can be acquired using at least one 3D acquisition device, preferably one incorporating a lidar, a stereoscopic camera, a monocular camera, and / or a thermal camera. The 3D information can be in the form of a point cloud. The 3D information can represent depths or altitudes.

[0024] The contact sensor may include a force sensor and / or a humidity sensor.

[0025] The suitability measure can be calculated using a neural network that takes as input data relating to soil moisture and / or hardness and is trained to provide as output a measure estimating the suitability of the soil for a landing.

[0026] The neural network can be trained to provide a binary output, for example, "adequate" or "unadequate." Alternatively, the neural network can be trained to provide a probability that the area is adequate. In specific embodiments, the neural network can be trained to provide an estimated value for soil hardness.

[0027] For example, the neural network is a convolutional neural network or multilayer perceptron (MLP).

[0028] The neural network is pre-trained using a database compiled according to the desired output and the inputs to be used. A person skilled in the art knows how to build such a database and how to train such a neural network. Any known method can be used.

[0029] As an alternative to neural networks, the fit measure can be calculated from physical models.

[0030] The suitability measure can be a binary value, a probability, or a predictive measure of soil hardness or density.

[0031] The method may further include an assessment of the suitability measure to determine, based on the suitability measure, whether the potential landing area is suitable for the drone and, consequently, generate a flight instruction for the drone indicating whether or not to land. The method may also include determining a landing trajectory for the drone within said area when the potential landing area is deemed suitable. The method may include selecting a new potential landing area or moving the drone to a new search perimeter when the potential landing area is deemed unsuitable.

[0032] The assessment can be carried out manually, remotely, by an operator to whom the adequacy measurement has been transmitted. Alternatively, the assessment can be carried out automatically, that is to say by computer means, for example by comparing the adequacy measurement with a threshold value.

[0033] In certain embodiments of the invention, when the potential landing area is considered not to be adequate for a landing, a flight instruction may be determined, for example so as to return the drone to its starting point, otherwise to its take-off point.

[0034] All or part of the steps of the process according to the invention can be implemented by computer, preferably all the steps of the process are implemented by computer.

[0035] Thus, the invention also relates to a computer program comprising code instructions which, when implemented, allow the execution of the steps of the process according to the invention.

[0036] The invention further relates to a recording medium readable by means of a computer comprising a computer program according to the invention.

[0037] In particular, the computer program can be programmed logic, the recording medium may include, among other things, FPGA-type components.

[0038] Correspondingly, the invention also relates to a drone landing zone detection system comprising: - An acquisition module comprising: at least one camera, at least one 3D information acquisition device, and at least one contact measurement unit comprising at least one contact sensor capable of acquiring humidity and / or hardness data; - A processing unit configured to implement a process according to the invention, based on the data and information acquired by the acquisition module; - Means of communication between the acquisition module and the processing unit.

[0039] Each contact measuring unit may include a force sensor and / or a humidity sensor. The system may include a plurality of contact measuring units, preferably substantially identical, in particular to allow for successive drops, with only one measuring unit being dropped at each drop.

[0040] At least one 3D acquisition device may include a Lidar, a stereoscopic camera, and / or a thermal camera, for example a monocular.

[0041] At least one camera may include an RGB camera and / or an infrared camera.

[0042] The invention also relates to a drone comprising a detection system according to the invention.

[0043] The aforementioned features and advantages, as well as others, will become apparent upon reading the detailed description that follows. This detailed description refers to the attached drawings. Brief description of the drawings

[0044] The attached drawings are schematic and are primarily intended to illustrate the principles of the presentation.

[0045] In these drawings, from one figure to another, identical elements (or parts of elements) are identified by the same reference symbols.

[0046] [Fig. 1] Figure 1 represents a drone comprising a detection system according to the invention,

[0047] [Fig. 2] Figure 2 schematically illustrates a functional architecture of a detection system according to the invention,

[0048] [Fig. 3] Figure 3 illustrates an example of steps in a process according to the invention,

[0049] [Fig. 4] Figure 4 illustrates an example of steps for implementing a potential landing zone selection step,

[0050] [FIG. 5] Figure 5 illustrates an example of the hardware architecture of a processing unit of a detection system according to the invention. Description of the implementation methods

[0051] To make the explanation more concrete, an example of a detection system and detection method according to the invention is described in detail below, with reference to the accompanying drawings. It should be noted that the invention is not limited to these examples.

