• 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:
    The invention relates to an autonomous vehicle comprising a chassis and carrying lugs, each of said lugs being composed of at least two segments, an upper segment, connected to said chassis by a first pivot, a lower segment carrying at its lower end a wheel, and connected to said upper segment by a second pivot; said lower segment being further connected by two opposing springs to a pulley secured to the frame, so as to exert an angular stiffness around a position of equilibrium corresponding to a vertical orientation of the lower segment; and, said upper segment being connected by two other counter-springs, at said first pivot, to a motorized pulley secured to said frame, adapted to exert a control of the moment exerted on said upper segment around a nominal position corresponding to an orientation. of said upper segment, said moment control being determined from a measurement of a first angle formed by the upper segment relative to its nominal position, of a measurement of a second angle formed by the lower segment relative to in its position of equilibrium, and the attitude of said frame.
    公开号:FR3055256A1
    申请号:FR1658067
    申请日:2016-08-31
    公开日:2018-03-02
    发明作者:Faiz Ben Amar；Christophe Grand；Arthur Bouton
    申请人:Centre National de la Recherche Scientifique CNRS；Universite Pierre et Marie Curie Paris 6；
    IPC主号:
    专利说明:

    Holder (s): PIERRE ET MARIE CURIE UNIVERSITY Public establishment, NATIONAL CENTER FOR SCIENTIFIC RESEARCH - CNRS - Public establishment.
    Extension request (s)
    Agent (s): NOVAGRAAF TECHNOLOGIES.
    FR 3 055 256 - A1
    INDEPENDENT SELF - CONTAINED VEHICLE AND ASSOCIATED CONTROL METHOD.
    ©) The invention relates to an autonomous vehicle comprising a chassis and carrying legs, each of said legs being composed of at least two segments, an upper segment, connected to said chassis by a first pivot, a lower segment carrying at its lower end a wheel, and connected to said upper segment by a second pivot;
    said lower segment being further connected by two opposing springs to a pulley secured to the chassis, so as to exert an angular stiffness around an equilibrium position corresponding to a vertical orientation of the lower segment; and, said upper segment being connected by two other opposing springs, at the level of said first pivot, to a motorized pulley, integral with said chassis, adapted to exercise control of the moment exerted on said upper segment around a nominal position corresponding to an orientation horizontal of said upper segment, said moment control being determined from a measurement of a first angle formed by the upper segment with respect to its nominal position, from a measurement of a second angle formed by the lower segment with respect to at its equilibrium position, and the attitude of said chassis.
    i
    AUTONOMOUS COMPLIANT VEHICLE AND METHOD FOR
    ASSOCIATED ORDER
    FIELD OF THE INVENTION
    The invention relates to the field of autonomous vehicles, or robot, and more particularly that of vehicles adapted to traffic on various terrains, not previously prepared for traffic and likely to include obstacles.
    BACKGROUND OF THE INVENTION
    The field of the invention is that of mobile robots, or autonomous vehicles, which can move on a ground neither prepared for traffic nor known in advance, independently.
    Unprepared soils are uneven soils, generally natural soils, which can therefore show roughness, irregularities and obstacles. A vehicle can be destabilized or come into abutment on these accidents on the ground, and it must be able to pass them autonomously, that is to say, without the help of a human operator. To do this, it must have an adapted physical structure, but also an appropriate command-control system to estimate and model the difficulties the vehicle faces, by means of sensors, and consequently control the actuators (or "actuators") to allow crossing .
    One field of application is, for example, that of astromobile robots, such as the Curiosity vehicle for exploring the planet Mars. But these autonomous vehicles can find applications on Earth also off the roads and artificially arranged grounds for the circulation of the vehicles, such as for applications of observation and civil or military intervention.
    Different techniques are usually used.
    Certain autonomous vehicles come in the form of massive vehicles, often with 4 or 6 wheels, whose weight and the presence of additional degrees of freedom in the chassis ensure the contact of all the wheels with the ground and the compensation of certain irregularities of the ground. In order to ensure the isostatism of the system, the connections are devoid of actuators but rather of visco-elastic organs.
    Other autonomous vehicles seek to determine their surroundings before advancing in order to position the carrying legs at appropriate locations on the ground to ensure their stability and progression.
    However, these solutions are not fully satisfactory for various reasons.
    First of all, certain solutions must implement sensors to determine their environment, that is to say typically video cameras, depth cameras or even the use of laser scanning. This type of mechanism increases the cost of the vehicle, since the camera or cameras must be of sufficiently good quality to allow modeling of the environment. In addition, the best models thus obtained are generally not sufficiently precise, require substantial computing power which increases the resources to be carried in the vehicle and therefore its cost, an imprecision of the terrain map obtained, but also an uncertainty on the position of the vehicle at all times in this map. These solutions may also not allow sufficiently short reaction times for good autonomous driving of the vehicle.
