• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    A method of perception of material bodies comprising the following steps: a) Acquisition of a plurality of distance measurements of said material bodies from one or more sensors (C1 ... CNC); b) Acquisition or calculation of a prior occupation probability value (P (o)) of the cells of a occupation grid; and c) Application of an inverse sensor model on said occupancy grid to determine a probability of occupancy of a set of cells of said grid d) Construction of a consolidated occupancy grid by merging the probabilities of occupation estimated in step c); characterized in that each said inverse sensor model is a discrete model, associating with each cell of the corresponding occupation grid, and for each distance measurement, a probability class chosen within the same set of cardinality finished and identified by an entire index; and in that said step d) is implemented by means of integer calculations performed on the indices of the classes of probabilities determined during said step c), and as a function of said prior occupation probability value. Hardware body perception system adapted to implement such a method.
    公开号:FR3062924A1
    申请号:FR1751246
    申请日:2017-02-16
    公开日:2018-08-17
    发明作者:Frederic Heitzmann
    申请人:Commissariat a lEnergie Atomique CEA；Commissariat a lEnergie Atomique et aux Energies Alternatives CEA；
    IPC主号:
    专利说明:

    Holder (s): COMMISSIONER OF ATOMIC ENERGY AND ALTERNATIVE ENERGIES Public establishment.
    Extension request (s)
    Agent (s): MARKS & CLERK FRANCE General partnership.
    PTY METHOD AND SYSTEM FOR CONTEXTUALIZED PERCEPTION OF MATERIAL BODIES.
    FR 3,062,924 - A1
    Process for the perception of material bodies comprising the following steps:
    a) Acquisition of a plurality of distance measurements of said material bodies from one or more sensors (Ci ... C NC );
    b) Acquisition or calculation of an a priori occupation probability value (P (o)) of the cells of an occupancy grid; and
    c) Application of a reverse sensor model to said occupancy grid to determine a probability of occupation of a set of cells of said grid
    d) Construction of a consolidated occupation grid by merging the occupation probabilities estimated during step c);
    characterized in that each said inverse sensor model is a discrete model, associating with each cell of the corresponding occupation grid, and for each distance measurement, a probability class chosen within the same set of cardinality finished and identified by a whole index; and in that said step d) is implemented by means of whole calculations carried out on the indices of the probability classes determined during said step c), and as a function of said a priori occupancy probability value.
    System for perceiving material bodies adapted to implement such a method.
    V MTD2
    METHOD AND SYSTEM FOR CONTEXTUALIZED PERCEPTION OF MATERIAL BODIES
    The invention relates to a method and a system for perceiving and estimating the position of material bodies performing, effectively in terms of computing power and energy consumption, a multi-sensor fusion.
    The term “material body” means any substance or material object having an individuality and which can be detected and identified by an appropriate sensor. Thus, material bodies are considered inanimate objects, whether natural or artificial, plants, animals, human beings, but also liquid or solid particles suspended in the air, such as clouds, or even masses. liquids or gases.
    The invention applies in particular to the field of navigation of robots, drones, autonomous vehicles, etc. and more generally to that of perception.
    With the explosion of computing means that can be integrated into a robot, robotics applications have multiplied in recent years, from industrial production to home automation, from space and underwater exploration to consumer toy drones. The tasks performed in robotic applications have gradually become more complex, implying more and more often for robots to be able to evolve in unknown environments; this has made it increasingly important to develop means and techniques of perception, that is to say allowing the discovery and interpretation of the surrounding space. An important application that uses perception in robotics is navigation, which consists in setting a target of destination for a robot, and letting it go there taking care to avoid unknown and potentially mobile obstacles; the robot is then responsible for planning its trajectory itself. A typical example, the subject of intense research, is the autonomous car.
    To allow a knowledge of the whole environment by limiting the blind spots as much as possible, and to compensate for a possible defect in a sensor, use is generally made of the integration of multiple sensors. When several sensors, possibly of different types, cover the same space, it is necessary to be able to combine the information extracted from each of them: this is called multi-sensor fusion.
    There are two main families of perception techniques: geometric methods, which aim to identify the geometry of objects in the surrounding space, and those based on an occupancy grid, which aim to determine whether a certain location is occupied by a obstacle (more generally, by a material body). The invention relates to techniques based on the occupation grid.
    The theoretical foundations of multi-sensor perception and fusion methods based on occupation grids are described in the article by A. Elfes, "Occupancy grids: a stochastic spatial representation for active robot perception" (Sixth Conference on Uncertainty in Al, 1990). This publication is not concerned with the practical implementation of methods, the direct application of which would require complex floating point calculations.
    The article by K. Konolige "Improved occupancy grids for map building" (Autonomous Robots, 4, 351-367, 1997), that by J. Adarve et al. "Computing occupancy grids from multiple sensors using linear opinion pools", (Proceedings - IEEE International Conference on Robotics and Automation, 2012), that of T. Rakotovao et al. “Real-time power-efficient integration of multi-sensor occupancy grid on many core” (2015 International Workshop on Advanced Robotics and its Social Impact, June 30, 2015) and that of E. Kaufman et al. "Bayesian Occupancy Grid Mapping via an Exact Inverse Sensor Model" (2016 American Control Conference - ACC, Boston Mariott Copley Place, July 6-8, 2016, Boston, pp. 5709 - 5716) describe improvements in grid-based techniques. 'occupation. Here again, the implementation of these techniques requires massive recourse to floating point calculation.
    Documents US 2014/035775, FR 2006/050860 and DE 102009007395 describe multi-sensor perception and fusion methods and systems based on occupancy grids, applied to the autonomous driving of land vehicles. All these methods require, for their implementation, floating point calculations.
    However, the use of floating point calculation requires significant resources in terms of computing power, which are hardly compatible with the constraints specific to on-board systems. As a reminder, the floating point format - defined by the IEEE 754 standard - represents a number by means of three elements: a sign (1 bit), a mantissa (23 or 52 bits) and an exponent (8 or 11 bits). Performing calculations using numbers represented in floating point is much more complex (that is, requires many more elementary operations) than performing calculations on integers. This therefore requires the use of a faster processor and / or a dedicated hardware accelerator, with an unfavorable impact in terms of cost, size and electrical consumption.
    The article by T. Rakotovao et al. “Multi-Sensor Fusion of Occupancy Grids based on Integer Arithmetic” (2016 IEEE International Conférence on Robotics and Automation - ICRA, Stockholm, May 16-21, 2016, pp. 1854 - 1859) describes a multi-sensor perception and fusion process based on an occupancy grid, using only integer calculations without introducing errors compared to approaches using floating point calculations. Such a process is also described in patent application in France 15-58919 of September 22, 2015 and in international application PCT / EP2016 / 072530 of September 22, 2016. This approach allows the use of simple on-board computing devices, not supporting not necessarily floating point operations; even if floating point calculations are supported, it makes it possible to reduce the energy consumption of the calculation device, by avoiding or greatly limiting the effective use of such calculations.
