CABLE-BASED PHOTOGRAMMETRIC ROBOT
Patent Information
- Application Number
- FR2019007402
- Authority / Receiving Office
- FR · FR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2019-07-03
- Publication Date
- 2024-12-13
- Estimated Expiration
- 2039-07-03
Smart Images
Figure 00000015_0000 
Figure 00000015_0001 
Figure 00000015_0002
Abstract
Description
Title of the invention: CABLE-BASED PHOTOGRAMMETRIC ROBOT technical field
[0001] The present invention relates to the field of photogrammetry and more particularly to a method for parameterizing a cable-driven photogrammetric robot. Prior art
[0002] Photogrammetry is a well-known technique for creating a 3D representation of a scene using the parallax obtained between images acquired from different viewpoints. This technique is commonly used in industrial metrology.
[0003] When the objects to be measured are large, a photogrammetry robot can be used to take views of the object to be reproduced from different points and angles. An example of a photogrammetry robot with an articulated arm is described in US patent application US-A-2019 / 0122425.
[0004] It has also been proposed to use a particular type of robot, called a cable-driven parallel robot or, more simply, a cable robot, to carry out photogrammetric measurement campaigns. An example of such a photogrammetric system can be found in the article by JD Deschênes entitled "A cable-driven parallel mechanism for capturing object appearance from multiple viewpoints" published in Proceedings of the 6th International Conference on 3D Digital Imaging and Modeling (3DIM 2007), August 21-23, 2007.
[0005] It should be noted that a cable robot is a specific type of parallel kinematic robot whose kinematic chains are not rigid segments but cables. These cables, attached to a mobile platform, are each connected to an independent winch via one or more guide pulleys. The points to which the cables are attached on the mobile platform are called attachment points. A cable therefore extends from an attachment point on the mobile platform to an anchor point on the frame (or base), represented by the winch around which it is wound. The point of contact of the cable on the last guide pulley (or eyelet) before being attached to the mobile platform is called the exit point (exilpoint) of a cable. The position and orientation of the mobile platform are thus determined by the lengths of the various cables between the attachment points and the exit points.
[0006] In some cases, the exit points may be movable relative to the frame so as to allow an adaptive configuration of the robot, as described in application EP-A-2982483.
[0007] A cable robot has the advantage of being able to move quickly in translation. However, adapting it for photogrammetry is a complex task. First, the inherent constraints of this type of robot must be taken into account, as certain parts of the workspace are inaccessible to the mobile platform due to collision phenomena (collisions between cables, collisions between cables and the platform, or collisions between cables and the environment). Second, it is necessary to ensure that the camera mounted on the mobile platform is sufficiently stable to allow image capture and that no cable or part of the mobile platform obstructs its field of view. Finally, the successive positions and attitudes of the camera must allow for the creation of a 3D representation of the scene and, more generally, for carrying out a campaign of position measurements of a predetermined set of points in the scene.
[0008] One object of the present invention is to provide a method for parameterizing a cable-driven photogrammetric robot that allows for the easy execution of a campaign to measure the position of a predetermined set of points in a scene. Presentation of the invention
[0009] The present invention is defined by a method for configuring a cable-driven photogrammetric robot comprising a frame, a mobile platform on which at least one camera is mounted, and a plurality of cables, each cable extending from an attachment point on the mobile platform to an exit point connected to the frame, the various cables being actuable to move and orient the platform, the positions of the exit points of the various cables being adjustable on the frame. This configuration method is specific in that: - we determine a set SeT of vectors representing all possible positions of the cable exit points, said set defining a maximum working space of the photogrammetric robot; - from a predetermined set of points to be measured in a scene, and said set of sep vectors, a set A of pairs of positions and orientations of the platform is determined to perform the measurements of the Sem points by photogrammetry; - an initial population of vectors is generated and cyclically subjected to crossover, mutation, and selection operations to obtain a new population in each cycle. These crossover, mutation, and selection operations are repeated until a stopping criterion is met. This selection operation is performed using a fitness function. The fitness function is a combination of objective functions, including at least one primary objective function evaluating a collision percentage and a second primary objective function evaluating the percentage of violation of a mobile platform equilibrium constraint, when the latter takes said different positions and setpoint orientations; - when the stopping criterion is satisfied, the output point vector corresponding to the highest value of the fitness function is selected and the output point positions are adjusted according to the values of the components of the vector thus selected.