    Other solutions are based on a priori knowledge of the field or on assumptions concerning it.
    In general, the solutions of the technique are not completely effective in that they do not avoid blocking or loss of balance positions.
    This could be the result of encountering an obstacle for which the robot was not prepared, or a gap between the real terrain and its prior modeling, for example.
    SUMMARY OF THE INVENTION
    The object of the present invention is to provide an autonomous robotic vehicle at least partially overcoming the aforementioned drawbacks.
    More particularly, the invention aims to provide a vehicle having great capacities for crossing uneven ground and more particularly straightforward obstacles such as a step. It allows an autonomous control of the attitude of the vehicle without a priori knowledge of the geometry of the ground, as well as the application of adaptation strategies to the very contact with obstacles.
    To this end, the present invention provides an autonomous vehicle comprising a chassis and carrying legs, each of said legs being composed of at least two segments, an upper segment, connected to said chassis by a first pivot, a lower segment bearing at its end. lower a wheel (R), and connected to said upper segment by a second pivot;
    said lower segment being further connected by two opposing springs to a pulley integral with said chassis, so as to exert an angular stiffness around an equilibrium position corresponding to a vertical orientation of the lower segment; and, said upper segment being connected by two other opposing springs, at the level of said first pivot, to a motorized pulley, integral with said chassis, adapted to exercise control of the moment exerted on said upper segment around a nominal position corresponding to an orientation horizontal of said upper segment, said moment control, τ, being determined from a measurement of a first angle, Ob ′ formed by the upper segment relative to its nominal position, from a measurement of a second angle, 0 C , formed by the lower segment with respect to its equilibrium position, and the attitude of said chassis.
    According to preferred embodiments, the invention comprises one or more of the following characteristics which can be used separately or in partial combination with one another or in total combination with one another:
    - each leg has a first encoder measuring the first angle 0b located on said first pivot, and a second encoder measuring said second angle and located on said second pivot;
    - Said attitude is determined by two inclinometers determining a pitch angle, 0 y , and a roll angle, 0 X , of the chassis, and by an elevation, p, of said chassis determined as an average of the elongations of said legs of said vehicle;
    - an action can be triggered to modify the actuation strategy of said vehicle according to the measurements of the deformations generated by the ground on said vehicle;
    - the forced fi measured for the leg i, with i = 1,2,3,4 is expressed by:
    Z = where T c , i is the moment exerted by the lower segment of said tab on the chassis at said second pivot, said moment is given by the expression:
    U ,, = -2 - ^ / -kh-0 c in which R c is the radius of said second pivot
    The invention also relates to a method for controlling an autonomous vehicle according to the invention, in which a control device determines a moment to be exercised by means of said motorized pulley () from the measurement of a first angle 0b formed by the upper segment with respect to its nominal position, by a second angle 0 C formed by the lower segment with respect to its equilibrium position, and with the attitude of said chassis by the expression:
    τ = G (0 b , 0 C ) + M (q) [K p q + K d q] -C (0 b , 0 c ) ~ N (q, q) in which
    G (0b, 0c) is a matrix representing the influence of the torque of the actuator on the dynamics of the chassis, and G (0b, 0 C ) + its pseudo-inverse;
    C (0 b , 0c) expresses the influence on the chassis of the stiffness of the springs at the level of the elbows;
    M (q) represents the inertia matrix;
    N (q, q) represents the gravitational, centrifugal and Coriolis forces;
    represents the difference between the state vector q and a reference attitude, determined by a reference elevation and reference pitch and roll angles.
    </ is the first derivative of the state vector q with respect to time;
    K p and Ka are two diagonal matrices formed by the correction gains.
    According to preferred embodiments, the invention comprises one or more of the following characteristics which can be used separately or in partial combination with one another or in total combination with one another:
    said control device can trigger an action to modify the actuation strategy of said vehicle based on measurements of the deformations generated by the ground on said vehicle.
    - the forced fi measured for leg i, with i = l, 2, 3, 4 is expressed by: fi = y ^ cos (^ ci ) 'c where T c , i is the moment exerted by the lower segment of said tab on the chassis at said second pivot (P2), said moment is given by the expression:
    C ,, = -2-R -Kh-0 c in which R c is the radius of said second pivot
    The subject of the invention is also a computer program having instructions allowing the implementation of the method described above when it is executed by an information processing platform.
    Other characteristics and advantages of the invention will appear on reading the following description of a preferred embodiment of the invention, given by way of example and with reference to the accompanying drawings.
    BRIEF DESCRIPTION OF THE DRAWINGS
    FIG. 1 schematically represents an example of architecture of an autonomous vehicle according to an embodiment of the invention.