    The invention aims to improve this method, and more particularly to make the occupancy grid calculated using sensor measurements more relevant, without increasing the complexity of the calculations performed in real time.
    According to the invention, this aim is achieved by making the calculation of the occupation probabilities dependent on an a priori probability whose value depends on the context. For example, in the case of an application to autonomous or assisted driving, the a priori probability (that is to say before any measurement) that obstacles are present near a vehicle is more or less high in depending on traffic density.
    On the other hand, the method described in the aforementioned article by T. Rakotovao et al. uses a probability of occupation a priori fixed and equal to 0.5, which translates a total ignorance of the context. The generalization in the case of any posterior probability, which can therefore change depending on the context, is not obvious. The inventors have managed to make such a generalization, and to note that it does not imply an increase in the complexity of the calculations to be carried out in real time.
    The value of the context-dependent a priori probability of occupation can be determined in several different ways. For example :
    It can be chosen manually by a user, for example the driver - or passenger - of a vehicle.
    It can be transmitted to a calculation device on board by a control station, for example a traffic monitoring center.
    It can be calculated based on position and / or time. For example, always considering the case of an application to assisted or autonomous driving, the probability of occupation a priori may be higher in the morning and late afternoon than in the middle of the day or at night; similarly, it will be reasonably higher in an urban area than in a rural area.
    It can also be calculated from measurements obtained from a dedicated sensor. For example, it can be a non-directional radar, which detects the presence of obstacles without locating them precisely.
    An object of the invention is therefore a method of perceiving material bodies comprising the following steps, implemented by a computer or a dedicated digital electronic circuit:
    a) Acquisition of a plurality of distance measurements of said material bodies from one or more sensors;
    b) Acquisition from an external device, or calculation from at least one signal received from the outside, of at least one value of probability of a priori occupancy of the cells of an occupancy grid;
    c) Application, at each said distance measurement, of an inverse model of the corresponding sensor on said occupancy grid providing a discretized spatial representation of an environment of said sensor, to determine a probability of occupation by a material body of a set of cells of said occupancy grid, each said inverse sensor model being a discrete model, associating with each cell of the corresponding occupancy grid, and for each distance measurement, a probability class chosen inside of the same set of finite cardinality, each said probability class being identified by an integer index; and
    d) Construction of a consolidated occupation grid, each cell of which has an occupation probability calculated by merging the occupation probabilities estimated during step c), the occupation probability of each cell of the consolidated occupation being determined by means of whole calculations carried out on the indices of the probability classes determined during said step c), and as a function of said or of said a priori probability of occupation.
    According to particular embodiments of such a method: Said step b) can comprise the acquisition of said or at least one said value of probability of occupation a priori by means of a user interface device .
    Said step b) may comprise the acquisition of said or at least one said a priori occupancy probability value by means of a radio receiver. In this case, said or at least one said a priori probability of occupation value, acquired by means of a radio receiver, can be transmitted by a control station (SST).
    At least said sensors are on board a land vehicle, and said or at least one said a priori occupancy probability value, transmitted by a control station, can be a function of traffic density on at least one axis. road on which said land vehicle travels.
    Said step b) can comprise the calculation of said or at least one said a priori occupancy probability value from a spatial or spatio-temporal positioning of said sensors.
    Said step b) can comprise the calculation of said a priori occupancy probability value from at least one presence signal of said material bodies from one or more sensors other than the sensors used to implement said step a ).
    Said set of finite cardinality of probability classes can be formed by the union of one or more subsets such that, during said step d), the fusion of two probability classes belonging to the same subset provides a result also belonging to said subset.
    Said set of finite cardinality of probability classes can constitute a non-uniform discretization of the probability interval [0, 1], with a discretization step increasing between 0 and 0.5, then decreasing between 0.5 and 1. More in particular, said set of finite cardinality of probability classes, designated by Gp (0) , can be defined by: Gp (0) = {(p n ) <ne Z}, Z being the set of relative integers, the classes of probabilities p n being defined as follows:
    Po = P (Oi);
    pi = p;
    p n + i = F (Pi, p) Vn> 1 ίο
    Ρ (θί) 2 (ι-ρ)
    P-i =
    P (Oj) 2 + (l-2P (Oi)) p '
    Pn-1 = F (Pn, Pl) Vn <-1 where P (Oi) is said a priori occupation probability value of the index cell "i" of the occupation grid, p a parameter of value strictly between 0 and 1-P (o) and F a function of occupation probability fusion; and in which, during said step d), the fusion between two classes of probabilities p n , p m θ being calculated by application of the following equation: F (p n , p m ) = p n + m
    Said step d) may include the implementation of landmark changes to construct said consolidated occupancy grid from probabilities of occupation of occupancy grid cells associated with non-co-located sensors.
    The method can also include a preliminary stage of construction of the inverse models of at least one said sensor on the corresponding occupation grid, implemented by application of the following equation:
    p (z | Oi) P (Oi)
    P (Oi | z) = ρ (ζ | θί) Ρ (θί) + p (z | vi) [l - Ρ (ο0] or:
    • P (oj | z) represents the probability of occupancy of the cell with index "i" of the occupancy grid, said cells being ordered by increasing distance from said sensor;
    • P (Oj) is said a priori probability of occupation value of the index cell "i" of the occupation grid;
    • ρ (ζ | θί) is the probability density of the measure “z >> provided by the sensor when the cell of index“ i ”of the occupation grid is occupied, function of Ρ (ο,) and the direct sensor model; and • p (z | Vj) is the probability density of the measure “z >> supplied by the sensor when the i cell of index“ i ”of the occupation grid is empty, function of Ρ (ο, ) and the direct sensor model.
    Advantageously, the probability of occupation a priori can be the same for all the cells of the occupation grid.
    Another object of the invention is a system for perceiving material bodies comprising:
    at least a first input port for receiving a plurality of signals representative of distance measurements of said material bodies from one or more sensors;
    at least a second input port for receiving a signal representative of at least a value of probability of occupation a priori of the cells of an occupancy grid, or allowing its calculation;
    a data processing module configured to receive said signals as input and use them to construct a consolidated occupation grid by applying a method as stated above; and at least one output port for a signal representative of said consolidated occupancy grid.
    According to particular embodiments:
    Such a system may also include a user interface device connected to said second input port and adapted to allow a user to choose said or at least one said a priori occupancy probability value.
    Such a system may also include a radio receiver connected to said second input port and adapted to receive a radio signal representative of said or at least one said a priori occupancy probability value.
    The system may also include a spatial or spatio-temporal positioning system connected to said second input port and adapted to generate positioning information, said data processing module being adapted to calculate said or at least one said probability value d occupancy a priori from said positioning information.
    The system may also include one or more sensors connected to said second input port and adapted to generate a signal for the presence of said material bodies.