[0052] A drone 1 comprising a detection system 2 according to the invention is shown in Figure 1. Figure 2 represents a functional architecture of the detection system 2.

[0053] The drone landing zone detection system 2 comprises an acquisition module 10, a processing unit 20 and communication means 30 of the acquisition module 10 with the processing unit 20.

[0054] The acquisition module 10 comprises a set of sensors 12 called "remote sensors" or "contactless sensors", and at least one contact measurement unit 14. The acquisition module 10 is mounted on the drone 1.

[0055] The non-contact sensor assembly 12 comprises at least one camera 121, for example an RGB and / or infrared camera; and at least one 3D acquisition device 122, for example a Lidar. Preferably, the sensor assembly comprises an RGB camera, an infrared camera, and a Lidar.

[0056] As an alternative or additional to the Lidar type device, it is possible to use a stereoscopic system or a monocular camera which provide information in three dimensions (3D).

[0057] The array of 12 non-contact sensors allows for data acquisition while remaining at a distance from a search area, even when the drone is in flight. Analysis of this data can enable the initial selection of a potentially suitable landing zone. In other words, the acquisitions made by this set of sensors make it possible to consider for the rest of the process only areas that meet the criteria classically considered to define a landing zone, in particular the absence of obstacles in the area considered, such as humans, vehicles, trees, animals, a sufficiently flat and clear ground, relatively gentle slopes.

[0058] At least one contact measuring unit 14 is capable of acquiring soil moisture and / or hardness data. Each contact measuring unit 14 may include a force sensor and / or a moisture sensor.

[0059] Preferably, at least one contact measuring unit 14 is capable of acquiring soil moisture and hardness data, each contact measuring unit 14 comprising a force sensor and a moisture sensor.

[0060] The force sensor, for example a compression sensor, can advantageously be placed at the top of the jettisonable module, in other words at the end intended to come into contact with the ground first.

[0061] At least one contact measurement unit 14 allows acquisitions to be made to estimate the hardness of the soil or its moisture content, or both, in order to assess whether the soil is suitable for a drone landing.

[0062] At least one contact measurement unit 14 can be arranged to be released by the drone. The contact measurement unit 14 can then have a fuselage shaped into a streamlined form optimized for ground penetration. The fuselage is advantageously made of an impact-resistant alloy to withstand impacts when the non-contact sensor makes contact with the ground after release.

[0063] When at least one contact measuring unit 14 is arranged to be dropped from the drone 1, the detection system 2 may advantageously comprise a plurality of contact measuring units 14 intended to be dropped independently of each other, each unit 14 being able to consist of a plurality of sensors.

[0064] Each contact measurement unit 14 may include an inertial measurement unit (IMU) capable of measuring linear accelerations and angular velocities.

[0065] Each contact measuring unit 14 may include a power supply, for example a battery.

[0066] The communication means 30 may include a wireless transmission system included in each contact measurement unit 14, configured to allow the data collected by the jettisonable contact measurement unit to be sent to the processing unit 20 then located remotely, for example in the drone.

[0067] The communication means 30 may be wired or wireless. In some embodiments, the communication means 30 may include wireless transmission systems and wired transmission systems.

[0068] In some embodiments of the invention, the processing unit 20 is not integrated into the drone 1 but can be located remotely.

[0069] The processing unit 20 is configured to implement a process 100 for determining a landing zone as described below and illustrated in Figure 3.

[0070] Processing unit 20 includes, in particular: - a module 22 for obtaining 3D images and information representing at least one search perimeter configured to implement a step E10 of a process 100 according to the invention; - a module 24 for selecting a potential landing zone within said search perimeter by analyzing images and 3D information obtained, configured to implement a step E12 of a process 100 according to the invention; - a module 26 for obtaining data relating to the humidity and / or hardness of the soil in the potential landing area configured to implement a step E50 of a process 100 according to the invention; - a module 28 for determining a measure of soil suitability in the potential landing area with the landing of the drone, from data relating to soil moisture and / or hardness, and possibly images and / or 3D information, configured to implement a step E60 of a process 100 according to the invention.