    FIG. 2 represents a diagram of a leg of an autonomous vehicle according to the invention.
    DETAILED DESCRIPTION OF THE INVENTION
    In Figure 1 is shown very schematically an autonomous vehicle according to an embodiment of the invention. It comprises a chassis 100 which may include a control device 110, as well as 4 legs 101, 103, 105 and 107 connected at one end to the chassis 100 and terminating, at their other end, by wheels, respectively 102, 104, 106 and 108. These 4 legs make it possible to carry the chassis and substantially form a quadrilateral, typically a rectangle, the center of which preferably corresponds to the center of gravity of the autonomous vehicle.
    Each leg is made up of two segments:
    an upper segment, connected to the chassis 100 by a first junction, and substantially horizontal at its nominal position, and
    - A lower segment carrying the wheel at its lower end, connected to the upper segment by a second junction and substantially vertical at its nominal position.
    These two junctions make it possible to provide two degrees of agreement, a horizontal and a vertical. "Compliance" is the ability of a robot, or of a robot's organ, to react according to the external forces that are applied to that organ. There is a distinction between active compliance and passive compliance. Passive compliance only uses mechanical devices, generally composed of springs, axes, joints, etc. Active compliance involves a feedback loop between force or torsion sensors and an actuator controlled by a control circuit taking as input the signals transmitted by the sensors. In other words, we speak of active compliance when there is an additional energy supply to that provided by the environment at the interface.
    Furthermore, the autonomous vehicle can comprise a central pivot 109, which can be controlled by the control device 110. It can in particular allow a degree of freedom between the front and rear sections of the chassis 100 along an axis of rotation perpendicular to the horizontal plane formed by this chassis.
    This architecture makes it possible to best model the natural division of the legs into two components, horizontal and vertical, in space.
    Passive compliance along the horizontal axis plays a role in maintaining the grip of the wheels on the ground when encountering vertical obstacles and in their detection by the control system. Vertical compliance can, for its part, be implemented by a series-elastic actuator (or SEA for “Serial Elastic Actuator” in English). A SEA takes as input a setpoint force which is compared to a force applied by elastic deformation and the difference (or "error") is used to calculate the variation in position of a linear actuator.
    The spring of the SEA passively absorbs irregularities in the terrain at high frequencies, and the actuation mechanism allows you to control the vertical forces of the vehicle regardless of the differences in height of the terrain. The effort control therefore ensures at the same time the maintenance of the contact of each wheel with the ground, since the slaving of the elongation of the springs intrinsically realizes the relief.
    Thus, the vertical forces are actively controlled while the horizontal position of the wheels is maintained at a reference point, by the return force of the springs and the assistance of the speed control of the wheels.
    The speed of the wheels can indeed be modulated in proportion to their deviation in position.
    The invention also proposes a possible and effective implementation of this general principle, by using segments linked by pivots and actuated by antagonistic springs.
    Figure 2 provides a diagram of a leg of a vehicle according to this implementation of the invention. The other three legs are similar to this described leg and the person skilled in the art is therefore perfectly capable of producing the autonomous vehicle from the description of a single leg.
    Each leg therefore has two segments.
    - The upper segment Ls is connected to the chassis by a first pivot Pt, allowing the segment to rotate around the axis of the pivot in the plane of the wheels, but integral in translation with the chassis. This arrangement forms 1 "shoulder" of the leg. This degree of freedom aims to adapt the height of the wheel relative to the chassis.
    - The lower segment Li carries at its lower end the wheel R, and is connected to the upper segment by a second pivot P2. This second pivot allows the lower segment to be secured to the upper segment and to rotate around this pivot, in the plane formed by the two segments. This arrangement forms the "elbow" of the leg. This degree of freedom makes it possible to absorb the horizontal forces exerted on the tab.
    In addition, the lower segment Li is connected to a pulley SI secured to the chassis by two opposing springs khi, kh2, of stiffness kh, so as to exert an angular stiffness around an equilibrium position corresponding to a vertical orientation of the segment lower, regardless of the position of the upper segment to which it is connected. In this way, any deviation from this equilibrium position is possible but causes a return movement. These antagonistic springs act as a return around the equilibrium position via the pivot P2, by exerting antagonistic pulls of stiffness kh at a distance R c from the center of the pivot.
    This configuration makes it possible to use only linear springs in tension, which is simpler to implement. In addition, in the case of the P2 pivot, this allows the orientation reference to be moved, which can then be given by the chassis, although the P2 pivot is not directly attached to it.
    ίο
    The upper segment Ls is connected to a motorized pulley S2 secured to the chassis by two other opposing springs kvl, kv2. The two springs allow, as for the first segment, to create an angular stiffness around an equilibrium position. This position is controlled by the rotation of the motorized pulley S2 and can thus move around the nominal position represented by a horizontal orientation of the upper segment.
