Method for autonomously controlling actuators of a device - Patents.com

Abstract

The present invention relates to a method for autonomously controlling at least one actuator of a self-propelled device, the actuator being intended to influence the trajectory of the self-propelled device and to control a parameter (V) related to the trajectory of the self-propelled device and the speed (V) of the self-propelled device. y , r, Ψ L , y L , δ, M z , Y ref ) and using a controller (K) to obtain the control setpoints (M z_ref , δ ref According to the invention, the controller used varies according to the speed of the vehicle.

Description

[Technical Field]

[0001] The present invention relates generally to automating path tracking in self-propelled vehicles.

[0002] The invention finds particularly advantageous application in the context of automated vehicle driver assistance, but may also be applied in the fields of aeronautics or robotics.

[0003] The invention more particularly relates to a method for autonomously controlling at least one actuator of a self-propelled device capable of influencing the trajectory of said device, comprising: obtaining parameters relating to the trajectory of the self-propelled device; using a computer to calculate control settings for each actuator as a function of said parameters using a controller; The present invention relates to a method comprising:

[0004] The invention also relates to an apparatus comprising a computer capable of carrying out this method.

[0005] The present invention has more particular, but not exclusive, application to autonomous vehicle obstacle avoidance trajectory tracking. [Background technology]

[0006] With the aim of making motor vehicles safer, such vehicles are now equipped with driver assistance or autonomous driving systems.

[0007] Such systems include in particular collision mitigation braking systems (better known by its abbreviation AEB) which are designed to avoid any collision with an obstacle in the path taken by the vehicle simply by acting on the conventional braking system of the motor vehicle.

[0008] However, there are situations in which such emergency braking systems are unable to avoid a collision or in which such emergency braking systems cannot be used (for example, when there is a vehicle following very closely behind the motor vehicle).

[0009] In such situations, automatic avoidance steering or automatic emergency steering systems (better known by its abbreviation AES) have been developed that make it possible to avoid the obstacle by acting on the vehicle steering or on the vehicle's differential braking system to deflect the vehicle from its trajectory. It will be noted that the obstacle may be in the same lane as the vehicle or in an adjacent lane, in which case it will be detected that this obstacle may be in the vehicle's trajectory in the very near future.

[0010] However, sometimes AES systems impose trajectories on the vehicle that are at the limit of controllability and do not allow the driver to regain control of the vehicle in full safety.

[0011] Therefore, document FR 3099450 describes a solution in which a controller is used to generate control settings that allow the vehicle to remain controllable by the driver of the vehicle if the driver wishes to regain control during the avoidance maneuver. For this, the controller limits the amplitude and rate of change of the direction imposed on the motor vehicle by a hyperbolic tangent function. Although this solution is effective in many configurations, there is room for improvement in its performance (i.e., correct following of the avoidance trajectory).

[0012] More specifically, it is desirable to find a solution that guarantees performance and robustness with regard to vehicle stability when the vehicle speed changes, which is manifested in particular by the following: - Accurate heading and configuration tracking throughout the entire avoidance trajectory; -Good stability Summary of the Invention

[0013] To this end, the invention proposes a control method as defined in the introduction, in which the controller used varies as a function of the vehicle speed (here meaning the speed of the vehicle relative to the road, this speed may have a longitudinal component along the vehicle axes and a lateral component).

[0014] More specifically, the controller continuously varies as a function of vehicle speed.

[0015] Thus, thanks to the invention, the controller used is not the same regardless of the vehicle speed, which allows the control laws for the actuators to be adapted to best suit the vehicle dynamics. Such a solution works well: in particular, the more adapted the controller is to the situation (i.e., the fewer contingencies it has to take into account), the more tolerance it has to control the vehicle, which allows it to avoid obstacles in the safest and most comfortable way for passengers.

[0016] This solution proves to be particularly important in situations where differential braking (in addition to steering) is used, since in such situations the vehicle speed is continuously changing (decreasing) throughout the entire avoidance trajectory, which has a considerable effect on the vehicle behavior.

[0017] Another feature of the present invention provides another advantage.

[0018] Therefore, the solution described below allows hybrid control of steering and differential braking. Specifically, the (raw, pre-saturation) steering set point is calculated based not only on the vehicle dynamic variables, the measured steering angle, and the saturated steering set point, but also on the saturated yaw moment set point and the measured (or estimated) yaw moment, and vice versa. This allows for good coherence between the two command set points (since they are interdependent).

[0019] The structure of the proposed controller is simple, which means that the controller is cheap to use, especially in terms of computing power.

[0020] The optimization is simple, as it simply involves presetting values ​​for some parameters (minimum and maximum vehicle speed in the case of the AES function, minimum and maximum curvature of the avoidance trajectory, ...) and solving a system of linearized matrix inequalities.

[0021] This controller can be varied as a function of the curvature of the avoidance trajectory to better suit the situation.

[0022] This controller is capable of maximizing the performance and robustness of the vehicle's effective trajectory within the limits of the vehicle's controllability.

[0023] As already mentioned, the method used provides good performance, i.e. accurate tracking of position and heading, thereby enabling the vehicle to follow with high precision the avoidance trajectory calculated to avoid obstacles.

[0024] The claimed solution also provides high stability, especially as long as the energy of the disturbances is not unlimited, i.e. as long as the trajectory to be followed has a curvature that stays within acceptable boundaries. In other words, the solution makes it possible to quickly determine whether the avoidance trajectory calculated in terms of dynamics can be achieved by the vehicle, so that the AES function is activated only if this is indeed the case.

[0025] More specifically, it can be realized that the AES function is not activated if the characteristics of the trajectory cross a predetermined threshold. Thus, the concept is not to retroactively deactivate the AES function, but to proactively choose whether to activate the AES function or not, thus making it possible to anticipate when the system is potentially unstable and when the generated trajectory is too imperfect (too much overshoot, too much oscillation, etc.).