[0010] Said collision percentage can be calculated from a collision percentage between any two cables of said plurality of cables and a collision percentage of a cable of said plurality of cables with the platform.
[0011] The percentage of violation of a platform equilibrium constraint can be calculated as the percentage of pairs of setpoint positions and orientations for which a capacity margin indicator indicates an impossibility of platform equilibrium, taking into account the maximum tensions of the cables and their attachment positions on the platform.
[0012] Advantageously, the combination of objective functions further involves a third objective function evaluating the percentage of pairs of setpoint positions and orientations for which the voltage of at least one of the cables is greater than a predetermined maximum voltage value.
[0013] The set A of setpoint position and orientation pairs can be determined so that each point to be measured of Sem belongs to the field of view of the camera for at least a first setpoint position and orientation pair of A and a second setpoint position and orientation pair of A, distinct from the first, and that the measurement of said point, obtained by photogrammetry from the images of the camera when it takes the position and orientation of the first pair, then the position and orientation of the second pair, is less than a predetermined maximum error value.
[0014] Preferably, after each selection operation, it is determined, for each of the pairs of setpoint positions and orientations, whether the camera's field of vision is obstructed by a cable, a platform element or a frame element, and that in case of obstruction, assembly A is modified by replacing this setpoint position and orientation pair with a new setpoint position and orientation pair.
[0015] Prior to modifying assembly A, it is advantageous to verify that, when the platform takes the position and orientation setpoint of the new couple, the measurement error by photogrammetry is less than said predetermined maximum error value.
[0016] After the modification of set A, the crossing, mutation and selection operations are continued cyclically, the objective functions then being evaluated on the different positions and orientations of the set A thus modified.
[0017] In all cases, the mobile platform can be in suspended configuration or in fully constrained configuration. Brief description of the figures
[0018] Other features and advantages of the invention will become apparent upon reading a preferred embodiment of the invention, described with reference to the accompanying figures, among which:
[0019] [fig. 1] schematically represents a cable-driven photogrammetric robot;
[0020] [fig.2] represents an example of a first type of attachment configuration of the mobile platform;
[0021] [fig.3A]
[0022] [fig.3B] and
[0023] [fig.3C] represent examples of a second type of mobile platform attachment configuration;
[0024] [fig.4] represents a maximum working space of the photogrammetric robot, for a given configuration of cable exit points;
[0025] [fig.5] represents a flowchart of a method for parameterizing a photo robot grammatical according to an embodiment of the invention;
[0026] [fig.6] represents all the positions and orientations of an on-board camera on the platform to conduct a measurement campaign of a scene;
[0027] [fig.7A] and
[0028] [fig.7B] represent a first and second trajectory for carrying out a photogrammetric measurement campaign. Description of the implementation methods
[0029] In the following we will consider a cable-driven photogrammetric robot as mentioned in the introductory part.
[0030] Such a robot is schematically represented in Fig. 1. It comprises a frame, 100, including here four vertical pylons, 110, 120, 130, 140, for example forming part of a metallic structure.
[0031] The mobile platform (or gondola), 150, represented here schematically as a cube, is connected to the frame by eight cables, 160, respectively attached to the platform's attachment points (here, the vertices of the cube). A pair of cables is associated with a pylon, and each cable extends from its attachment point on the platform to an anchor point (generally an actuator such as a winch, for example) via its exit point (rear pulley or eyelet). The anchor points are designated here as A1-A8 and the exit points as EX1-EX8. Advantageously, each rear pulley can pivot around the axis of the pylon on which it is mounted, so that its groove is always aligned with the cable. The height of Each return pulley can be fixed or, advantageously, independently adjusted on the frame.
[0032] The positions of the exit points of the different cables and the attachment configuration of the platform, the minimum and maximum tensions that can be exerted by the cables and the wrench of external mechanical forces, define the working space of the robot.