    The motorized pulley S2 is further adapted to exercise control of the moment τ exerted on the upper segment Ls around the pivot PI via the springs kvl and kv2.
    The angular actuation is determined by a control device from measurements provided by sensors arranged on the junctions between the two segments and between the upper segment Ls and the chassis.
    These sensors include
    - an encoder on the axis of the first PI pivot, measuring the angle 0b. This angle is that formed by the upper segment Ls relative to the longitudinal axis of the chassis, that is to say relative to its nominal position;
    - an encoder on the axis of the second pivot P2 measuring the angle 0 C. This angle is that formed by the lower segment Li with respect to the vertical, that is to say with respect to its position of equilibrium.
    - an encoder on the axis of the motorized pulley S2 measuring the angle 0 m , which is the angle of the pulley with respect to the orientation which makes the equilibrium position of the upper segment coincide with its nominal horizontal position.
    According to one embodiment of the invention the encoder at the second pivot can not directly determine the angle 0b + 0 c. The angle 0 C is deduced by difference with the angle 0b measured by the first encoder.
    These sensors are present on each leg of the vehicle, so 4x3 = 12 sensors are present on all 4 legs of the vehicle.
    In addition, two inclinometers, not shown in the figure, are arranged on the chassis, preferably near its center of gravity. An inclinometer is used to measure a pitch angle 0 y of the chassis. Another inclinometer is used to measure a roll angle θ χ of the chassis. A practical implementation consists in using an inertial unit which has an integrated Kalman filter and provides these filtered angles directly.
    Kalman filters are part of filtering techniques well known to those skilled in the art in the field of automation and digital signal processing. They are notably exposed in a general way on the web page:
    https://fr.wikipedia.org/wiki/Filtre_de_Kalman
    The original article defining Kalman filters is. A New Approach to Linear Filtering and Prediction Problems by Kalman, R. E, in Transactions of the ASME - Journal of Basic Engineering Vol. 82, p. 35-45 (1960)
    These sensors make it possible to deduce the instantaneous configuration of the robot, and to have information to control its stability in all circumstances via the control of the torques applied to the upper segments of the legs.
    In particular, these sensors make it possible to detect the presence of an obstacle and immediately generate a reaction at the level of each leg by determining a control of the moment to be exerted by the actuator (that is to say the motorized pulley S2). These torques are controlled so as to maintain the values of the angles θ χ , 0 y , as well as the elevation p of the autonomous vehicle at reference values.
    This elevation p of the vehicle can be deduced from the measurements made by the mentioned sensors. Indeed, the vertical elongation pi of each leg i can be determined by:
    Pi = Lb * sin (0b, i) + L c * cos (0c, i) with Lb and L c the lengths of the lower and upper segments of the leg, and 0b, i and 0 c , i the values of the angles 0b , 0 C for leg i.
    The elevation p of the chassis can then be deduced from these elongations of each leg, by determining the average value.
    We note q the state vector that the control device must seek to control, and τι, T2, τ 3 , u the components along the ordinate direction of the torque τ determined by this control device and to be applied to the actuator previously mentioned.
    So :
    q = [, 0x, 0 y ] T , and τ = [τι, τ2, τ 3 , τ4] Τ τ is therefore the vector grouping the four moments τι, τ2, τ 3 , τ 3 to be applied to each of the segments upper through actuators
    In general, we can write the following equation, which determines the dynamics of the autonomous vehicle:
    M (q) q = G (0 b , 0 c ) T + C (0 b , 0 c ) + G (q) + N (q, q) + q in which:
    G (0b, 0c) is a matrix representing the influence of the actuator torque on the chassis dynamics,
    C (0 b , 0c) expresses the influence on the chassis of the stiffness of the springs at the level of the elbows;
    M (q) represents the inertia matrix;
    N (q, q) represents the vector of the gravitational, centrifugal and Coriolis forces;
    q represents the second derivative of the state vector q, and // represents disturbance forces.
    The matrix G (0 è , 0 c ) can be written:
    Λ Λ
    There
    - - L x Z l - - L x Z 2
    G <& M =
    And the vector C (0 b , 0 c ) can be written:
    -Λ
    -!, Ζ ~ 1 _ / -yA
    -Λ - / y Ai
    -1 - ZA with
    C (^ A) = 27 X
    Al + A2 + A3 + A4 / -, (// - / + + / + “A4)
    Σ ^ + 4 (-Αι-Α 2 + Α3 + Α 4 ) / = 1 cos θ „
    L b cos (0 bi -θ α Υ and
    A, = sin / 9,
    L c cosfô, -θ α )
    According to one embodiment of the invention, one can make approximations on the above expression as well as on the determination of the matrices, as soon as the angles are close to the initial angles.