[0026] According to the invention, the controller operates even if the initial state of the vehicle (initial heading, initial yaw rate, etc.) at the moment of triggering the AES function is non-zero, which occurs when the vehicle already has a certain dynamic behavior (for example because the vehicle is cornering at the moment of triggering the AES obstacle avoidance function) and which is not the case with the solution described in document FR 3099450. To achieve this result, the controller is synthesized to take into account the initial state of the vehicle.

[0027] Other advantageous, non-limiting features of the method of the invention, considered individually or in any technically possible combination, are: - the self-propelled device is a vehicle having wheels, a power steering actuator, and a differential braking actuator; - the controller includes several components for determining a control setpoint for the power steering actuator and a control setpoint for the differential braking actuator; the controller includes several components including at least one state return gain to be applied to said parameters and at least one saturation compensation gain to be applied to the value of the control setpoint determined in the previous iteration; -The controller can be written in the form of a sum of multiple products of speed-dependent variables and speed-independent local controllers; - each local controller is determined for determined values ​​of a vector of two variation (scheduling) parameters, preferably one of said variation (scheduling) parameters being equal to the speed V of the vehicle and preferably the other of said variation (scheduling) parameters being equal to the reciprocal of said speed; - the controller complies with at least one saturation function modeled for each nonlinear sector; The saturation function follows the setpoint amplitude limiter model and can be expressed in the following form: ψ1(x)=sat η (Κ(ρ).x) The saturation function follows a model that limits the variation of the control set point and can be expressed in the following form: ψ2(x)=sat v (A1x+B1sat η (Κ(ρ).x))-(A1x+B1sat η Κ(ρ).x) In the above equation, K is the controller and sat v is an amplitude limiting function, A1 and B1 are predetermined matrices, and x is a state vector of the free-running machine including the parameters; The controller is based on a modeling of the device in which one output to be minimized is a function of the trajectory tracking error and the heading angle error; - calculating a parameter related to the curvature of the trajectory, and then performing the calculation step on condition that the parameter is included within a predetermined interval; - a control setting for the power steering actuator is calculated as a function of a previously calculated control setting for the differential braking actuator; A control setting for the differential braking actuator is calculated as a function of a previously calculated control setting for the power steering actuator.

[0028] The invention also proposes a device comprising at least one actuator capable of influencing the trajectory of the self-propelled device, and a computer for controlling said actuator, programmed to carry out the method defined above.

[0029] It should be understood that various features, modifications, and aspects of embodiments of the present invention may be combined with one another in various combinations, provided that they are not incompatible or mutually exclusive.

[0030] The following description, with reference to the accompanying drawings given as non-limiting examples, will make it easier to understand what the invention consists of and how it can be put into practice. [Brief explanation of the drawings]

[0031] [Figure 1] 1 is a schematic view from above of a motor vehicle suitable for carrying out the method according to the invention, moving along a road; [Figure 2] 2 is a graph showing parameters used in the context of the method of FIG. 1; [Figure 3] 2 is a schematic diagram of the motor vehicle of FIG. 1 shown in four successive positions along an obstacle-avoidance trajectory from above. [Figure 4] FIG. 2 illustrates a polytope used in the context of the method of FIG. 1. [Figure 5] FIG. 2 illustrates a closed-loop transfer function used to control the motor vehicle of FIG. [Figure 6] 2 is a graph showing the saturation polygon and region of attraction of the controller used in the context of the method of FIG. 1 and an example of how the vehicle state varies in the absence of external disturbances; [Figure 7] 5 is a graph similar to FIG. 4, but showing an example of how the vehicle state fluctuates in the presence of a disturbance. [Figure 8] FIG. 2 illustrates a method for selecting an appropriate controller in the context of the method of FIG. 1. DETAILED DESCRIPTION OF THE INVENTION

[0032] 1 shows a motor vehicle 10 comprising, in a conventional manner, a chassis separating a cabin, two steered front wheels 11, and two non-steered rear wheels 12. Alternatively, these two rear wheels may also be steered through adaptation to a control law.

[0033] The motor vehicle 10 comprises a conventional steering system capable of influencing the orientation of the front wheels 11 so as to be able to turn the vehicle. This conventional steering system in particular comprises a steering wheel connected to a link for rotating the front wheels 11. In the example considered, the motor vehicle 10 also comprises an actuator capable of influencing the orientation of the front wheels as a function of the orientation of the steering wheel and / or as a function of a request received from the computer 13. For that purpose, this actuator may act on the steering column of the vehicle (fixed to the steering wheel) and on the steering rack (connecting the steering column to the steered wheels). Of course, the actuator may be implemented in different ways.

[0034] In addition, the motor vehicle comprises a differential braking system capable of differentially acting on the rotational speed of the front wheels 11 (and, if applicable, the rear wheels 12) so as to slow down the motor vehicle while it is turning. In the example considered, this differential braking system comprises at least one actuator, for example formed by a controlled differential or an electric motor, located at the wheels of the vehicle.

[0035] The computer 13 is therefore intended to control the power steering actuators and the actuators of the differential braking system, and for that purpose the computer 13 comprises at least one processor, at least one memory and various input and output interfaces.

[0036] Thanks to the input interface, the computer 13 is able to receive input signals originating from various sensors.

[0037] These sensors include, for example: -Devices such as front cameras that can determine the position of the vehicle relative to the lane in which it is moving; - a device such as a radar or LIDAR remote detector capable of detecting an obstacle 20 (Fig. 3) in the path of the motor vehicle 10; - at least one lateral device, such as a radar or LIDAR remote detector, capable of observing the environment along the side of the vehicle; a device such as a gyrometer capable of determining the rate of rotation of the yaw (around the vertical axis) of the motor vehicle 10; -Steering wheel position and angular rate sensors, and -A sensor that can estimate the yaw moment that the vehicle is subjected to.