[0033] Different types of mobile platform attachment configurations can be considered:
[0034] In a configuration, known as a suspended configuration, an example of which is shown in Fig. 2, the cable exit points are located above the robot's workspace. The correspondence between the attachment points and the cable numbers is indicated at the bottom of the figure. In this configuration, the mobile platform is always suspended by the cables, and in a static position, no downward vertical force, other than gravity, is exerted on the platform. The weight of the platform therefore plays a crucial role in ensuring the robot's mechanical stability.
[0035] In a fully constrained configuration, various examples of which are shown in Figs. 3A-3C, the robot's workspace is contained within the polyhedron formed by the exit points of the different cables. In this configuration, the cables work in opposition, on either side of the mobile platform. Thus, for a platform represented schematically by a cube, the exit points of return pulleys mounted on the same pylon can be connected to opposite or adjacent vertices of the cube in various possible combinations. Different combinations are possible depending on how the platform's attachment points are associated with the cable exit points, as shown in the correspondence tables at the bottom of Figs. 3A-3C.
[0036] It is understood that, depending on the different possible configurations and combinations, the situations of cable collisions between themselves, with the platform or with the environment are different.
[0037] The positions of the different cable exit points, together with the cable attachment configuration on the platform, define a working space for the photogrammetric robot.
[0038] More specifically, Fig. 4 schematically represents a workspace of the photogrammetric robot. This workspace is defined as the envelope of positions of the geometric center (or isobarycenter) of the mobile platform when the cables are actuated by the winches. In other words, the workspace is that swept by the geometric center of the mobile platform when the cable lengths between their attachment points and their respective exit points take all possible values guaranteeing the static and dynamic equilibrium of the robot.
[0039] It is noted that this envelope is contained within a polyhedron (here a parallelepiped) whose vertices are the exit points of the various cables. For given positions of the exit points, the offset between the workspace and the polyhedron is due to the respective distances between the geometric center of the platform and the various cable attachment points.
[0040] It should be noted that the focal point of the camera mounted on the mobile platform may be distinct from the geometric center of the latter. In some cases, the camera's focal point may be indexed along three axes; in other words, its position may be controlled relative to a frame of reference attached to the platform. Similarly, the Euler angles defining the camera's orientation with respect to a trihedral frame attached to the platform may be controlled to vary the camera's attitude.
[0041] It should be noted that the mobile platform can be equipped with a plurality of cameras, in particular to simultaneously capture images in a plurality of directions. In this case, the preceding remarks apply to each of the cameras.
[0042] Figure 420 shows the polyhedron corresponding to the extreme positions of the exit points, that is, when these exit points are furthest apart. In the following, the surface of this polyhedron is referred to as the extremal surface of the exit points. The workspace, 410, of the photogrammetric robot corresponding to the extreme positions of the exit points is referred to as the maximum workspace.
[0043] Taking into account the extreme surface area of the exit points and the maximum working space that can be deduced from it, the parameterization method according to the present invention aims to find, for a given set of points to be measured in a scene or on an object, the optimal positions of cable exit points to operate the photogrammetric robot.
[0044] Fig. 5 schematically represents the flowchart of a method for parameterizing a photogrammetric robot according to an embodiment of the invention.
[0045] The parameterization method assumes knowledge of a certain number of geometric and physical characteristics of the photogrammetric robot.
[0046] It will be assumed in particular that the geometric and physical characteristics of the platform are known or that they have been the subject of a working hypothesis. This working hypothesis can be iterated if necessary.
[0047] Geometric characteristics here mean the dimensions of the mobile platform, the attachment configuration of the platform and its type (suspended or fully constrained configuration), the positions of the attachment points on the platform and, where applicable, the position and orientation of the camera on the platform.
[0048] The physical characteristics of the platform are understood to mean the weight of the platform, including that of its payload (i.e., the camera(s)) as well as the moment external forces acting upon them.
[0049] We also have the geometric (diameter for example) and physical characteristics of the cables, in particular their specific mass and the maximum tensions they can withstand.
[0050] In all cases, the positions of the cable exit points are among the parameters to be determined.