[0071] The processing unit 20 may also include all or part of the following modules, not shown in the figures: - a dropping module for a contact measurement unit, and optionally a moving object(s) detection module, respectively configured to implement a step E40 and a step E30 of a process 100 according to the invention; - a suitability measurement evaluation module configured to implement a step E70 of a process 100 according to the invention; - a landing trajectory determination module configured to implement a step E80 of a method 100 according to the invention.

[0072] An example of a method for determining a drone landing zone is now detailed with reference to Figure 3.

[0073] Process 100 is implemented during the drone flight, essentially in real time.

[0074] Procedure 100 can be triggered as soon as a drone landing instruction is received. This landing instruction can be planned or triggered by the detection of any drone malfunction.

[0075] Following receipt of this instruction, at step E00, process 100 includes a step E10 of obtaining 3D images and information, the images having been acquired by means of at least one camera, preferably an RGB and / or infrared camera, and the 3D information having been acquired by means of at least one 3D acquisition device, for example of the Lidar type.

[0076] The process 100 further includes a step E20 of selecting a potential landing zone, determined from the analysis of the images and 3D information obtained in step E10.

[0077] An example of the implementation of this step E20 is illustrated in figure 4.

[0078] Step E20 may include the following sub-steps: - a substep E201 of semantic segmentation of the pixels, or groups of pixels, of the IMG images obtained in step E10. This segmentation can be performed using a deep neural network specifically trained to perform this segmentation. The classes, or labels, assigned to the pixels by the neural network can be binary, i.e., the pixels can be classified as "landable" or "non-landable". Alternatively, the number of classes can be greater than 2; for example, the classes can identify various components of the images and include, for example, all or part of the following classes: "tree", "built structure", "road", "agricultural area", "water", "sand", etc. This list is not exhaustive; - a sub-step E202 for labeling the 3D information obtained in step E10, notably in the form of a point cloud, for example acquired using a Lidar. For this, the 3D information can be projected onto the segmented IMGs in step E201, for example using an extrinsic calibration matrix linking the coordinate system of the 3D information acquisition device to that of the camera. The classes or labels associated with the pixels or groups of pixels are then associated with the points of the point cloud by correspondence; - a substep E203 for updating the labeled 3D information based on the drone's movements. Substep E203 may include analyzing the 3D information at regular intervals and identifying correspondences between the different 3D information in order to estimate the drone's movements. These movement estimates can then be used to update the drone's position and the labeled 3D information in real time; - a substep E204 for generating an elevation map. The generation of the elevation map may involve transforming labeled and possibly updated 3D information into an elevation map, commonly called a "Digital Terrain Model" (DTM), by projecting the points of the point cloud onto a horizontal (XY) plane. Each point is used to calculate an elevation based on its Z coordinate, representing the height of the terrain at that point. The surface resulting from the points projected onto the horizontal plane is then discretized into a regular grid, and for each cell of the grid, the elevation value is determined, either by interpolation or by taking the average of the points present in that cell. In addition to the elevation, a label can be associated with each cell based on the labeling obtained in substep E202, which assigns a label to each point.Each cell can thus be labeled according to the majority of the points it contains, allowing the different classes to be identified (for example "road",. "vegetation", "building", etc.). Cells for which no point is measured can be marked as "not measured" or "empty". The resulting elevation map offers not only a topographic representation of the terrain, but also a semantic one, as well as an indication of uncovered areas; - A substep E205 of elevation map filtering can be performed to identify and exclude areas unsuitable for landing. This E205 filtering step can be semantic filtering, retaining only areas of the elevation map corresponding to classes, or labels, deemed compatible with drone landing. To do this, each cell is examined based on the cell's class, or the classes of the points within the cell, possibly taking into account points in neighboring cells within a radius equivalent to the size of the drone, with a safety margin added to this radius. In some embodiments, if at least one pixel of a cell belongs to a class deemed incompatible with landing (for example, "obstacle," "dense vegetation," etc.), or if the cell is marked as "not measured" or "empty," then the cell associated with that point is excluded; - a sub-step E206 for estimating one or more parameters for cells not excluded in sub-step E205. Preferably, sub-step E206 includes the estimation of all or