[0038] In practice, no sensors are actually provided to measure the yaw moment, but rather a low-level calculation unit is provided that is able to estimate the yaw moment as a function of the braking torque applied to the wheels of the vehicle.

[0039] Thanks to the output interfaces of the computer 13, the computer 13 can send setpoints to the power steering actuator and to the actuators of the differential braking system.

[0040] Therefore, the vehicle can be forced to follow a predefined avoidance trajectory T0 to avoid the obstacle 20 (see FIG. 3).

[0041] By virtue of the memory of the computer 13, the computer 13 stores data used in the context of the methods described below.

[0042] The computer 13 stores in particular a computer application consisting of a computer program containing instructions which, when executed by the processor, enable the computer to carry out the methods described below.

[0043] Before describing this method, an introduction to the various variables used is given, some of which are shown in FIGS.

[0044] The total mass of the motor vehicle is indicated as "m" and expressed in kg.

[0045] The moment of inertia of the motor vehicle about a vertical axis passing through the center of gravity CG of the motor vehicle is designated "J" and is expressed in Nm.

[0046] The distance between the center of gravity CG and the front axle of the vehicle is "l f " and expressed in meters.

[0047] The distance between the center of gravity CG and the rear axle is "l r " and expressed in meters.

[0048] The cornering stiffness coefficient for the front wheels is "C f " and expressed in N / rad.

[0049] The cornering stiffness coefficient for the rear wheels is "C r " and expressed in N / rad.

[0050] The cornering stiffness coefficient for these wheels is a concept familiar to those skilled in the art. As an example, the cornering stiffness coefficient for the front wheels is calculated using the formula F f =2.C f .α f This allows F f is the side-slip force on the front wheels, and α f is the slip angle of the front wheels.

[0051] The steering angle that the steered front wheels make with the longitudinal axis A1 of the motor vehicle 10 is designated "δ" and is expressed in rad.

[0052] Variable δ expressed in rad ref refers to the saturated steering angle setpoint sent to the power steering actuator.

[0053] Variable δ expressed in rad Κ refers to the non-saturated steering angle setpoint. At this stage, the concept of saturation does not necessarily apply to the variable δ Κ However, the variable δ refIt can simply be specified that the steering angle according to is related to the steering speed limit.

[0054] Variable δ expressed in rad sat refers to the quasi-saturated steering angle setpoint. sat is the unsaturation setting value δ Κ and saturates only with the steering angle. ref is the quasi-saturation setting value δ sat It is calculated based on:

[0055] The orthonormal frame of reference for the vehicle (here defined when the vehicle is on a horizontal surface) has as its origin the vehicle centre of gravity CG. Its abscissa axis X v is oriented along the longitudinal axis A1 of the motor vehicle 10, and its ordinate axis Y v The vertical axis passing through the center of gravity is Z v It is shown as follows.

[0056] Axis Z by differential braking system v The yaw moment expressed in Nm exerted around z It is shown as follows.

[0057] Variable M expressed as Nm z_ref indicates the yaw moment setting to be applied to the wheels using the differential braking means.

[0058] Variable M expressed as Nm zΚ indicates the unsaturated yaw moment. At this stage, the concept of saturation is not necessarily related to the variable M zΚ However, the variable M z_ref , which may simply be specified to be related to the limits of variation of the yaw moment.

[0059] Variable M expressed in rad z_sat refers to the quasi-saturated yaw moment setting value. M z_sat is the variable M zΚ and saturates only in amplitude. The saturation setting M z_refis the quasi-saturation setting value M z_sat It is calculated based on:

[0060] The yaw rate of the vehicle (about a vertical axis passing through the vehicle's center of gravity CG) is denoted "r" and is expressed in rad / s.

[0061] The relative angle between the longitudinal axis A1 of the vehicle and the tangent to the avoidance trajectory T0 (the desired trajectory of the vehicle) is "Ψ L " and expressed in rad.

[0062] The lateral deviation between the longitudinal axis A1 of the motor vehicle 10 (passing through the center of gravity CG) and the avoidance trajectory T0 at the aiming distance "ls" located in front of the vehicle is "y L " and expressed in meters.

[0063] The set value for the lateral deviation between the longitudinal axis A1 of the motor vehicle 10 (passing through the center of gravity CG) and the avoidance trajectory T0 at the aiming distance "ls" located in front of the vehicle is "y L-ref " and expressed in meters.

[0064] The trajectory tracking error is "e yL ” and expressed in meters. The tracking error is calculated by the lateral deviation setpoint y L-ref and lateral displacement y L is equal to the difference between

[0065] The aforementioned aiming distance "ls" is measured from the center of gravity CG and is expressed in meters.

[0066] The slip angle of the motor vehicle 10 (the angle that the motor vehicle's velocity vector makes with the vehicle's longitudinal axis A1) is designated "β" and is expressed in rad.

[0067] The speed of the motor vehicle is indicated as "V" and is expressed in m / s.

[0068] Axis Y v The lateral velocity of the motor vehicle is expressed as "V y " is displayed.

[0069] The constants "ξ" and "ω" represent the dynamic characteristics of the steering angle for the front wheels of the vehicle.

[0070] The constant "g" is -2 This is the acceleration due to gravity, expressed as

[0071] Steering rate indicates the angular rate of rotation of the steered front wheels.

[0072] The method according to the invention is intended to enable the vehicle to autonomously follow the avoidance trajectory T0 as precisely as possible. The method is executed when the automatic obstacle avoidance AES function is triggered and then the avoidance trajectory T0 is calculated. It will be noted that the manner in which the AES function is triggered and the avoidance trajectory T0 is calculated does not strictly speaking form part of the subject matter of the present invention and will therefore not be described here.

[0073] The method is intended to be performed in a loop, in successive "iterations" (where these iterations have a duration of approximately 10 ms).