[0051] The geometric data of the frame are shown in Figure 510. From this geometric data, the ranges of possible positions for the various output points, defining the extremity surface of the output points, can be deduced in Figure 520. Thus, for example, in the case illustrated in Figure 4, these possible positions are located on the vertical edges of the parallelepiped 420. Other frame geometries can, of course, be considered without departing from the scope of the present invention. In particular, the ranges of possible positions for the output points are not necessarily on vertical edges.
[0052] In the following, the cable exit point vector will be any N-tuple (ep, ..., epN) where ePn represents the coordinates of the exit point with index n. Without loss of generality, we will assume that the exit points are located on vertical edges and that their possible position ranges are identical. In other words, the exit point vector can be expressed in the form (z, ..., zN) where zmin - zn - zmax = zmin, ..., N in the Oxyz frame attached to the structure, the Oz axis being the vertical axis and [zmin, zmax] being the common range of possible z values.
[0053] From the ranges of the cable exit points, the maximum working space of the photogrammetric robot can be deduced in 540, that is to say, as indicated above, the envelope of possible positions of the geometric center of the mobile platform, for one or more given orientations of the platform.
[0054] In step 530, the parameterization method also receives as input a set, Sem, of points to be measured in a scene (or more specifically, in an object). These points are defined by their coordinates in the Oxyz coordinate system, with, where applicable, a tolerance margin on these coordinates. It should be noted that this is indeed a matter of at least approximate knowledge of the positions of the points to be measured, allowing, as indicated later, the determination of the target positions and orientations of the camera (or cameras, where applicable).
[0055] From the maximum working space of the photogrammetric robot, the Sem assembly, and the characteristics of the camera, a photogrammetric simulation module determines in 550, a set of camera setpoint positions and, for each of these setpoint positions, the orientation or orientations that the camera must adopt in order to be able to measure by triangulation the points of the assembly Such a simulation module is available as a software product marketed under the name V-STARS™ by the company Geodetic Systems.
[0056] The photogrammetric simulation module ensures in particular that each point to be measured (taking into account the margin of tolerance) is well covered by a predetermined number of shots with a sufficient angular deviation to be able to carry out the triangulation with the minimum degree of accuracy required.
[0057] In other words, as long as the degree of precision constraint is not verified in 570, the set of position and orientation pairs is corrected in 550.
[0058] Fig. 6 shows an example of a set of set positions and orientations for a photogrammetric robot. It can be seen that for each set position the platform must take a plurality of set orientations to perform a picture of the scene to be measured.
[0059] The set of camera position and orientation commands is provided in 580 to the genetic algorithm, in the form of a file, A.
[0060] The genetic algorithm shown on the right of the figure has, on the one hand, the ranges of possible positions of each of the cable exit points and, on the other hand, the geometric and physical characteristics of the platform and the cables as defined above.
[0061] From the ranges of possible positions of the output points, a population of output point vectors is generated in step 525. This population of vectors can be obtained by randomly selecting a vector from the interval [zmm, zmax] within the interval [zmm, zmax], the components of which are the results of N random selections within this interval. Each vector in the population can be considered a potential solution to the problem of parameterizing the photogrammetric robot.
[0062] Once the population of vectors has been established, this population is subjected to successive operations of crossing, mutation and selection as described below.
[0063] In step 535, the crossover operation between two parent vectors of the population may consist of forming a vector whose first half is made up of the first N / 2 components of one of the parent vectors and whose second half is made up of the last N / 2 components of the other parent vector. Other cut-point positions are of course also conceivable.
[0064] In 545, the mutation operation can be performed by adding a Gaussian noise vector. For example, if we denote eP as a vector of output points at the output of the crossover operation, the mutated vector will be given by:
[0065] [Math.l] ep '-ep + toJN (0.1 ) ( 1 ) where N(0,l) is a vector whose components are iid random variables obeying a centered normal distribution, and where is the mutation rate, giving the amplitude of the Gaussian noise.
[0066] Output point vectors having a component outside the interval [zmm, zmax] are directly eliminated.