    The system may also include one or more distance sensors adapted to generate signals representative of a plurality of distance measurements from said material bodies and connected to said input port (s).
    Said data processing module can comprise at least one hardware block for calculating occupation probabilities comprising a memory storing, in the form of a correspondence table, an inverse model of a sensor associating with each distance measurement a vector of integers representing indices of probability classes associated with respective cells of an occupation grid.
    Said data processing module can comprise an entire hardware block, called consolidation, configured to receive as input a plurality of integers representing indices of probability classes associated with cells of respective occupation grids, and to calculate an index of a probability class associated with a cell of said consolidated occupation grid.
    Other characteristics, details and advantages of the invention will emerge on reading the description made with reference to the attached drawings given by way of example and which represent, respectively:
    Figure 1, the concept of "direct" model of a distance sensor;
    Figure 2, the concept of occupancy grid;
    Figure 3, the concept of "reverse" model of a distance sensor;
    Figure 4, the spatial discretization of an inverse model on an occupation grid;
    ίο
    FIGS. 5A to 5D, a method for choosing the optimal spatial resolution of an occupancy grid;
    FIGS. 6A and 6B, various systems of probability classes;
    FIGS. 7A and 7B, a system for perceiving obstacles according to two variants of a first embodiment of the invention;
    FIGS. 8A and 8B, a system for perceiving obstacles according to a second embodiment of the invention; and
    FIG. 9, a system for perceiving obstacles according to a third embodiment of the invention.
    In the detailed description which follows, reference will be made to the case of the perception of obstacles. However, all that is described applies more generally to the perception of all kinds of material bodies.
    Most often, the sensors used for navigation provide information on the distance from surrounding obstacles; this is called distance sensors. To account for the accuracy of a sensor, its possible error or its resolution, we introduce a probabilistic model. The idea is that a measurement at the sensor output does not necessarily indicate the exact distance between the obstacle and the sensor, and that consequently it is necessary to reason about the probability that the obstacle is at a given distance knowing the sensor response.
    If we denote by D the real distance between an obstacle and the sensor, and z the sensor output, we are interested in the conditional probability density function p (z | D) which models the link between the real position of a obstacle and its estimation seen by the sensor ("direct model"). Figure 1 shows an example of a direct model of a sensor; we consider a linear space of 50 m long and assume that an obstacle is at D = 25m from the sensor. For a sensor with an error that can be modeled by a Gaussian function, the most likely z response will be close to 25 m, but other values will be possible, with a probability density defined by the curve. In the case of an ideal sensor we would have p (z | D) = ô (z-D), where δ is a Dirac Delta, and the measurement would always be equal to the true distance. The direct model of a sensor can be obtained experimentally, by performing series of measurements for one or more distances D. It can also be constructed empirically, typically from data supplied by the manufacturer (in the Gaussian case, the value of the standard deviation is sufficient to describe the model).
    In the following, we will denote by Ω a one, two or three dimensional spatial referential; an occupancy grid GO is a partition of a continuous and bounded subset of Ω into a number N of parts, called cells and designated by an index ie [0, N-1]. We indicate by c, the cell with index i. Without loss of generality, we will consider in the following a one-dimensional occupation grid observed by a single distance sensor C (or a plurality of co-located sensors), the index i increasing with the distance from the sensor (co being therefore the cell closest to the sensor and Cn-i the farthest). This configuration is illustrated in Figure 2.
    An obstacle A is a bounded continuous subset of Ω. We say that a cell c, is occupied by an obstacle A if Anc ^ 0, that it is not occupied by A if AnCi = 0. In other words, if the obstacle even partially covers the cell, it is considered occupied. Other conventions are possible, but in any case a cell must be either free or occupied.
    We consider for each cell of the grid, the binary random experiment "state" which can have one of the two outcomes {occupied; empty} consisting in knowing if the cell contains an obstacle or not. We will note e, the state of the cell c ,, o, the realization e, = occupied and v, the realization e, = empty. In a grid, we consider that all the cells are independent, so that
    V i, je [0, N-1], P (Oi A Oj) = P (0i) -P (0j) (1) where a is the logical operator "and >> and P (.) Denotes the probability of an event (not to be confused with a probability density, designated by a "p"
    tiny). In the following, for the sake of simplicity, we will replace "a >>
    with a comma, so Ρ (ο,, 0j> P (0i a Oj).
    It is also considered that the position of the obstacles can only be known using uncertain distance sensors, characterized by a probabilistic model as described above which can be written more generally p (z | x), x being the position of an obstacle (in several dimensions, it is a vector, expressed in Cartesian, spherical, polar coordinates, etc. and not a simple scalar). These sensors can be range finding lasers (also called lidars), sonars, infrared radars, time of flight cameras, etc.
    A measurement z from a sensor makes it possible to determine the probability of occupation Ρ (ο, | ζ) of a cell c ,. For a given measure z, the set of probabilities P (o, | z) V i e [0, N-1] constitutes the inverse model of the sensor on the grid. While the direct model of the sensor provides information on the response of the sensor as a function of the physical world, the inverse model expresses the impact of the measurement on the occupancy grid, which is the model of the physical world that we adopt, which justifies the reverse model designation.
    Figure 3 shows a typical example of an inverse model of a distance sensor, in a case where z = 25m. It can be checked that the probability of occupancy is almost zero for the cells which are at a distance less than 24.25 m from the sensor and reaches a peak for a distance of 25 m (corresponding to the measurement provided by the sensor) . Beyond 25 m, the probability of occupation decreases until it stabilizes at a value of 0.5, indicative of a total ignorance of the state of occupation of the cells which, being located beyond the obstacle, are hidden by the latter and therefore inaccessible to the sensor.
    In accordance with the practice prevailing in the literature, Figure 3 shows the inverse model by means of a smoothed curve. A more correct representation would be to display only the points corresponding to the limits of the cells of the grid: indeed, one cannot distinguish a cell "partially" occupied from another which would be "totally", in all cases the distance to the obstacle will be estimated as the distance to the corresponding cell. This is the spatial error introduced by the grid.
    A fairer version of the inverse model of Figure 3, taking into account this spatial discretization induced by the grid is presented in Figure 4.
    It should be noted that the concepts of "occupation" and "distance from the obstacle" are not entirely equivalent. Indeed, to say that an obstacle is at a distance z from the sensor does not only mean that a certain cell is occupied, but also that all the other cells of the grid closer to the sensor are free (otherwise, the first obstacle would have been seen at a distance less than z). In FIG. 2 above, the obstacle A is in the cell of index i (in black); cells with index less than i are shown in white to indicate that they are free, those with index greater than i are shown in gray to indicate that their occupancy is unknown.