[0074] It will be noted here that it is intended that path following is performed autonomously by the computer 13, but also needs to be able to be interrupted at any moment to allow the driver to regain control of the vehicle. It also needs to be able to be used as a driving aid when the driver is holding the steering wheel but is not applying the torque to the steering wheel required to avoid an obstacle.

[0075] Before describing the methods performed by computer 13 to implement the invention itself, the first part of this description describes the calculations that led to the invention, in order to allow a comprehensive understanding of where these calculations originate and what they are based on.

[0076] The purpose of this first part of the description is to effectively explain how it is possible to synthesize a controller that, once implemented in computer 13, is capable of controlling the vehicle so that it follows the avoidance trajectory T0 in a stable and well-behaved manner.

[0077] Here, it is considered that the dynamic behavior of the vehicle is modeled by the following equation 1: [Formula 1] TIFF2025527378000002.tif23170

[0078] This model is a classic bicycle model.

[0079] From this equation, the term γ ref It will be noted that allows the vehicle trajectory (and therefore the curvature of the road) to be taken into account when modelling the dynamic behaviour of the vehicle.

[0080] From this equation, in the preliminary state vector used, the first state variable is the lateral velocity V y It will also be noted that the slip angle β could have been used instead. However, this allows for a reduction in the number of variable (scheduling) parameters in the nonlinear model described below, so the lateral velocity V y The use of is preferred.

[0081] The steering of the front wheels 11 can be simply modeled using the following formula: [Formula 2] TIFF2025527378000003.tif9170

[0082] Differential braking, for its part, exhibits a dynamic behavior that can be modeled using the following differential equation: [Formula 3] TIFF2025527378000004.tif6170

[0083] In this equation, τ is the dynamic characteristic of the yaw moment.

[0084] These three equations allow us to express a new model of vehicle behavior. [Formula 4] TIFF2025527378000005.tif66170

[0085] The variation in curvature of the avoidance trajectory T0 can also be modeled using the following equation: [Formula 5] TIFF2025527378000006.tif7170

[0086] In this equation, the variable ω r represents the dynamics of the variation of the trajectory curvature, and the term w refers to the random input that is assumed to be bounded.

[0087] This equation (similar to equation 3) allows us to model the variations in the curvature of the orbit in the form of a low-pass filter.

[0088] Therefore, the previous two equations allow us to express an enhanced model of the vehicle. [Formula 6] TIFF2025527378000007.tif77170

[0089] Note that in this new model, the vehicle speed V is included. Indeed, speed has been found to be an important parameter to take into account, since the differential braking system not only affects the vehicle trajectory, but also the vehicle speed. Vehicle speed also depends on other factors, such as braking or acceleration settings imposed on the vehicle by the driver or computer 13.

[0090] Therefore, in order to synthesize a trajectory tracking controller that is robust in terms of stability and performance (as will be explained in more detail below), it is important to take into account the vehicle's velocity changes.

[0091] So here, the idea is to reformulate the above equation 6 as an LPV (Linear Parameter-Varying) model, which then makes it simple to find the controller that needs to be optimized for this LPV model.

[0092] The improved model is as follows: [Formula 7] TIFF2025527378000008.tif7170

[0093] In this equation, the term ρ 1 / v is equal to the reciprocal of the velocity V (i.e., 1 / V), and the term ρ v is equal to the velocity V. These two terms are commonly called "variation parameters" or "scheduling parameters." Then the vector ρ is (ρ 1 / v ,ρ v ) is written in the form

[0094] State vector x p is defined as follows: [Formula 8] TIFF2025527378000009.tif8170

[0095] Vector u p For, this can be written as follows: [Formula 9] u p =(δ ref ,M z_ref )

[0096] The matrix used itself can be expressed as: [Formula 10] TIFF2025527378000010.tif44170[Formula 11] TIFF2025527378000011.tif39170[Formula 12] TIFF2025527378000012.tif36170[Formula 13] TIFF2025527378000013.tif35170[Formula 14] TIFF2025527378000014.tif35170

[0097] It will be noted that these equations are clearly defined and explained in M. Corno, G. Panzani, F. Roselli, M. Giorelli, D. Azzolini, and S.M. Varesi, "An LPV Approach to Autonomous Vehicle Path Tracking in the Presence of Steering Actuation Nonlinearities," IEEE Transactions on Control Systems Technology, doi:10.1109 / TCST.2020.3006123.

[0098] At this stage, we can rewrite Equation 7 using a very specific form, namely in the form of a polytopic LPV system. To do this, we first need to determine if the vehicle speed is at least V min and maximum limit V max We assume that the eigenvalue is bounded between two boundaries:

[0099] The variation parameter vector ρ is then constrained by the contour of a geometric figure shown in Figure 4, called polytope P1. Here, this polytope takes the form of a triangle. The vertices ρ1, ρ2, and ρ3 of this polytope P1 are defined by the following equations: [Formula 15] TIFF2025527378000015.tif24170

[0100] As will become clearly evident in the rest of this description, the idea of ​​this polytope is to take into account that the controller K (which allows the calculation of the setpoint to be sent to the actuator) depends on the velocity, and that this is made easy to calculate by approximations obtained from the theory of projections onto closed convex spaces.

[0101] Next, the function ρ shown in Figure 4 is a curve C1 that varies between the ends ρ1 and ρ3. A convex space chosen to enclose this curve C1 as tightly as possible is then a polytope P1. Specifically, the goal is to enclose the curve tightly without increasing the number of contingencies to be taken into account, as this would make the controller overly conservative. A triangular shape is the shape that gives the best results.

[0102] Therefore, depending on the theory chosen, a minimum of V min and maximum limit V max For any velocity V that falls between these two bounds, it is possible to compute the controller K as a function of the three values ​​of the controller K at the vertices ρ, ρ, and ρ of the polytope P. All that is needed then is to find these three values ​​of the controller K.