[0067] In step 555, the value of the fitness function is evaluated for each eP vector of the current population (after the crossover and mutation operations), based on the set of camera setpoint positions and orientations.
[0068] This aptitude function takes into account one or more objective function(s), including:
[0069] A first objective function, F^, represents the percentage violation of an equilibrium constraint of the platform when it assumes the different setpoint positions and orientations. This equilibrium constraint is advantageously calculated as the capacity margin index (CMI). A detailed description of the calculation of this index can be found in the article by F. Guay entitled "Measuring how well a structure supports varying extremal wrenches," 2nd Conference on Mechanisms, Transmissions and Applications, Bilbao, Spain, 2013. This index represents the minimum degree of satisfaction of the static equilibrium constraints of the platform, taking into account the wrench of the forces external to the structure (primarily due to gravity) and the wrench due to the application of cable tensions at the attachment points. It assumes knowledge of the positions of the attachment points on the platform as well as the maximum cable tensions.
[0070] This indicator is calculated for each setpoint position and orientation pair of the platform. When this indicator is positive, there are positive cable tension values lower than the aforementioned maximum values, ensuring an equilibrium position of the platform (equipped with its payload). Conversely, when this indicator is negative, there are no N-tuples of positive cable tensions lower than the aforementioned maximum values, ensuring the static equilibrium of the platform.
[0071] A second objective function, F^, represents the percentage of collision between at least two cables when the mobile platform takes the different positions and orientations of the setpoint.
[0072] A third objective function, F^, represents the percentage of collision between at least one cable and the mobile platform when the mobile platform takes the different setpoint positions and orientations.
[0073] A fourth objective function, F^, represents the percentage of collision between the mobile platform and the building environment when it takes the different positions and orientations of the setpoint.
[0074] Other objective functions may be taken into account where appropriate. Thus, for example, a fifth objective function may be used representing the percentage of violation of the maximum cable tension when the platform takes the different setpoint positions and orientations.
[0075] These objective functions generally represent a collision frequency or a violation of a constraint when the mobile platform takes the different positions and orientations of the setpoint.
[0076] The aptitude function, F, combines these K different objective functions, for example in the form of a linear combination with weighting coefficients. Thus, for example, the aptitude function could be:
[0077] [Math.2] Z' ) = ^ ^(1 — (ep D (2) where z = son( |cs different weighting coefficients.
[0078] In step 565, for each vector, eP, in the population, it is determined whether there is an obstruction in the camera's field of view when the platform assumes the different setpoint positions and orientations. This obstruction may be due to the moving platform or the cables, for example. When such an obstruction is observed for a setpoint and orientation, the combination of this point and orientation is removed from the set of position-orientation pairs of the platform. This pair can then be replaced by another pair corresponding to a different setpoint position and orientation, allowing the same field of view of the object to be covered. This situation is shown in Fig. 7B, where the circled areas indicate the setpoint positions (and, if necessary, the setpoint orientations) that were modified in Fig. 7A.If applicable, when the camera is angularly indexed, the platform's setpoint orientation can remain the same, with only the camera's orientation being modified.
[0079] The deletion or replacement of one or more position pair(s), setpoint orientation in file A is indicated by the feedback 567. If necessary, the position tolerances of the points to be measured can be checked by calling routine 560. If these tolerances are not respected, step 550 is repeated.
[0080] In the absence of obstruction of the camera's field of vision when the platform takes the different positions and orientations of A setpoint, we proceed to the selection step.
[0081] In step 575, a selection is made of the population vectors exhibiting the highest values of the fitness function. These vectors constitute a new population of vectors, defining the population of the next generation.
[0082] Finally, in step 585, we test whether a stopping criterion is satisfied, for example, if a predetermined number of iterations is reached.
[0083] In the affirmative, the cable exit point vector, ep"'", corresponding to the highest value of the suitability function is selected in 595. The components of this vector z°ptzopt, provide the respective adjustment heights of the different cable exit points along the pylons of the frame.
[0084] On the other hand, if the negative result is not achieved, we return to 535, to start a new cycle of crossover-mutation-selection.