    If we take into account the notion of an uncertain sensor characterized by its (direct) model p (z x) and note d, the distance from cell c, with respect to the measurement point and x [the point of cell c, the closest to said measurement point, we have:
    Vi <N, p (z xf) = p (z v 0 , ..., ν ^ Οΐ) (2)
    Equation (2) indicates that the sensor model evaluated in one
    0 point which is at the border of a cell of the grid (x,) is equal to the density of probability of response of the sensor for a corresponding grid configuration, namely a grid where the cells closest to the cell i are empty, cell i is occupied, and the states of occupation of cells further away than cell i are not determined. The inverse model of the sensor can be constructed by exploiting this information. An explanation of this process is given below.
    Bayes' theorem allows to express the inverse model of a sensor Ρ (ο, | ζ) in the following way:
    p (z) ρ (ζ θΐ) Ρ (θΐ) + ρ (ζ νΐ) Ρ (νΐ) (3) where P (Oi) and P (Vi) denote the a priori probabilities (i.e. say without knowing the position of obstacles, or the output of the sensor) that the cell is occupied or free, respectively. In the following, we will hypothesize a priori equal probabilities for all cells: Ρ (ο,) = Ρ (ο) and P (vî) = P (v) = 1-P (o) Vi. This assumption simplifies the calculation of the inverse model, but is by no means essential. The inverse model remains calculable if P (Oi) is not constant, provided that the a priori probability of occupation of each cell is known. The calculations are certainly more complex, but should only be done once, and not in real time; complexity is therefore not unacceptable.
    Equation (3) then becomes:
    ni rz1 η Λ D ί
    To be able to apply equation (4) and effectively determine P (Oj | z), the terms p (z | Oj) and p (z | v,) must first be calculated. For this we will consider the case of a one-dimensional grid, but the generalization is without difficulty. In the case where the a priori probability of occupation is not the same for all cells, equation (4) can be generalized without difficulty:
    rz1 r D ί r
    Calculation of p (z | Oj)
    Let “d” be the first occupied cell in the grid (0 <d <N), and
    P (dk) the probability of the event d = k. We can then write:
    ρΟΙθί) = ZLoP (z d k ) P (Moi) + p (z V) P (VM where V corresponds to the event "all cells are empty". By definition P (dk I ο,) = 0 if k> i and P (V | ο,) = 0, then:
    P (z ° ù = Zk = oP ( z K) p ( d kl 0 i) (6)
    The term P (d k | oj) takes the following values, according to the indices k and i:
    <P (d 0 | o 0 ) = 1 P (d 0 | Oi) - P (oq) for i> 0 <P (d k | Oi) = [riy = o ^ ©)] ^ (Ok) For 0 <k <i
    P (di Oi) = Πγ = ο p ( v y) <P (dk c> i) = 0 for k> i (7)
    At this point we introduce the aforementioned hypothesis of a priori equal probabilities for all cells:
    P (Oi) = P (o) and P (vi) = P (v) = 1-P (o) Vi. (8)
    By replacing (7) and (8) in (6) we find:
    (p (z | o 0 ) = p (z | d 0 ) [ρ (ζ | θ;) = P (o) p (z | d 0 ) + Zfc = o [l - PG)] k P &) p (z dk) + [1 - P (o) Ÿp (z di): i> 0 10 (9)
    Calculation of p (z | vQ
    Similar to (6) we can write:
    P (z vî) = lk ^ oP (z d k ') P (d k v i ') + (z | V) F (K | v £ ) (10) however in (10) the second term is generally not zero.
    The term P (d k vi) takes the following values, according to the indices k and i:
    <P (d 0 | r 0 ) = 0
    P (d 0 1Vf) = P (o 0 ) for i> 0 <P (d k v i ) = [Y [ 1 j / py j ')] PÇoGpourO <k <i (11)
    P (di Vi) = 0 <P (.d k Vi) = [n ^ + i ^ fe)] P (Ofc) for k> i
    By replacing (11) in (10) and introducing the aforementioned hypothesis of a priori equal probabilities for all cells:
    P (Oi) = P (o) and P ( Vi ) = P (v) = 1 -P (o) Vi. (8) we find:
    f p (z | v 0 ) = PÇp ^ pWdJ + Zk = 2 [l “P (o)] fe Ko) p (z | 4) + + [lP (o)] ] V - 1 p (z | 7 ) <
    p (z | e f ) = PÇo) pÇz d 0 ) + ZLoU “P (o ')] k P (o') pÇz d k ) + l + ZEViU - Pto ^ PÇoM ^ dk) + [1 - PÇoT ^ pÇzW) (12)
    To obtain the inverse model, it suffices to replace (8), (9) and (12) in (4).
    The construction of the inverse model strongly depends on the definition of the grid. It is therefore interesting to study what is the impact of a variation in spatial resolution on the inverse model. Figures 5A - 5D show the inverse models of the same sensor on four different spatial resolution grids: 1 m (6A), 50 cm (6B), 25 cm (6C), 12.5 cm (5D). These inverse models were constructed by taking P (o) = P (v) = 0.5. We can notice that when the resolution increases (the grid step decreases), the maximum of the inverse model decreases and tends towards 0.5. Indeed, one should not expect to be able to know the position of an obstacle with a precision greater than that of the sensor. Conversely, if we are content to know the occupation with an accuracy much lower than that of the sensor, we can determine with great certainty the presence or absence of an obstacle (case of Figure 5A, where the maximum reverse model is 0.999994). These considerations make it possible to optimize the spatial resolution of the grid: we can indeed carry out an exploration making it possible to determine the maximum resolution of the grid for which the maximum of the inverse model remains above a threshold (strictly greater than 0.5 and strictly less than 1) considered "significant".
    From the inverse models of two sensors on the same occupancy grid, the data from the two sensors is merged using the following equation:
    Z2> - [1 - »(O) (13) where zi and z 2 are the measurements provided by the two sensors (the generalization to more than two sensors is immediate - just consider P (o i z 1 , z 2 ) as the inverse model of a "virtual" sensor and merge it with the measurement provided by a third sensor, and so on).
    In the case where the a priori probability of occupation is not the same for all the cells, equation (13) can be generalized without difficulty:
    (= [1 - ^ (θί)] ^ (θίΙζι) ^ (θίΙζ 2 ) (Ο ^ ζ, .ζ,) - tl _ ρ (ο . )] Ρ (ο . | Ζι) ρ (ο . | Ζ2 ) + PC ^ PCrJzJPCrilzJ (13bis)
    Directly applying equation (13) or (13bis) to the merging of data from several sensors is difficult to envisage in an on-board system, as this would require the execution of numerous floating point calculations for each cell of the grid at a frequency at least as fast as the acquisition frequency of the sensors. Significant computing power would therefore be necessary.
    In accordance with the approach already disclosed by the aforementioned article by T. Rakotovao et al., The invention exploits the concept of "probability classes" to perform the merging of data from several sensors using only whole calculations.