[0103] In fact, it is possible to rewrite Equation 7 as follows: [Formula 16] TIFF2025527378000016.tif7170The above formula, [Formula 17] TIFF2025527378000017.tif7170The above formula, [Formula 18] TIFF2025527378000018.tif8170[Formula 19] TIFF2025527378000019.tif8170[Formula 20] TIFF2025527378000020.tif8170[Formula 21] TIFF2025527378000021.tif8170[Formula 22] TIFF2025527378000022.tif9170[Formula 23] TIFF2025527378000023.tif9170

[0104] However, this model does not in itself make it possible to limit the steering angle and steering speed of the vehicle's front wheels 11, or the amplitude of the yaw moment applied by differential braking and the speed at which this yaw moment varies. Nevertheless, such limits have proven to be particularly important in order to ensure that the vehicle driver will be able to regain control of the vehicle at any moment.

[0105] Such a limit can be expressed using the following formula:

[0106] To limit the steering speed, it can be expressed as follows: [Formula 24] TIFF2025527378000024.tif7170

[0107] To limit the steering angle amplitude, it can be expressed as follows: [Formula 25] |δ ref |≦η δ

[0108] In Equation 24, the coefficient v δ is a constant representing the steering speed that should not be exceeded. This constant is defined by calculation or after a test run carried out on the test vehicle. This constant is equal to, for example, 0.0491 rad / s, which corresponds to 0.785 rad / s (i.e. 45° / s) at the steering wheel when the steering gear ratio is set to 16.

[0109] In Equation 25, the coefficient η δ is a constant representing the steering angle that should not be exceeded. This constant is defined by calculation or after a test run carried out on the test vehicle. This constant is equal to, for example, 0.0328 rad, which in this example corresponds to 0.524 rad (i.e., 30°) at the steering wheel.

[0110] The constraint expressed in Equation 25 allows us to limit the torque exerted by the power steering actuator so that the average driver can manually counteract this torque.

[0111] Specifically, the greater the steering angle, the greater the force applied by the power steering actuator. This limit therefore ensures that the user can regain control of the vehicle without having to apply excessive counter torque. This angle will then depend on the force applied by the type of actuator chosen.

[0112] The constraint expressed in Equation 24 ensures that the driver is not startled by excessively rapid changes in steering wheel orientation.

[0113] It will be noted that the aforementioned values ​​are given as examples and can alternatively be smaller (eg 25° / s and 20° to ensure greater comfort).

[0114] To limit the rate of change of the yaw moment, it can be expressed as follows: [Formula 26] TIFF2025527378000025.tif6170

[0115] To limit the amplitude of the yaw moment, it can be expressed as follows: [Formula 27] |M z_ref |≦η M

[0116] In Equation 26, the coefficient v M is a constant that describes the variation of the yaw moment that should not be exceeded. This constant is defined by calculation or after a test campaign carried out on the test vehicle. This constant is, for example, equal to 2500 Nm / s.

[0117] In Equation 27, the coefficient ηM is a constant representing the yaw moment that should not be exceeded. This constant is defined by calculation or after a test campaign carried out on the test vehicle. This constant is, for example, equal to 2000 Nm.

[0118] These two equations allow the driver to regain control of the vehicle's steering in a controllable manner at any moment, and in particular limit the element of surprise caused by sudden lateral acceleration.

[0119] In accordance with the present invention, it is desirable to limit these four parameters by gradually saturating their values ​​rather than by imposing abrupt thresholds, and it is desirable to take these four parameters into account in the synthesis of the controller, in particular so that the controller behaves in a more stable manner and in a more comfortable manner for the vehicle passengers.

[0120] FIG. 5 shows the control architecture used to control the power steering actuators and differential braking actuators so that the vehicle follows the avoidance trajectory T0 as closely as possible while obeying the constraints mentioned above.

[0121] In this diagram, the controller K is shown as being dependent on the vehicle speed V and consisting of two components, one of which (K Mz ) is associated with differential braking, while the other (K δ ) is associated with power steering.

[0122] These two components result in a yaw moment M zΚ and the steering angle δ Κ It is then possible to calculate the desaturation setpoints for and , respectively.

[0123] Advantageously, each of these two components provides an unsaturated setpoint at its output, which depends on the state of the vehicle (state return unit KMz p , K δ p the state return term (originating from ) and the saturated yaw moment setpoint M calculated in the previous iteration. z_ref or saturated steering angle setting value δ ref depends on the saturation compensation unit K Mz aw , K δ aw and a summer receiving at its input a saturation compensation term (originating from

[0124] The saturation compensation term enhances the stability of the controller in non-linear modes, i.e., when the power steering actuator or yaw moment actuator commands saturate in amplitude or in velocity.

[0125] The unit SAT1 shown in FIG. 5 is a non-saturated steering angle setting value δ Κ The unit SAT1 receives as input the output from the corresponding component of the controller K and calculates the quasi-saturated steering angle setpoint δ sat as the output. You can observe that this unit operates in open loop.

[0126] Unit SAT2 is the unsaturated yaw moment setting value M zΚ The unit SAT2 receives as input the output from the corresponding component of the controller K and calculates the quasi-saturated yaw moment setpoint M z_sat as its output. It can be observed that this unit also operates in open loop.

[0127] The unit SAT3 sets the quasi-saturated steering angle setting value δ sat A set of units SAT3 receives this quasi-saturated setpoint as input and calculates the saturated steering angle setpoint δ ref as the output. It can be observed that this is a closed loop.

[0128] The setting of the unit SAT4 is the quasi-saturated yaw moment setting value M z_sat A set of units SAT4 receives this quasi-saturated setpoint as input and calculates the saturated yaw moment setpoint M z_ref as the output. It can be observed that this is also a closed loop.