    In this situation, the matrix G (0b, 0 c ) can be written:
    G (0 b , 0 c ) =
    -1 -1 1 1 L b cos0 bl Z, cos ^ 2 L b cos 0 h 3 Z, cos ^ 4-L L L -L y y y y L b cos0 bl L b cos0 bl L b cos0 b3 L b cos0 bAL L L L 1 + 1 + 1 + 1 + A.cos ^ L b cos 0 b 3 L b cos 0 b 3 Z, cos ^ 4
    And the vector C (0b 0 c ) can be written:
    c (<u) =
    -2R 2 ck b ^
    In other words, the matrix G (0b, 0 c ) then only depends on the angle Ob, and the vector C (0b 0 c ) only depends on the angle 0 c .
    The control device can then determine all of the moments to be exercised by the expression:
    τ = G (0 b , 0 C ) + · M (q) [K p q + K d q -C (0 b , 0 c ) -N (q, q) in which q represents the difference between the vector state q and a reference attitude, determined by a reference elevation and reference pitch and roll angles. This attitude can be chosen with harmful angles in order to distribute the weight of the chassis evenly on the 4 legs of the vehicle;
    q is the first derivative of the state vector q with respect to time; K p and Ka are two diagonal matrices formed by the correction gains.
    and G (0b, 0 c ) + the pseudo-inverse of G (0b, 0 c );
    The control device can thus receive the measurements from the sensors in the form of a data flow and continuously determine the control of the torque τ to be exerted by the actuator by means of the motorized pulley S2.
    In addition, in addition to this compliance mechanism for each leg, allowing each leg to react in real time to the terrain on an individual basis, a behavioral adaptation mechanism can be implemented to adapt the strategy to activate the autonomous vehicle synchronously. In particular, this involves playing on the distribution of the vehicle's lift forces, when an equal distribution (or "equal distribution" of effort) on each wheel is not enough to overcome a particularly steep obstacle that presents itself.
    According to this mechanism, an action can be triggered to modify the distribution of the actuation on the different legs as a function of measurements of the deformations generated by the terrain on the structure of the vehicle.
    All of the actions should make it possible to offer a solution to each type of difficulty that may be encountered, so that whatever the conditions in which the robot is located, it can continue to progress.
    Basic actions can include:
    - Modulation of lift efforts independently of postural control.
    As the vehicle rests on 4 non-aligned wheels and is therefore hyperstatic, we have a degree of freedom to modulate the distribution of the torques on the legs without this having any effect on the control of the posture.
    - Swinging of the chassis and passage on three supports.
    From the point of view of static balance, when the wheels are engaged with an obstacle and exert a torque, only those located at the front can be completely relieved by the modulation of the forces described above. This is why it is necessary to add a mechanism for tilting the chassis forward and to the side when it is desired to effectively relieve one of the rear wheels.
    This action therefore consists in modifying the reference attitude used in the setpoint of the postural control described above, by modifying the angles 0b and 0 C , so as to tilt the robot forward and on the side opposite to the wheel to be released or relieved . Thus, the center of gravity is moved in the triangle formed above the other three wheels.
    Once the new position has been reached, the column corresponding to the tab to be relieved in the matrix linking the actuating torques to the torsor of the resulting actions exerted on the chassis is replaced by a zero vector in order to remove the influence of this tab in the calculation of the couples to be applied. To this is also added the change to a control in position of the opposite rear wheel in order to keep the center of gravity of the chassis above the support polygon without having to resort to excessive tilting.
    - Bending of the chassis
    The autonomous vehicle may have a central pivot in order to steer it. This central pivot can be used to move the relative gap between the balance positions of the front and rear wheels. This is particularly useful when encountering a bilateral obstacle: it is then possible to sequence the crossing of the obstacle by the right and left wheels, and therefore isolate the problem posed by the obstacle on a single wheel. The term “bilateral” obstacle is used here to mean an obstacle which occurs simultaneously (or substantially simultaneously) for the right and left wheels of the vehicle.
    When the robot is not subjected to a particular difficulty and the horizontal forces are consequently well balanced between the 4 wheels, the robot maintains its attitude horizontally by distributing the actuations as well as possible on all the legs. As this reference control mode is the most stable and since carrying out the actions generates an additional energy cost, it is preferable to keep the autonomous vehicle there as much as possible. Behavioral policy should therefore not lead to the implementation of a global adaptation action only if it considers that the vehicle has a high probability of remaining blocked by implementing only the compliance mechanisms of carrying legs.