[0085] According to one variant, the optimization using the genetic algorithm can be complemented by an optimization phase using the gradient algorithm. In this case, the genetic algorithm provides an approximate solution to the optimum. The optimum is then sought through successive iterations by calculating the gradient of the cost function at each iteration and updating the positions of the exit points in the opposite direction of this gradient, in a manner known per se.
Claims
Claims
1. Method for configuring a cable photogrammetric robot comprising a frame, a mobile platform on which at least one camera is mounted and a plurality of cables, each cable extending from an attachment point of the mobile platform to an exit point linked to the frame, the different cables being able to be actuated so as to move and orient the platform, the positions of the exit points of the different cables being adjustable on the frame, characterized in that: a set Sep of vectors representing all the possible positions of the exit points of the cables, called exit point vectors, is determined, said set defining a maximum working space of the photogrammetric robot;from a predetermined set of points to be measured in a scene, and from said set of vectors Sep, a set A of pairs of positions and orientations of the platform is determined (550) for carrying out the measurements of the points by photogrammetry;a genetic algorithm is applied to an initial population of said output vectors by cyclically applying crossover (535), mutation (545) and selection (575) operations to it, to obtain, at each cycle, a new population, the crossover, mutation and selection operations being repeated until a stopping criterion is satisfied, said selection operation being carried out by means of a suitability function, the suitability function being a combination of objective functions, including at least a first objective function evaluating a collision percentage and a second objective function evaluating the percentage of violation of an equilibrium constraint of the mobile platform, when the latter takes said different set positions and orientations;when the stopping criterion is satisfied (585), the vector of output points corresponding to the highest value of the aptitude function is selected and the positions of the output points are adjusted (595) according to the values of the components of the vector thus selected.;
2. Method for setting up a cable photogrammetric robot according to claim 1, characterized in that said collision percentage is calculated from a collision percentage between two cables any of said plurality of cables and a percentage of collision of a cable of said plurality of cables with the platform.
3. Method for parameterizing a cable photogrammetric robot according to claim 1 or 2, characterized in that said percentage of violation of a platform equilibrium constraint is calculated as the percentage of pairs of set positions and orientations for which a capacity margin indicator indicates an impossibility of equilibrium of the platform, taking into account the maximum tensions of the cables and their attachment positions on the platform.
4. Method for parameterizing a cable photogrammetric robot according to one of the preceding claims, characterized in that the combination of objective functions also involves a third objective function evaluating the percentage of pairs of set positions and orientations for which the tension of at least one of the cables is greater than a predetermined maximum tension value.
5. Method for parameterizing a cable photogrammetric robot according to one of the preceding claims, characterized in that the set A of pairs of set positions and orientations is determined in such a way that each point to be measured of 8^ belongs to the field of vision of the camera for at least a first pair of set position and orientation of A and a second pair of set position and orientation of A, distinct from the first, and that the measurement of said point, obtained by photogrammetry from the images of the camera when it takes the position and orientation of the first pair, then the position and orientation of the second pair, is less than a predetermined maximum error value.
6. Method for parameterizing a cable photogrammetric robot according to one of the preceding claims, characterized in that after each selection operation, it is determined, for each of the pairs of set positions and orientations, whether the field of vision of the camera is obstructed by a cable, an element of the platform or an element of the frame, and that in the event of obstruction, the set A is modified by replacing this pair of set position and orientation with a new pair of set position and orientation.
7. Method for setting the parameters of a cable photogrammetric robot according to claim 6, characterized in that, prior to modifying the assembly A, it is verified that when the platform takes the position and orientation set for the new pair, the measurement error by photogrammetry is less than said predetermined maximum error value.
8. Method for parameterizing a cable photogrammetric robot according to claim 6 or 7, characterized in that after modification of the set A, the crossing, mutation and selection operations are continued cyclically, the objective functions then being evaluated on the different positions and orientations of the set A thus modified.
9. Method for configuring a cable photogrammetric robot according to one of the preceding claims, characterized in that the mobile platform is in a suspended configuration.
10. Method for parameterizing a cable photogrammetric robot according to one of claims 1-8, characterized in that the mobile platform is in a fully constrained configuration.