[0129] Each of these two sets of units SAT3, SAT4, corresponding to a "pseudo speed limiter" function, therefore has at its input an adder that makes it possible to calculate the difference Δ between the quasi-saturated setpoint and the saturated setpoint of the previous iteration. The adder comprises a multiplier unit that can multiply this deviation by the parameter λ, a saturation unit that ensures that the derivative of the saturated setpoint is not exceeded, and an integration unit to obtain (via the Laplace transform) the saturated setpoint.

[0130] The parameter λ describes the dynamic behavior of units SAT3 and SAT4 (in this application, λ=500 can be considered); the higher the value of λ, the more closely the behavior of this pseudo rate limiter corresponds to that of a rate limiter.

[0131] In Figure 5, unit P sys represents an open-loop system describing the dynamic behavior of the vehicle, the behavior of the power steering actuators, the behavior of the differential braking actuators, and the position of the vehicle relative to the avoidance trajectory T0.

[0132] This unit is the disturbance w, saturated yaw moment setting value M z_ref , and the saturated steering angle set value δ ref It can be observed that the unit receives as input the vector y and the error z as outputs.

[0133] In fact, the output vector y is calculated by dividing the state vector x introduced earlier. p Corresponds to.

[0134] The error z has a value that should be minimized for its part.

[0135] Here, this error z is the tracking error e, which is known to need to be minimized. yL and the relative heading angle between the vehicle's longitudinal axis A1 and the tangent to the avoidance trajectory T0 (hereinafter referred to as the heading error Ψ L is a function of Therefore, it is possible to write: [Formula 28] TIFF2025527378000026.tif6170

[0136] In this equation, the term α ψ is an adjustment coefficient for adjusting the error (heading (or heading angle) error or position tracking error) that should be minimized as a priority. The value of this adjustment coefficient will be explained later in this explanation. This selection of the output error z makes it possible to guarantee both correct position tracking and correct heading tracking.

[0137] Therefore, the objective is to calculate the preliminary state vector x, taking into account the vehicle speed V. p Based on the unsaturated yaw moment setting value M zΚ and the unsaturated steering angle setting value δ Κ The goal is to find a form of controller K that is a state return regulator that can calculate

[0138] To understand how to find a controller K that is suitable in terms of both stability and performance, we first consider the system P without saturation. sys When operating in open loop, i.e., in linear mode, the system P sys (saturation steering angle setting value δ ref and the unsaturated steering angle setting value δ Κ are equal, and the saturated yaw moment setting value M z_ref and the unsaturated yaw moment setting value M zΚ (An equal scenario).

[0139] Here, the system can be expressed in the following general form: [Formula 29] TIFF2025527378000027.tif15170

[0140] Given the contents of Equation 28, the matrix C pz is known. Principal component matrix A p , the command matrix B p , and the disturbance matrix B w It will be noted that can be deduced from Equation 6.

[0141] The controller K, defined as a static state return regulator, seeking the optimal gain for which it satisfies a control criterion, can itself be expressed in the following form: [Formula 30] u Κ =Κ(ρ).x

[0142] In this equation, the term x is the state vector augmented by the saturated steering angle setpoint and saturated yaw moment setpoint. It is this augmented state vector that is considered in the remainder of this description. It can be expressed as: [Formula 31] x=[x p u p ] T

[0143] For the reasons explained above with reference to FIG. 5, the state return gains to be optimized form a 2×2 matrix whose terms depend on the vehicle speed V and can be expressed as follows: [Formula 32] TIFF2025527378000028.tif11170

[0144] As mentioned above, the controller K may be expressed as a function of its values ​​at the vertices ρ, ρ, and ρ of the polytope. More specifically, the controller K may be expressed in the following polytope form: [Formula 33] K(ρ)=K ρ1 α1+Κ ρ2 α2+Κ ρ3α3

[0145] In this equation, the terms α1, α2, and α3 are as defined in equations 21 through 23.

[0146] The matrix K based on which the controller K can be calculated ρ1 , K ρ2 , K ρ3 must itself be determined by the optimization method proposed in the following section.

[0147] Thus, here the idea is to synthesize a reduced number (three) of controllers rather than an infinite number of controllers K. The values ​​of the other controllers can be interpolated as functions of these three matrix controllers via the convex combinations mentioned above.

[0148] Next, System P sys The system Psys can be described for when is operating in closed loop, i.e., in a nonlinear mode with saturation (a scenario where the saturated and unsaturated set points are not equal).

[0149] Given the contents of equations 29 and 30, it can be written as follows: [Formula 34] TIFF2025527378000029.tif6170

[0150] In this formula, the terms A, B, A1, and B1 are defined as follows: [Formula 35] TIFF2025527378000030.tif9170[Formula 36] TIFF2025527378000031.tif8170

[0151] I is the identity matrix. [Formula 37] A1=[0 -Λ] [Formula 38] B1=Λ [Formula 39] TIFF2025527378000032.tif11170

[0152] To take into account the controllability constraints defined in Equations 24–27, two new saturation functions Ψ1(x) and Ψ2(x) are introduced, which represent the crossing of the limits due to unsaturated command inputs.

[0153] Therefore, two saturation functions Ψ1(x) and Ψ2(x) can be defined as follows: [Formula 40] Ψ1(x)=sat η (Κ(ρ).x)-Κ(ρ).x [Formula 41] Ψ2(x)=sat v (A1x+B1sat η (Κ(ρ).x))-(A1x+B1sat η (Κ(ρ).x))

[0154] In these two equations, sat f0 A saturation function, denoted as (f), is used, sat f0 (f) can be defined as follows: [Formula 42] TIFF2025527378000033.tif12170

[0155] It will therefore be noted that these two saturation functions Ψ1(x) and Ψ2(x) take on the value zero in the unsaturated mode and non-zero values ​​otherwise.