    The central pivot 109 of the autonomous vehicle is used for its steering as long as it does not encounter any obstacle or even if these are considered to be capable of being managed and absorbed by the compliance control of equiparti. As soon as a global reaction mechanism has to be implemented, the pivot is then used as a priority for crossing the obstacle. Once the obstacle has been crossed, the robot can return to its nominal operation and use the central pivot to correct the course, which could have been modified during the crossing of the obstacle. This mechanism has the advantage of requiring only one actuation to perform two functions, which are applied exclusively: steering the vehicle on course and overcoming certain obstacles.
    The observation of the horizontal forces exerted on the four wheels suffices to determine which action should make it possible to overcome the obstacle encountered.
    The calculation of the inverse kinematics makes it possible to know exactly the horizontal force exerted at the end of each leg from the moments of torsion on the articulations of the leg. These torsional moments can be deduced from the angular deviations from the equilibrium positions of the pairs of opposing springs. All 4 forces are then brought to a zero average so as to eliminate the influence of chassis oscillations and to keep only the forces describing the static balance of the structure.
    By noting fi the horizontal force measured for the leg i, and i = l, 2,3,4, we can approximate it by the expression:
    Z = where T c , i is the moment exerted by the lower segment of the leg i on the chassis at the level of the second pivot P2 (or "elbow").
    This moment is given by the expression:
    L ,, = -2 - ^ / 'M in which R c is the radius of the second pivot
    The forces fl, 12, 13, f4 are positive when the wheel is driven back. They respectively represent the left front wheel, the right front wheel, the left rear wheel, and the right rear wheel.
    The choice of the appropriate action to trigger for each situation encountered can be done using an algorithm evaluating the value of the different horizontal forces compared to a set of thresholds. This algorithm can be described as follows:
    If fl> ca and 12> ob, then the control device triggers a modulation of the forces so as to relieve a front wheel. The front wheel to be relieved is the left wheel if fl> 12 and the right wheel if 12> f 1. Line slow rotation of the central pivot can also be triggered in the opposite direction. In order to avoid oscillations, the direction of this command is kept equal to that chosen when it is triggered as long as the criterion "fl> ca and 12> ab" is fulfilled.
    If 13-f4> cb and 13> 0, then the control device triggers a tilt to the front and to the right to relieve the left rear wheel.
    If f4-13> cb and f4> 0, then the control device triggers a tilting forwards and to the left to relieve the right rear wheel.
    If 13> cc and f7> cc, then the control device triggers a rapid rotation of the central pivot to the left if f3> f4 or to the right if f4> 13.
    Toggle actions can be combined with the last action. The first action is not compatible with the others: when several premises of these rules are met, it is primarily applied.
    The thresholds ca, cb, these are the parameters of the command that should be chosen so that the robot performs an action if and only if it would find itself blocked otherwise. It is therefore these thresholds that guarantee to continue in the equiparti compliance control mode as much as possible by limiting the triggering of inappropriate global actions at the chassis level.
    These thresholds can be adjusted using an evolutionary algorithm for which the cost function is expressed by the sum of the time taken to overcome a series of canonical obstacles, the number of action changes and the time spent in different modes of the reference one.
    It is also possible to synthesize a command in the form of a connectionist network by taking inspiration from the previous logical relationships. The final choice of the action to be executed is then made by the principle of "winner takes ail" (the winner takes everything). This has the advantage of providing a differentiable function with respect to its inputs, which makes it possible to set up a search for optimal parameters by descending the gradient.
    It also offers a more general policy class and therefore increases the flexibility of behavioral policies that can be obtained. On the other hand, this increases the number of parameters to be adjusted and therefore the learning time.
    It is also possible to use a learning technique without a priori on the form of the command, such as a Q-learning algorithm. Q-learning techniques are techniques known to those skilled in the art in this field, and for example exposed on the web page:
    https://fr.wikipedia.org/wiki/Q-Leaming
    One can also consult the original reference on this technology which is the thesis “Leaming from delayed rewards” by C. J. Watkins, Ph.D. thesis, Kings College, Cambridge, England, May 1989.
    In this case, it is a question of constructing as many functions from R4 to R as possible discrete actions. The function returning the largest numerical value from the four horizontal forces exerted on the wheels determines the action to be performed. These functions can be modeled using “approximators” of functions such as a Gaussian mixing network or a neural network. There are eleven actions to be taken in this case to cover all the useful possibilities:
    - initial reference mode;
    -modulation of efforts to relieve the left front wheel;
    - force modulation to relieve the right front wheel;
    - tilting to relieve the left rear wheel;
    - tilting to relieve the right rear wheel;
    - rotation of the central pivot to the left;
    - rotation of the central pivot to the right;
    - modulation of efforts to relieve the left front wheel and rotation of the central pivot to the right;
    - modulation of efforts to relieve the right front wheel and rotation of the central pivot to the left;
    - tilting to relieve the left rear wheel and rotation of the central pivot to the left;
    - tilting to relieve the right rear wheel and rotation of the central pivot to the right.