    In what follows, we will call “probability class system” S = {p n , ne Z} a countable subset of [0; 1], whose elements p n can therefore be characterized by a relative integer index "n". If we call “F” the data fusion function expressed by equation (13) above, we can write:
    F (JPgP2) [1-Ρ (ο)] · ΡΓΡ2 (14)
    In the case where the a priori probability of occupation is not the same for all cells, equation (14) can be generalized without difficulty:
    _ [l-P (Ot)] - Pi-P2_
    F (0i) -Pi-P2 + F (0i) - (l-Pi) - (lp 2 ) (14bis)
    A particularly interesting case is that of a class system such that the result of the fusion of two probability classes of the system also belongs to the system; formally: V Pi> Pj e S, F (pi, p 2 ) eS. We then speak of a class system "without error", because the fusion does not introduce any error or approximation. It is therefore possible to identify the probability values by the indices of the corresponding classes, and the result of a fusion is also identified by an index. The problem of merging then amounts to determining an appropriate function F d which, with two integer indices, associates another integer index.
    Formally:
    V (fo Z) e Z 2 , 3 ie Z: F (p k , p t ) = p t and we denote F d (k, T) = i.
    The calculation of F d (k, l) requires only the knowledge of the indices k and I and of the arithmetic of whole index; no floating point calculation is necessary for the calculation of the fusion of the information p k and p /.
    0 In addition, if the class system is considered without error, the index obtained using F d (k, l) designates a probability value strictly identical to that obtained - using floating point numbers - in applying equation (14). The method thus allows the fusion of probability classes without error compared to a floating point calculation.
    A trivial example of an error-free system is S = {1/2, 1}. Any error-free class system with probabilities other than 1/2, 1 and 0 necessarily has an infinite number of elements. In practice, for obvious reasons of implementation, we will consider only systems of classes of probabilities of finite cardinality. However, given that the sensors are in a finite number and that their outputs (quantified and digitized) can only take a finite number of values, we demonstrate that it is possible to carry out "error-free" merges even from of finite cardinality probability class systems.
    It is important to note that the inverse model in Figure 4 was spatially discretized, but the probability of occupancy of each cell could take any value in the interval [0,1]. The use of class systems also implies a discretization of the probability values. Thus, it is necessary to approach the probability values of the inverse model of Figure 4 by elements of a class S system.
    A first possibility consists in replacing the values of the inverse model by the closest elements of S, so as to minimize the quantization error. This approach can lead to underestimating the probability of occupying a cell, which may not be acceptable in an obstacle detection application. An alternative consists in approaching the values of the theoretical inverse model by the smallest upper bound of the class system S; thus the probability of occupation is never underestimated, which can be an advantage for the detection of obstacles. In other applications, such as person counting, this type of approximation can, however, lead to the generation of false devices.
    Whatever the type of approximation considered, the spatially discretized inverse system takes a number of values which is very small (from less than 10 to a few tens at most). Consequently, the number of elements of the S system necessary to approach the inverse model is also very small. We can therefore limit ourselves to considering a finite, small-sized subset of the class system S whose cardinality is, in principle, infinite and countable.
    It has been shown above, with reference to FIGS. 5A - 5D, that - for a given sensor - the more the spatial occupation grid is resolved, the closer the maximum value of the inverse model is to 0.5. However, if we limit ourselves to probability values belonging to a system of classes S, there exists a probability class p m in which corresponds to the smallest value greater than 0.5. The optimal resolution of the occupation grid is therefore that for which the maximum value of the inverse model is equal to p m in- Any subsequent reduction in the grid pitch increases the computational load without bringing any gain in terms of information.
    A class system suitable for the implementation of the invention can be defined by induction.
    Let p be a probability of occupation strictly between 10 0 and 1-P (o). We then define by induction the sequence p n as follows:
    Po = P (o); pi = p;
    p n + i = F (pi, p) Vn> 1 _ p (q) 2 (ip)
    P -1 P (o) 2 + (l-2P (o)) p '
    Pn-1 = F (p n , Pl) Vî <-1 (15) where F is the fusion function defined by equation (14).
    In the case where the a priori probability of occupation is not the same for all the cells, the system (15) can be generalized without difficulty:
    Po = P (Oi);
    pi = p;
    p n + 1 = F (p n , p) Vi> 1 _ P (Qj) 2 (lp)
    P -1 P (Oj) 2 + (l-2P (Oj)) p '
    Pn-i = F (p n , p.-ι) Vi <-1 (15bis)
    F then being defined by equation (14bis).
    In addition, the parameter "p", which fixes the value of pi, can also be different from one cell to another.
    By construction, these classes are error free. Indeed, if we denote f p the function with a variable which, with a probability x, associates f p (x) = F (x, p), we immediately see that Vn e M *, p n = fp (p) , where the exponent "n >>means" composed n times with itself >>. Therefore :
    KPri'Pm) = F (fp (p) = fp + m (.P) = Pn + m (16)
    The merge formula for these classes is extremely simple:
    F d (n, m) = n + m V n, me Z (17)
    In addition, the parameter p allows fine control of the error introduced by the quantification; indeed if we set p = 1/2 + ε we have:
    Ε (ρ) = ρι-ρ 0 = ε (18)
    This system of classes Gp (0) is therefore very interesting because it makes it possible to carry out the whole fusion in the simplest possible way (a whole addition) and without error, and to master the error introduced by the quantification by the choice of the parameter p.
    Figure 6A illustrates the probability class system Gp (0) for three values of the a priori probability (considered constant for all cells): P (o) = 0.2 - corresponding, for example, in the case of an automotive application, in a light traffic situation; P (o) = 0.5 corresponding to a traffic situation of medium intensity, or
    0 unknown; P (o) = 0.8 - corresponding to a situation of heavy traffic. FIG. 6B illustrates the engraved class system P (o) = 0.5 for three values of the parameter p: 0.52; 0.55 and 0.7.
    The Gp (0) class system can also be defined directly (not recursively):
    e L i.
    Ρί = ^ ϊ-, ίΕΖ (19) where L <= 1 ( l ° B fe) - l ° S Gfe)) + l ° S Gfe) , 20)
    FIG. 7A illustrates the application of the invention to a land vehicle VT equipped with a distance sensor C, for example a laser rangefinder with mechanical scanning (LIDAR). This sensor is configured to perform a plurality of one-dimensional scans of the space in front of the vehicle, each scan defining an acquisition "tablecloth". Preferably, a plurality of layers N1, N2, etc. are produced at different heights. During each scan, the sensor produces a vector of z measurements, each of which is indicative of the presence of an obstacle - pedestrian, other vehicle, tree on the side of the road ... - in a respective direction and its distance (for example, when z takes its maximum value, this means that there is no obstacle detected within the range limit of the sensor). As a variant, a plurality of co-located sensors (that is to say having the same point of origin for measuring the distance) can make it possible to simultaneously produce a plurality of acquisition layers.