part of the following parameters: roughness, slope, highest step. Roughness estimation: Substep E206 may include calculating a roughness measurement for each cell of the elevation map. For each cell, a neighborhood can be defined, preferably corresponding to the size of the drone. From the points included in this neighborhood, a local plane is determined; in other words, a plane is generated from these points, for example, by applying an optimization method that minimizes the deviation of these points from the plane. Roughness can be calculated as the standard deviation of the residual errors, that is, the difference between the elevation values ​​at each considered point on the elevation map and those predicted by the local plane. This measurement allows for the quantification of ground irregularities in the studied area. Slope estimation: Substep E206 may involve calculating a slope for each cell of the elevation map by analyzing neighboring cells within a radius corresponding to the drone's size. Similar to roughness estimation, a local plane can be determined. The slope can be determined by calculating the inclination of this local plane relative to a horizontal plane. Highest step estimation: Substep E206 may include calculating a highest step for each cell in the elevation map. A neighborhood can be defined around the cell in question, preferably with a radius equal to the drone's diameter. The highest step of a cell can be calculated by determining a local plane based on the points within the neighborhood, comparing the elevations of the points within the neighborhood to the local plane. The highest step can be defined as the maximum value of these elevation comparisons relative to the local plane. This estimation allows for the detection of steep drops or obstacles. A substep E207 filters cells based on the parameter(s) determined in substep E206 to determine a potential landing zone.Substep E207 preferably includes filtering the cells of the elevation map based on three parameters: slope, roughness, and highest step. If the value of any of these parameters exceeds a predefined threshold, the corresponding cell is preferably considered unsuitable for landing. This predefined threshold can be adjusted according to safety constraints and / or the drone's capabilities. Specifically, the threshold is adjusted to exclude areas that are too steep, too rough, or have excessively high steps. This filtering ensures that only areas meeting certain criteria are retained for potential landing. Substep E207 also includes the selection of an optimal landing zone (ZPOT).The optimal landing zone can be selected using conventional optimization methods, taking as input all or part of the previously defined parameters and the distance between the drone and the cells under consideration. To this end, the process may involve assigning a value, called the difficulty value, to each cell of the elevation map not excluded by the implementation of the preceding substeps. For each cell, the difficulty value can be determined from the distance between the drone and that cell and at least one parameter defined in the preceding substep, preferably from all the parameters defined in the preceding substep calculated for that cell. The difficulty value can be obtained by applying the following formula: [Math 1] vt = p * di + X a k with V| the difficulty value for cell i, di pk-ref the distance between cell i and the drone, pk the parameters for cell i, and pk ref threshold values ​​for each of the parameters pk, 6 and o k These are weighting coefficients. The optimal landing zone may correspond to the cell with the lowest difficulty value. In the example described, a low difficulty value represents an area with a relatively flat, relatively smooth surface, relatively few major obstacles, and relatively close to the drone's current position.

[0079] The optimal landing zone can be defined as the potential landing zone.

[0080] When no optimal landing zone is found, the process may include a step E92 for generating a flight instruction that directs the drone to move to a new search perimeter. The flight instruction may provide a movement path, preferably following a predefined pattern. For example, the predefined pattern may be a zigzag or spiral pattern.

[0081] In general, when no optimal landing area is found, thus providing no potential landing area, no data relating to soil moisture and / or hardness in the potential landing area is acquired by means of the contact sensor.

[0082] Once a potential landing zone has been selected, the process includes a step E50 for obtaining data relating to soil moisture and / or hardness in the potential landing zone, acquired by means of at least one contact measuring unit comprising at least one contact sensor. This data acquisition step E50 is preferably preceded by a step E40 for dropping the contact measuring unit into the potential landing zone.

[0083] In order to carry out a safe drop, the process may include pre-drop and post-selection steps of the potential landing zone consisting of verifying that the conditions for dropping a unit of contact measurement are met.

[0084] In particular, process 100 may include a step E20 for verifying the drone's altitude to enable the release of the contact sensor and the contact measurement unit incorporating said sensor. The altitude may be calculated from 3D information. The altitude may be adjusted if necessary.