[0156] The aim is then to model the saturation function by modelling it using a non-linear sector ("dead zone non-linearity"), based on the works disclosed in the following references: -S. Tarbouriech, G. Garcia, JM Gomes da Silva Jr, and I. Queinnec, "Stability and Stabilization of Linear Systems with Saturating Actuators", 1st edition, London: Springer 2011 -Alessandra Palmeira, Joao Manoel Gomes da Silva Jr, Sophie Tarboureich, and I. Ghiggi, "Sampled-data control under magnitude and rate saturating actuators", International Journal of Robust and Non-linear Control, Wiley, 2016, 26(15), pp. 3232-3252

[0157] To model the limits of controllability of the system, two polygons can be defined, one denoted S1(x,η) to model the amplitude saturation behavior and the other denoted S2(x,Ψ1,v) to model the velocity saturation behavior. These two polygons can be modeled in the following form: [Formula 43] S1(x,η)={x∈R n ,|(Κ(ρ)-G1(ρ))x|≦η} [Formula 44] S2(x,ψ1,v)={x∈R n ,ψ1∈R n ,|(A1+B1Κ(ρ)-G2(ρ))x+(B1-G3(ρ))ψ1(x)|≦v}

[0158] This modeling includes matrices G1, G2, and G3, which are the same size as the matrices of the controller K. These matrices indicate the excursions that are allowed to exceed the saturation condition.

[0159] For example, the matrix G1 is composed of the matrix K and the scalar α Κ In that case, Equation 44 can be considered to be equal to the product of the term (1-αΚ ), which then constitutes a setting for the permissible exceedance of the limit. The extent to which this limit can be exceeded can be set to, for example, 10%. To set this limit range, road tests need to be carried out.

[0160] However, by preference, this matrix G1 is not a function of the matrix K. This matrix G1 should be optimized not by road testing but by calculation, for example using the linear matrix inequality method, so as to avoid the conservative tendency of the solution as explained in the previous section.

[0161] In a first step, we can assume that the system is operating in closed loop (as defined in Equation 7), but that the state vector x and the variable Ψ1 lie within these two polygons, which can be written as follows: [Formula 45] TIFF2025527378000034.tif11170

[0162] These two polygons are now represented in Figures 6 and 7 and therefore represent the two spaces in which system stability and performance are guaranteed, as clearly explained above.

[0163] At this stage, it will be noted that these figures show two-dimensional graphs that simplify the above solutions in a more easily understood representation.

[0164] In other words, this two-dimensional representation is only valid if the state vector x contains only two state variables, one of which forms the abscissa value and the other the ordinate value of each of these graphs.

[0165] In practice, the state vector here contains 10 state variables, so our representation needs to be plotted in 10 dimensions.

[0166] Together with the assumptions made in the above formula, the following inequality is satisfied for all positive diagonal matrices U1 and U2, where U1 and U2 are more specifically positive scalar quantities: [Formula 46] TIFF2025527378000035.tif13170

[0167] It will be noted here that these scalar quantities U1 and U2 are introduced here simply to facilitate subsequent calculations (using the S procedure).

[0168] In summary, Equations 43 and 44 are two saturation models for amplitude saturation and velocity saturation, respectively, which ensure that Equation 46 is valid as well, while still being valid within the meaning of Equation 45.

[0169] These models are then modeled as state return main gains K that are linked to the state variables of the system P operating in open loop. δ p , K Mz p and the saturation compensation gain ("anti-windup" gain) K of the controller K. δ aw , K Mz aw and can be used to synthesize (i.e., optimize)

[0170] 34 in closed-loop operation, and the system P as shown in Figure 5 sys By using this expression, it is possible to write it as follows: [Formula 47] TIFF2025527378000036.tif11170

[0171] We assume that the disturbance w is bounded in energy, i.e., bounded by a limit, which allows us to write: [Formula 48] TIFF2025527378000037.tif7170

[0172] This assumption is linked to the fact that the curvature of the trajectory (here considered a disturbance) and the duration for which the AES function is activated are always bounded by limits.

[0173] Therefore, in Equation 48, where the term w is linked to the curvature of the escape trajectory, the maximum value of this term w max is known (because the dynamic limits of the vehicle and therefore the maximum curvature that the vehicle can follow in complete safety and thus can be imposed by the trajectory planning are known).

[0174] To obtain the optimal solution for the controller K, the Lyapunov function V(t) for the stability condition is first defined. [Formula 49] V(t)=x T Px

[0175] In this equation, the matrix P is defined to be positive and symmetric.

[0176] The advantage of this Lyapunov function V(t) is that if the first derivative is strictly negative, it is guaranteed that the system will always be stable in the absence of disturbances.

[0177] 6 and 7 show two polygons that guarantee system performance and stability. Therefore, a space that is simultaneously located in both of these two polygons is sought, which can be, for example, the intersection of these two spaces.

[0178] However, we prefer to model this space in the form of an ellipse called a region of attraction, which is defined as follows: [Formula 50] ε(P,μ)={x∈R n ,x T Px≦μ -1}

[0179] In addition to the stability condition on this Lyapunov function V(t), this region of attraction must be included within two polygons S1 and S2 that model amplitude saturation and velocity saturation, respectively, to guarantee system stability (which remains a priority to be guaranteed).

[0180] The domain of attraction ε is therefore the space of stability (or invariance) of the system involved. In other words, it is the space within which the trajectories of the state variables (i.e., the components of the state vector x) remain, provided that they are initialized within this space (even when the system is subject to actuator disturbances and saturation).

[0181] With the above equations in mind, the space through which the system state variables can move is shown diagrammatically in FIGS.

[0182] As can be seen from the diagram, a controller K is then synthesized to meet three objectives.

[0183] The first objective is that in the absence of disturbances, the controller K should ensure that the trajectories of the state variables of the system operating in closed loop remain within the region of attraction ε (thereby guaranteeing stability) and, in particular, converge asymptotically towards the origin (thereby guaranteeing performance) within a predefined length of time.