    Of course, the present invention is not limited to the examples and to the embodiment described and shown, but it is susceptible of numerous variants accessible to those skilled in the art.
    权利要求:
    Claims (9)
    [1" id="c-fr-0001]
    1. Autonomous vehicle comprising a chassis and carrying legs, each of said legs being composed of at least two segments, an upper segment (Ls), connected to said chassis by a first pivot (Pi), a lower segment (Li) bearing in its lower end a wheel (R), and connected to said upper segment by a second pivot (P2);
    said lower segment being further connected by two opposing springs (khi, kh2) to a pulley integral (SI) with said chassis, so as to exert an angular stiffness around an equilibrium position corresponding to a vertical orientation of the lower segment; and, said upper segment being connected by two other opposing springs (kvl, kv2), at the level of said first pivot, to a motorized pulley (S2), integral with said chassis, adapted to exercise control of the moment exerted on said upper segment around 'a nominal position corresponding to a horizontal orientation of said upper segment, said moment control (τ) being determined from a measurement of a first angle (0b) formed by the upper segment with respect to its nominal position, d' a measurement of a second angle (0 C ) formed by the lower segment with respect to its equilibrium position, and of the attitude of said chassis.
    [2" id="c-fr-0002]
    2. Autonomous vehicle according to the preceding claim, wherein each leg has a first encoder measuring the first angle 0b located on said first pivot, and a second encoder measures said second angle (0 C ) and located on said second pivot.
    [3" id="c-fr-0003]
    3. Vehicle according to one of the preceding claims, wherein said attitude is determined by two inclinometers determining a pitch angle (0 y ) and a roll angle (0 X ) of the chassis, and by an elevation (p) of said chassis determined as an average of the elongations of said legs of said vehicle.
    [4" id="c-fr-0004]
    4. Vehicle according to one of the preceding claims, in which an action can be triggered to modify the actuation strategy of said vehicle as a function of the measurements of the deformations generated by the ground on said vehicle.
    [5" id="c-fr-0005]
    5. Vehicle according to the preceding claim, in which the force fl measured for the leg i, with i = 1,2,3,4, is expressed by:
    'c where T c , i is the moment exerted by the lower segment of said tab on the chassis at said second pivot (P2), said moment is given by the expression:
    L ,, = -2 - ^ / -kh-0 c in which R c is the radius of said second pivot
    [6" id="c-fr-0006]
    6. Method for controlling an autonomous vehicle according to claim 4, in which a control device determines a moment to be exercised by means of said motorized pulley (S2) from the measurement of a first angle (0b) formed by the upper segment with respect to its nominal position, with a second angle (0 C ) formed by the lower segment with respect to its equilibrium position, and with the attitude of said chassis by the expression:
    τ = G (0 b , θ ( ) M (q) [K p q + K d q] -C (0 b , 0 c ) - N (q, q) in which
    G (0b, 0c) is a matrix representing the influence of the torque of the actuator on the dynamics of the chassis, and G (0b, 0 C ) + its pseudoinverse;
    C (0b, 6c) expresses the influence on the chassis of the stiffness of the springs at the elbows;
    M (q) represents the inertia matrix;
    N (q, q) represents the gravitational, centrifugal and Coriolis forces; </ represents the difference between the state vector q and a reference attitude, determined by a reference elevation and reference pitch and roll angles.
    q is the first derivative of the state vector q with respect to time;
    K p and Ka are two diagonal matrices formed by the correction gains.
    [7" id="c-fr-0007]
    7. Control method according to the preceding claim, wherein said control device can trigger an action to modify the actuation strategy of said vehicle as a function of measurements of the deformations generated by the ground on said vehicle.
    [8" id="c-fr-0008]
    8. Control method according to the preceding claim, in which the force fi measured for the leg i, with i = l, 2, 3, 4 is expressed by:
    'c where T c , i is the moment exerted by the lower segment of said tab on the chassis at said second pivot (P2), said moment is given by the expression:
    L ,, = ~ 2-R c 2 -kh-6 c in which R c is the radius of said second pivot
    [9" id="c-fr-0009]
    9. Computer program having instructions allowing the implementation of the method according to one of claims 6 to 8 when it is executed by an information processing platform.
    1/1
    类似技术:
    公开号 | 公开日 | 专利标题
    WO2018042131A1|2018-03-08|Compliant autonomous vehicle and method for controlling same
    EP3344506B1|2019-07-17|Device and method for correcting the direction of travel of a motor vehicle
    EP2740012B1|2017-11-22|Robot with joints of variable rigidity and method for calculating said optimised rigidity
    FR2976686B1|2019-07-05|METHOD AND SYSTEM FOR CONTROLLING ATTITUDE CHANGE
    EP2551191A1|2013-01-30|Method and device for detecting the deployment of an aircraft control surface.