    The vehicle VT is also equipped with a radio receiver RR which receives an SPO signal enabling it to determine the a priori probability of occupation value, P (o). The SPO signal is transmitted by an SST monitoring station which can be a traffic control center or even autonomous equipment of the road infrastructure. This signal can directly convey the value of P (o), or else a parameter representative of a traffic condition allowing a processor on board the vehicle to calculate this value.
    Figure 7B illustrates an obstacle perception system suitable for this application. This system comprises said sensor C (or a set of collocated sensors) and a data processing module MTD1 receiving as input the measurement vectors corresponding to each acquisition layer of the sensor and providing at its output a signal (typically a vector of 'integers) representative of an occupation grid obtained by merging the data from these acquisition layers.
    In the embodiment of FIG. 7B, the data processing module MTD1 comprises a plurality of hardware blocks for calculating probabilities of occupation, COi ... COnc and a hardware block for calculating F called consolidation or fusion. Each block of calculation of occupation probabilities COk comprises a memory storing, in the form of a correspondence table, an inverse model of the layer of index k of the sensor C, discretized by means of a system of probability classes , for example Si. We speak here of “inverse model of the tablecloth” because these are the measurements of the various tablecloths that we merge. If a single sensor is used to acquire several layers, this single sensor is in fact equivalent to a plurality of sensors each acquiring a layer, and each having its own inverse model (even if all these inverse models can be identical).
    The data processing module also includes an RPO register making it possible to store a value P (o) of the probability of occupation a posteriori. Unlike the embodiment of FIG. 7A, in this case this value is chosen by a user by means of a user interface UI. The UI interface can for example be the control panel of an on-board computer, also allowing you to adjust the radio, configure the browser, etc. For ergonomic reasons, it is preferable that the user does not have to enter a numerical value; for example, they may have to choose between a finite number of traffic states, or place a cursor on a bar.
    Each processing block COk therefore receives as input the measurements corresponding to a respective acquisition layer Zk (references zi ... znc) as well as the value P (o) stored in the register RPO, and provides an output of occupation, in the form a vector of integers gk representing the indices of the probability classes associated with the different cells of the grid. Thus, the grid gk contains the occupation information estimated using the measurements of the tablecloth k only, that is to say the vector of measurements z k .
    The consolidation hardware block F is a very simple arithmetic circuit, which implements equation (17). It receives at its entry the occupation grids gi ... gNc and provides at its output a "consolidated" occupation grid, represented in turn by a vector of integers, indices of the probability classes associated with the different cells of this consolidated grid.
    If the inverse models associated with the different acquisition layers are identical, the blocks COi ... COnc are also identical, and can be replaced by a hardware block for calculating probabilities of single occupation, performing the treatments sequentially.
    The MTD1 data processing module can also be combined with any other type of distance sensor.
    FIGS. 8A and 8B relate to another embodiment of the invention, using several sensors arranged at different locations which cooperate to provide an occupancy grid constructed using measurements made from different points of view. The sensors can be technologically heterogeneous, in precision, range, field of vision and / or speed of acquisition. In this embodiment, the distance from the first obstacle is information relating to the sensor making the measurement. A schematic example of the scenario is shown in FIG. 8A, showing two sensors C1 and C2 placed at different positions and having different ranges and fields of vision. Thus the obstacle O is seen at completely different distances by C1 and C2.
    In this embodiment, the main difficulty lies in the fact that the occupancy grid on the one hand and the sensors on the other hand each have their own mark associated with them. Thus, the evaluation of the location of obstacles requires changes in benchmarks.
    FIG. 8B illustrates a system for perceiving obstacles according to such an embodiment of the invention. This system generally includes "NC" non-co-located and potentially heterogeneous sensors Ci, C 2 ... Cnc and a data processing module MTD2. The latter differs from the data processing module MTD1 of FIG. 7B in that it also includes, interposed between the hardware blocks for calculating the probabilities of occupation COi ... COnc and the hardware block for consolidation F, blocks of change of reference Ri ... Rnc- Each of these blocks Rk contains calculation units, generally in floating point, to effect the change of the reference of a respective sensor towards the reference of the occupation grid called "consolidation" With respect to which the data fusion is performed. The calculation carried out for the change of coordinate system consists in reassigning the occupation of a known location in the coordinate system of a sensor Ck (expressed by an integer vector g k ) to the corresponding cell in the coordinate system of the consolidation grid . We represent by gi ... g NC the integer vectors representative of the occupations of the cells of the consolidation grid. This reassignment involves the calculation of translations, rotations, etc. The processing of the Rk blocks can for example be carried out using a floating arithmetic unit of an on-board processor (FPU: Floating Point Unit). In this case the same hardware block can perform the calculation for all of the blocks Rk (sequential processing).
    As a variant, the equations for changing the reference point can be stored in conversion tables stored in memories contained in the modules Rk. So even in this case, we can get away from floating point calculation and only perform operations on integers. On the other hand, these conversion tables can be quite large and their storage can have a non-negligible cost in terms of silicon surface.
    In addition, unlike the two previous embodiments, the value of the a priori occupation probability P (o) is calculated by an EPO processor from data from a sensor Cpo distinct from the sensors Ci - Cnc providing the measurements. z. For example, it can be a non-directional radar, emitting an electromagnetic pulse and counting the number of echoes received. The higher this number, the higher the value of P (o).
    FIG. 9 illustrates a third embodiment of the invention in which a single sensor C, not shown, acquires scalar or vector measurements z t , z t + i, ... z t + m ... at a rate d N times higher acquisition of what is required for a specific application. A hardware block for calculating CO occupation probabilities produces an occupation grid g t , g t + i, ... gt + m ··· for each of these measurements. Then a fusion material block F merges N of these grids, acquired at successive instants, into a single consolidated grid gf US ; the gf US consolidated occupancy grids are therefore generated at a rate N times lower than the rate of application of the sensor. For example, if the sensor C operates at a rate of 100 Hz and a rate of 10 Hz is sufficient for the intended application, one can merge 10 successive acquisitions.
    In this embodiment, the EPO processor calculates the a priori occupancy probability value P (o) as a function of the space-time positioning of the sensor, i.e. its position and / or the time of the day. This data is provided by a POS positioning device, such as a GPS receiver. As explained above, knowing the position and / or the time makes it possible to estimate an expected traffic density, and therefore to choose an optimal value of P (o).
    Different means of acquiring and / or calculating the value of P (o) have been described with reference to different embodiments of the invention. However, each embodiment can use any of these means - or even more of them. We can even consider using different acquisition or calculation means for the a priori occupation probabilities of different cells of the grid. In addition, the means described do not constitute an exhaustive list.
    In the embodiments of FIGS. 7B, 8B and 9 we have considered the case where the processing operations - or at least some of them - are carried out by hardware calculation blocks, that is to say dedicated digital circuits . However, the invention also lends itself to fully or partially software implementation, in which the processing operations - or at least some of them - are carried out by a generic processor programmed in a timely manner.