[0085] In particular, method 100 may include a step E30 for detecting moving object(s) in the potential landing zone to enable the release of the contact sensor and the contact measurement unit incorporating said sensor. To avoid collision of the contact sensor with moving objects such as humans, vehicles, or animals, it is preferable to perform detection prior to release. This detection can be carried out using images, preferably images acquired with a thermal camera, and / or 3D information. Any known method may be used. The detection can be performed in an area including the potential landing zone, preferably with a safety margin that can be defined according to the speed of the detected object.When a moving object is detected in the potential landing zone, the release of the contact sensor can be delayed, for example, until the moving object crosses said zone. In some embodiments, the release is delayed by a predefined time. Optionally, a new detection can be performed once the predefined time has elapsed. In some embodiments, when a moving object is detected in the zone, an instruction to modify the potential landing zone (step E90) or to modify the search perimeter (step E92) can be generated.

[0086] Once the contact sensor is in contact with the ground, and preferably after the inertial measurement unit (IMU) data has stabilized, the sensor acquires data relating to soil hardness and / or soil moisture, preferably both. The acquired data is then sent to the processing unit. The data preferably includes forces, inertial data, and moisture levels. The data can be sent via a wireless communication system, such as a radio communication system.

[0087] The process 100 also includes a step E60 of determining a measure of soil suitability in the potential landing area with the drone landing, from data relating to soil moisture and / or hardness, and possibly images and / or 3D information.

[0088] The determination of the adequacy measure can be carried out by neural network or by modeling via physical equations.

[0089] The neural network can be trained using experimental data, such as data collected in a laboratory. This experimental data might come from dropping a contact sensor onto different soil types with varying moisture levels, while simultaneously measuring the ground's actual capacity to support a drone landing. Optionally, soil hardness measurements can be obtained during the experiments using a durometer. The neural network can be trained to take as input a sequence of inertial data, forces, and moisture levels. Additional input information can be provided to the neural network, such as classes / labels derived from semantic segmentation and / or images showing the potential landing zone.In some embodiments, the trained neural network can predict in a binary fashion whether the potential landing zone is suitable for landing or not. In other embodiments, the neural network can be trained to output a measure of ground hardness. The suitability measure can correspond to the output provided by the neural network.

[0090] An example of determining a measure of adequacy using physical equations is described below. The following steps can be implemented: - Measure an impact force using a force sensor, included in the contact sensor; - Calculate the pressure exerted by the contact sensor by dividing the measured force by the contact area of ​​the contact sensor with the ground, the pressure representing the resistance of the ground to the impact of the sensor; - Evaluate ground stability by analyzing post-impact motion using IMU data. This data can be integrated, for example, via a Kalman filter to detect vertical and horizontal displacements of the contact unit. Thresholds can be established for these displacements, as well as for the duration of the motion; prolonged motion may indicate ground instability, while minor adjustment immediately after impact may be considered acceptable; - Estimate the pressure exerted by the drone during landing based on landing speed, drone weight, and the drone's contact area with the ground; - Compare the estimated pressure exerted by the drone with the ground resistance, i.e., the pressure exerted by the contact sensor. The result of this comparison can correspond to the suitability measurement. Alternatively, the suitability measurement can be binary and correspond to "adequate" if the ground resistance is greater than the estimated pressure exerted by the drone during landing, and "inadequate" otherwise. - Optionally, check the soil moisture level. If the soil moisture level is above a certain threshold, the suitability measurement may then indicate unsuitability of the soil in the potential landing area.

[0091] Alternatively, only the humidity level can be considered.

[0092] Finally, the method 100 may include a step E70 for evaluating the suitability measure to determine whether the potential landing area is suitable for the drone landing. The method may also include determining a landing trajectory for the drone within said area when the potential landing area is deemed suitable, and selecting a new potential landing area or moving the drone to a new search perimeter. The method may further include a landing step E80 for the drone, when the potential landing area is deemed suitable, preferably following a predetermined landing trajectory.When the potential landing zone is assessed as unsuitable, the process may involve selecting a new potential landing zone, possibly after modifying the search perimeter and obtaining new 3D images and information, then dropping a new contact measurement unit in the new potential landing zone and determining a new suitability measurement. These steps may be repeated a predetermined number of times until a suitable landing zone is found, or until the system has no more contact measurement units available for deployment.