    WO2006103030A1|2006-10-05|Vehicle steering control system comprising no mechanical linkage between the steering wheel and the steered wheels
    WO2006103029A1|2006-10-05|Vehicle steering control system comprising no mechanical linkage between the steering wheel and the steered wheels
    EP2855240A1|2015-04-08|Vehicle trajectory control device
    EP2855238A2|2015-04-08|Device for controlling the course of a vehicle
    WO2015177173A1|2015-11-26|Wheelchair propulsion method, kit, and wheelchair implementing such a method
    WO2021058918A1|2021-04-01|Methods for learning parameters of a neural network, for generating a trajectory of an exoskeleton and for setting the exoskeleton in motion
    EP3107624A1|2016-12-28|Optimisation of a method for calculating doses deposited by an ionising beam
    EP2766251B1|2016-12-28|Methode for controlling a robot with at least two legs
    FR3094324A1|2020-10-02|Autonomous piloting method of a device trajectory control member
    EP3274780A1|2018-01-31|Improved method for correcting a trajectory in a personal movement assistance device equipped with sensors
    EP2027004B1|2014-05-21|Process and control system for vehicle steering wheel
    FR3099450A1|2021-02-05|Autonomous control method of an actuator of an automotive device
    EP3734392A1|2020-11-04|Methods for guiding an autonomous vehicle
    EP3589533A1|2020-01-08|Device for controlling the trajectory of a vehicle
    FR3093687A1|2020-09-18|Method of autonomous piloting of a mobility device
    FR3076530A1|2019-07-12|STEER BY WIRE-TYPE STEERED STEERING SYSTEM USING TRANSPARENT ACTUATED ACTUATORS USING LOCAL TORQUE AND / OR EFFORT LOCAL LOADS
    FR3082163A1|2019-12-13|METHOD AND DEVICE FOR ASSISTING THE DRIVING OF A VEHICLE BY DETERMINING THE TURNING ANGLE AS A FUNCTION OF A PREDICTIVE GAIN
    FR3100779A1|2021-03-19|A method of controlling an autonomous vehicle comprising a discretization of environmental data.
    FR3020492A1|2015-10-30|METHOD FOR DETERMINING HANDLE TORQUE, PROGRAM FOR IMPLEMENTING SAID METHOD, AND BIKE STIMULATOR
    EP1223104A1|2002-07-17|Guiding process for a gyroscopic moment control system
    同族专利:
    公开号 | 公开日
    WO2018042131A1|2018-03-08|
    FR3055256B1|2018-08-17|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    BE569442A|1957-10-17|1958-07-31|
    FR1493241A|1965-09-16|1967-08-25|Lockheed Aircraft Corp|Articulated all-terrain vehicle|
    FR2508391A2|1981-06-29|1982-12-31|Hildebrand Georges|All-terrain amphibious motor vehicle includes supplementary endless track wheels which can be raised when not required|
    EP1118531A1|2000-01-21|2001-07-25|Ecole Polytechnique Federale De Lausanne|Uneven terrain vehicle|
    CN101380978A|2008-08-08|2009-03-11|山东科技大学|Shrimp-shaped six-wheel mobile robot|
    CN103448831B|2013-09-16|2016-01-13|北京交通大学|A kind of obstacle detouring carrying robot|
    CN108839822B|2018-04-13|2020-04-10|北京控制工程研究所|Wheel-leg composite mobile robot capable of flying repeatedly|
    CN108945520B|2018-07-10|2021-01-29|上海交通大学|Leg type landing patrol robot|
    CN111220113B|2020-01-13|2021-10-19|哈尔滨工程大学|Pipeline corner bending angle detection method|
    法律状态:
    2017-06-16| PLFP| Fee payment|Year of fee payment: 2 |
    2018-03-02| PLSC| Publication of the preliminary search report|Effective date: 20180302 |
    2018-08-30| PLFP| Fee payment|Year of fee payment: 3 |
    2020-05-08| ST| Notification of lapse|Effective date: 20200406 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1658067A|FR3055256B1|2016-08-31|2016-08-31|COMPLIANT AUTONOMOUS VEHICLE AND CONTROL METHOD THEREOF|
    FR1658067|2016-08-31|FR1658067A| FR3055256B1|2016-08-31|2016-08-31|COMPLIANT AUTONOMOUS VEHICLE AND CONTROL METHOD THEREOF|
    PCT/FR2017/052312| WO2018042131A1|2016-08-31|2017-08-31|Compliant autonomous vehicle and method for controlling same|
    [返回顶部]