    Other probability class systems than c / © can be used for the implementation of the invention. In particular, it is not essential that the class system depends on the a priori probability P (o) or Ρ (ο,).
    权利要求:
    Claims (21)
    [1" id="c-fr-0001]
    1. Method of perception of material bodies (O) comprising the following steps, implemented by a computer or a dedicated digital electronic circuit (MTD1, MTD2):
    a) Acquisition of a plurality of distance measurements (zt ... z N c) of said material bodies from one or more sensors (Ci ... Cnc);
    b) Acquisition from an external device, or calculation from at least one signal (SST) received from the outside, of at least one a priori probability of occupation value of the cells of an occupation grid;
    c) Application, at each said distance measurement, of an inverse model of the corresponding sensor on said occupancy grid (GO) providing a discretized spatial representation of an environment of said sensor, to determine a probability of occupation by a body material of a set of cells of said occupancy grid, each said inverse sensor model being a discrete model, associating with each cell of the corresponding occupancy grid, and for each distance measurement, a probability class chosen at within the same set of finite cardinality, each said probability class being identified by an integer index; and
    d) Construction of a consolidated occupation grid, each cell of which has an occupation probability calculated by merging the occupation probabilities estimated during step c), the occupation probability of each cell of the consolidated occupation being determined by means of whole calculations carried out on the indices of the probability classes determined during said step c), and as a function of said or of said a priori probability of occupation.
    [2" id="c-fr-0002]
    2. The method as claimed in claim 1, in which said step b) comprises the acquisition of said or at least one said a priori occupancy probability value by means of a user interface device (UI).
    [3" id="c-fr-0003]
    3. Method according to claim 1 wherein said step b) comprises the acquisition of said or at least one said a priori occupancy probability value by means of a radio receiver (RR).
    [4" id="c-fr-0004]
    4. The method of claim 3 wherein said or at least one said a priori probability of occupation value, acquired by means of a radio receiver, is transmitted by a control station (SST).
    [5" id="c-fr-0005]
    5. Method according to claim 1 wherein at least said sensors are on board a land vehicle, and said or at least one said a priori probability of occupation value, transmitted by a control station, is a function of a density of traffic on at least one road axis on which said land vehicle travels.
    [6" id="c-fr-0006]
    6. Method according to claim 1 wherein said step b) comprises the calculation of said or at least one said value of probability of occupation a priori from a spatial or spatiotemporal positioning of said sensors.
    [7" id="c-fr-0007]
    7. The method as claimed in claim 1, in which said step b) comprises the calculation of said or at least one said value of probability of occupation a priori from at least one signal of the presence of said material bodies from a or several sensors (Cpo) other than the sensors used to implement said step a).
    [8" id="c-fr-0008]
    8. Method according to one of the preceding claims, in which said set of finite cardinality of probability classes is formed by the union of one or more subsets such that, during said step d), the merging of two classes of probability belonging to the same subset provides a result also belonging to said subset.
    [9" id="c-fr-0009]
    9. Method according to one of the preceding claims, in which said set of finite cardinality of probability classes constitutes a non-uniform discretization of the probability interval [0, 1], with a discretization step increasing between 0 and 0.5 , then decreasing between 0.5 and 1.
    [10" id="c-fr-0010]
    10. The method as claimed in claim 9, in which said set of finite cardinality of probability classes, designated by Gp (0) , is defined by: Gp (0) = {(p n ), ne Z}, Z being the set relative integers, the probability classes p n being defined as follows:
    Po = P (Oi); pi = p;
    p n + i = F (p n , p) Vn> 1 _ P (o) 2 (l ~ P)
    P -1 P (o) 2 + (l-2P (o)) p '
    Pn-1 = F (p n , Pl) VÏC-1 where P (Oi) is the a priori probability of occupation value of the index cell "i" of the occupation grid, p a parameter of value strictly between 0 and 1 -Ρ (ο,) and F a function of fusion of probability of occupation; and in which, during said step d), the fusion between two classes of probabilities p ,, pj e G p (0) is calculated by applying the following equation: F (Pri'Pm) ~ Pn + m
    [11" id="c-fr-0011]
    11. Method according to one of the preceding claims, in which said step d) comprises the implementation of landmark changes to construct said consolidated occupation grid based on occupation probabilities of occupation grid cells associated with non-co-located sensors.
    [12" id="c-fr-0012]
    12. Method according to one of the preceding claims also comprising a preliminary stage of construction of the inverse models of at least one said sensor on the corresponding occupancy grid, implemented by application of the following equation:
    ia Ρ ( ζ Ι ° ί) Ρ ( ο ί) l ° i | Zj p (z | Oi ) P ( Oi ) + p (z | Vi ) [l - P ( Oi )] where:
    • P (oj | z) represents the probability of occupancy of the cell with index "i" of the occupancy grid, said cells being ordered by increasing distance from said sensor;
    • P (Oj) is said a priori probability of occupation value of the index cell "i" of the occupation grid;
    • ρ (ζ | θί) is the probability density of the measure “z >> provided by the sensor when the cell of index“ i ”of the occupation grid is occupied, function of Ρ (ο,) and the direct sensor model; and • p (z | Vj) is the probability density of the measure "z >> supplied by the sensor when the cell of index" i >> of the occupation grid is empty, function of Ρ (ο,) and the direct sensor model.
    [13" id="c-fr-0013]
    13. Method according to one of the preceding claims, in which the a priori probability of occupation is the same for all the cells of the occupation grid.
    [14" id="c-fr-0014]
    14. System for perceiving material bodies comprising: at least a first input port for receiving a plurality of signals (z-i, ..., znc) representative of distance measurements of said material bodies originating from one or more sensors;
    at least a second input port for receiving a signal representative of at least a value of probability of occupation a priori of the cells of an occupancy grid, or allowing its calculation;
    a data processing module (MTD1, MTD2) configured to receive said signals as input and use them to construct a consolidated occupation grid by applying a method according to one of the preceding claims; and at least one output port for a signal (g fus ) representative of said consolidated occupation grid.
    [15" id="c-fr-0015]
    15. The system of claim 14 also comprising a user interface device (UI) connected to said second input port and adapted to allow a user to choose said or at least one said probability of occupation probability a priori.
    [16" id="c-fr-0016]
    16. The system of claim 14 also comprising a radio receiver (RR) connected to said second input port and adapted to receive a radio signal (SPO) representative of said or at least one said value of probability of occupation a priori .
    [17" id="c-fr-0017]
    17. The system as claimed in claim 14 also comprising a spatial or spatio-temporal positioning system (POS) connected to said second input port and adapted to generate positioning information, said data processing module being adapted to calculate said or at minus a said a priori occupancy probability value from said positioning information.