[0093] In particular embodiments, the processing unit 20 has the hardware architecture of a computer, as shown in Figure 5. It should be noted that some elements of this architecture may be confused with corresponding elements of the drone.

[0094] More specifically, the processing unit 20 may include a PC processor, ROM read-only memory, RAM random-access memory, and communication means.

[0095] Read-only memory constitutes a recording medium readable by the processor and on which is recorded a computer program according to the invention, comprising instructions for the execution of the steps of the method of determining 100 of a landing zone according to the invention detailed above and of which an example is in particular illustrated in figures 3 and 4.

[0096] Although the present invention has been described with reference to specific embodiments, it is evident that modifications and changes can be made to these examples without departing from the general scope of the invention as defined by the claims.

[0097] In particular, the selection of a potential landing zone can be performed by any known method and is in no way limited to the example in Figure 4. Similarly, a detection system according to the invention is not limited to the illustrated example and may, in particular, comprise more than one contact measuring unit. The invention is also not limited to an onboard contact measuring unit intended to be dropped. The invention also relates to a drone comprising an onboard contact measuring unit arranged to be lowered, brought into contact with the ground of the potential landing zone, and then raised again, for example, via a spring system or a retractable cable.

[0098] Furthermore, individual features of the various embodiments illustrated / mentioned can be combined in additional embodiments. Therefore, the description and drawings should be considered in an illustrative rather than restrictive sense. [0099JII is also evident that all the characteristics described with reference to a process are transposable, alone or in combination, to a device, and conversely, all the characteristics described with reference to a device are transposable, alone or in combination, to a process.

Claims

Demands

1. A method (100) for determining a landing zone for a drone comprising the following steps: a) obtaining (E10) images representing at least one search perimeter of a landing zone; b) obtaining (E10) 3D information representing at least said search perimeter; c) selection (E12) of a potential landing zone within said search perimeter by analysis of the images and 3D information obtained; the process being characterized in that it further comprises the following steps: d) dropping (E40) at least one contact sensor (14) and obtaining (E50) data relating to soil moisture in the potential landing area and / or soil hardness in the potential landing area acquired by means of said at least one contact sensor (14); e) determination (E60) of a measure of soil suitability in the potential landing area with the drone landing, from data relating to soil moisture and / or soil hardness, and possibly images and / or 3D information.

2. Method according to claim 1 comprising a step (E30), prior to drop, of detecting moving object(s) in the potential landing zone, such as an animal or a vehicle, the drop step being conditioned on the absence of moving object in said potential landing zone, preferably further the detection step being carried out from images acquired by means of a thermal and / or RGB camera.

3. A method according to any one of claims 1 and 2 wherein the suitability measure is calculated by means of a neural network taking as input data relating to soil moisture and / or soil hardness and trained to provide as output a measure estimating the suitability of the soil for a landing.

4. A method according to any one of claims 1 and 2 wherein the adequacy measure is calculated from physical modeling.

5. A method according to any one of claims 1 to 4 comprising the evaluation (E70) of the suitability measure so as to determine whether the potential landing area is suitable or not for the landing of the drone, the method comprising determining a landing trajectory for the drone in said area when the potential landing area is considered to be adequate, the process involving the selection (E90) of a new potential landing area or the determination (E92) of an instruction to move the drone to a new search perimeter when the potential landing area is considered inadequate.

6. Computer program comprising code instructions which, when implemented, enable the execution of the steps of the process (100) according to any one of claims 1 to 5.

7. A drone landing zone detection system (2) comprising: - An acquisition module (12) comprising: at least one camera (121), at least one 3D acquisition device (122), at least one contact measurement unit comprising at least one contact sensor (14) capable of acquiring soil moisture and / or soil hardness data; - A processing unit (20) configured to implement a process according to any one of claims 1 to 5; - Communication means (30) of the acquisition module with the processing unit.

8. System according to claim 7 in which said contact measuring unit comprises a force sensor and / or a humidity sensor.

9. Drone (1) comprising a system (2) according to any one of claims 7 and 8.

Images