    [18" id="c-fr-0018]
    18. The system of claim 14 also comprising one or more sensors (Cpo) connected to said second input port and adapted to generate a signal of presence of said material bodies.
    [19" id="c-fr-0019]
    19. System according to one of claims 14 to 18 also comprising one or more distance sensors (Ci, ..., Cnc) adapted to generate signals representative of a plurality of distance measurements of said material bodies and connected to said or to said ports of entry.
    [20" id="c-fr-0020]
    20. System according to one of claims 14 to 19 wherein said data processing module comprises at least one hardware block for calculating probabilities of occupation (COi ... COnc) comprising a memory storing, in the form of a correspondence table, a
    5 inverse model of a sensor associating with each distance measurement a vector of integers representing indices of probability classes associated with respective cells of an occupancy grid.
    [21" id="c-fr-0021]
    21. The system as claimed in one of claims 14 to 20, in which said data processing module comprises an entire calculation hardware block, known as a consolidation block (F), configured to receive as input a plurality of integers representing indices of probability classes associated with cells of respective occupancy grids, and for calculating an index of a probability class associated with a cell of said grid
    15 of consolidated occupation.
    1/7 p (z / d)
    类似技术:
    公开号 | 公开日 | 专利标题
    JP2019050035A|2019-03-28|Methods and systems for object detection using laser point clouds
    FR3041451A1|2017-03-24|METHOD AND SYSTEM FOR COLLECTING MATERIAL BODIES
    EP3364213B1|2021-08-04|Method and system for contextualised perception of material bodies
    EP3137355B1|2021-09-01|Device for designating objects to a navigation module of a vehicle equipped with said device
    EP3252615B1|2021-09-08|Method and system for determining cells crossed by a measuring or viewing axis
    EP3435029A1|2019-01-30|Resolution adaptive mesh for performing 3-d metrology of an object
    US9767361B2|2017-09-19|Bathymetric techniques using satellite imagery
    FR2953313A1|2011-06-03|OPTRONIC SYSTEM AND METHOD FOR PREPARING THREE-DIMENSIONAL IMAGES FOR IDENTIFICATION
    Bergman et al.2020|Deep adaptive lidar: End-to-end optimization of sampling and depth completion at low sampling rates
    EP3126864B1|2020-07-15|Method for geolocating the environment of a carrier
    FR2977023A1|2012-12-28|GENERATION OF CARD DATA
    CA2400676A1|2001-08-23|Telemetric equipment for two-dimensional or three-dimensional cartography of a volume
    EP3189389B1|2021-09-08|Location and mapping device with associated method
    Galloway et al.2014|Automated crater detection and counting using the Hough transform
    EP3384462B1|2020-05-06|Method for characterising a scene by calculating the 3d orientation
    Yu et al.2013|Registration and fusion for ToF camera and 2D camera reading
    FR3096637A1|2020-12-04|Method and device for determining a path for an autonomous vehicle
    Hewitt2018|Intense navigation: Using active sensor intensity observations to improve localization and mapping
    WO2021005096A1|2021-01-14|Method for three-dimensional representation of the spatial coverage of one or more detection systems
    FR3081598A1|2019-11-29|MAPPING AND SIMULTANEOUS LOCATION OF AN OBJECT IN AN INTERNAL ENVIRONMENT
    WO2021214016A1|2021-10-28|Method for georeferencing of a digital elevation model
    FR3090126A1|2020-06-19|Method for locating, in an absolute coordinate system, a device, computer and associated vehicle.
    Zini et al.2010|Relative pose estimation for planetary entry descent landing
    WO2014111442A1|2014-07-24|Method and system for absolute 3d modelling of all or part of a vehicle passing in front of a camera
    FR2964198A1|2012-03-02|Method for e.g. monitoring terrestrial targets moving at ground surface by using position measurements from radar sensor onboard on e.g. aircraft, involves defining tracking plane associated to target, so as to be tangent to surface
    同族专利:
    公开号 | 公开日
    EP3364213B1|2021-08-04|
    FR3062924B1|2019-04-05|
    US11105911B2|2021-08-31|
    US20180231650A1|2018-08-16|
    EP3364213A1|2018-08-22|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    WO2013087067A1|2011-12-14|2013-06-20|Continental Teves Ag & Co. Ohg|Free space information in an occupancy grid as basis for determining a maneuvering space for a vehicle|
    EP2891899A1|2014-01-06|2015-07-08|Honeywell International Inc.|Mathematically combining remote sensing data with different resolution to create 3D maps|
    FR2890774B1|2005-09-09|2007-11-16|Inst Nat Rech Inf Automat|VEHICLE DRIVING ASSISANCE METHOD AND IMPROVED ASSOCIATED DEVICE|
    DE102009007395B4|2008-03-25|2015-11-26|Volkswagen Ag|Method for map-based environment representation of a vehicle|
    US8106792B2|2009-07-10|2012-01-31|Telcordia Technologies, Inc.|Program and method for adaptive mobile ad-hoc wireless communication|
    US9429650B2|2012-08-01|2016-08-30|Gm Global Technology Operations|Fusion of obstacle detection using radar and camera|
    FR3041451B1|2015-09-22|2018-02-16|Commissariat A L'energie Atomique Et Aux Energies Alternatives|METHOD AND SYSTEM FOR COLLECTING MATERIAL BODIES|
    US20180203100A1|2017-01-19|2018-07-19|Honeywell International Inc.|Quality metric for ranging sensor in a degraded visual environment for a situational awareness system|EP3734388A1|2019-04-29|2020-11-04|Commissariat à l'Energie Atomique et aux Energies Alternatives|Method and apparatus for performing simultaneous localization and mapping|
    FR3097972B1|2019-06-28|2021-12-10|Aptiv Tech Ltd|Method and system for mapping a physical environment using an occupancy grid|
    US20210365961A1|2020-05-25|2021-11-25|Shopify Inc.|Systems and methods for measuring traffic density in a region|
    法律状态:
    2018-02-26| PLFP| Fee payment|Year of fee payment: 2 |
    2018-08-17| PLSC| Publication of the preliminary search report|Effective date: 20180817 |
    2020-02-28| PLFP| Fee payment|Year of fee payment: 4 |
    2021-02-26| PLFP| Fee payment|Year of fee payment: 5 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1751246|2017-02-16|
    FR1751246A|FR3062924B1|2017-02-16|2017-02-16|METHOD AND SYSTEM FOR CONTEXTUALIZED PERCEPTION OF MATERIAL BODIES|FR1751246A| FR3062924B1|2017-02-16|2017-02-16|METHOD AND SYSTEM FOR CONTEXTUALIZED PERCEPTION OF MATERIAL BODIES|
    EP18153468.6A| EP3364213B1|2017-02-16|2018-01-25|Method and system for contextualised perception of material bodies|
    US15/892,316| US11105911B2|2017-02-16|2018-02-08|Method and system for contextualized perception of physical bodies|
    [返回顶部]