[0184] Figure 6 considers the disturbance-free case, where the space for the system's initial conditions, denoted E0, coincides with an estimate E1 of the region of attraction ε. Here, we see that the trajectory T1 of the state vector x starting from any initial state indeed converges towards the origin.

[0185] The second objective is that in the presence of disturbances, the controller K should guarantee that the trajectories of the state variables of the system operating in closed loop remain within the region of attraction ε (thereby guaranteeing stability) independent of the disturbance w, provided that this disturbance is bounded in energy terms (within the sense of Eq. 48).

[0186] Figure 6 considers the case where there are no disturbances. In this case, the space for the initial conditions of the system, denoted E0, must be contained within the estimate E1 of the domain of attraction ε. Here, we see that starting from any initial state contained within the space E0, the trajectory T2 of the state variables will indeed remain within the domain of attraction ε.

[0187] The third objective is that in the linear mode (without amplitude or velocity saturation), the controller K should guarantee the performance of the system, and then the performance of the system is prioritized over stability by contriving that the resultant of the norms H∞ is smaller than a predetermined scalar quantity.

[0188] This composition gives the term F l (P sys Recall that the goal is to find a controller K such that the norm H∞ of (K, K) is minimized. [Formula 51] z=F l (P sys ,Κ).w

[0189] This formula can be written in the following form: [Formula 52] TIFF2025527378000038.tif7170

[0190] Note that γ is the norm H∞ of the transfer function w→z. In linear mode (without saturation), this constraint results in a guarantee on performance (in norm H∞).

[0191] This synthesis suggests good rejection of the disturbance w and good tracking of the avoidance trajectory (with error z close to zero).

[0192] In practice, there are several methods that can be used to meet these three objectives.

[0193] Preferably, the method used is the use of linear matrix inequalities (LMIs), which are implemented based on convex optimization criteria together with linear matrix inequality constraints (the linearity of the terms in the matrices used guarantees that the mathematical problem can be solved without excessive computational workload).

[0194] More specifically, the objective is to optimize the gain of the closed loop defined by the controller K by varying the pole selection.

[0195] Specifically, if there exist matrices R(ρ) Q, L1(ρ), L2(ρ), T1(ρ), T2(ρ) with appropriate dimensions such that the following optimization problem is feasible, then a controller K that satisfies the three objectives mentioned above can be obtained.

[0196] These matrices are calculated as a function of the matrix P.

[0197] The matrix inequalities used here are three, defined by the following inequalities, where γ should be minimized: [Formula 53] TIFF2025527378000039.tif16170[Formula 54] TIFF2025527378000040.tif12170[Formula 55] TIFF2025527378000041.tif12170

[0198] In these inequalities, i is a consecutive integer equal to 1, then 2, and so on.

[0199] term X (i) corresponds to the i-th row of the matrix X.

[0200] Furthermore, in these inequalities, A matrix in the form of TIFF2025527378000042.tif9170 is It is written in the format TIFF2025527378000043.tif9170.

[0201] The matrix variables R, Q, L1, L2, T1, and T2 are expressed in the form of matrices of appropriate dimensions.

[0202] The matrix variables are expressed in the following form: [Formula 56] TIFF2025527378000044.tif30170

[0203] To solve the optimization problem in the problem situation described below, it is necessary to solve matrix inequalities (LMIs) only for each vertex (ρ1,ρ2,ρ3) of the polytope shown in Figure 4, which results in solving nine inequalities.

[0204] At this stage, the term K ρ1 , K ρ2 , K ρ3 To optimize each of these matrices, the vehicle speed V is assumed to be constant (hence all of these matrices are considered constant). Specifically, each term is optimized with respect to the determined vector ρ, i.e., the determined speed.

[0205] The three inequalities in Equations 53 through 55 ensure that the closed-loop dynamic range remains bounded, i.e., the system remains stable in the absence or presence of disturbances (thus satisfying the first two conditions).

[0206] Furthermore, the first inequality in Equation 53 guarantees the performance (in the H∞ norm sense) of the system operating in closed loop when the system is subjected to disturbances. This inequality therefore guarantees that the third condition is met.

[0207] Once the nine matrix inequalities are solved, the local controllers at the vertices of polytope P1 are calculated as follows: [Formula 57] K ρ1 =R(ρ1)Q -1 [Formula 58] K ρ2 =R(ρ2)Q-1 [Formula 59] K ρ3 =R(ρ3)Q -1

[0208] Therefore, it is possible to obtain the controller K(ρ) since it is a convex combination of these local controllers. More specifically, the controller is found using equations 33 and 21-23.

[0209] At this stage, it should be noted that, as a preference, the controller K to be used to define the control settings for the actuators may depend not only on the vehicle speed V but also on the shape of the avoidance trajectory T0.

[0210] To understand what this is, first consider Equation 48, where the term w represents the curvature of the avoidance trajectory, and (since the dynamic limits of the vehicle, and therefore the maximum curvature that the vehicle can follow completely safely and thus be imposed by the trajectory plan, are known) the maximum value for this curvature, w max It must be noted that is known. As a result, the parameter σ can be defined as: [Formula 60] TIFF2025527378000045.tif11170

[0211] In this equation, time T AES corresponds to the maximum duration of activation of the AES function, which is generally between 1 and 3 seconds, and in particular corresponds to the duration of an evasive maneuver.

[0212] Therefore, when the avoidance trajectory T0 is defined, it is possible to calculate the value of this parameter σ.

[0213] The maximum interval for the curvature of the avoidance trajectory T0 that the vehicle can achieve can be divided into N uniform subintervals. Therefore, the term σ can be considered to belong to one particular interval and can be written as follows: [Formula 61] σ∈[σi ,σ i+1 ]

[0214] In the above formula, i is a natural number ranging from 1 to N.

[0215] Each interval can be defined by its mean value and can be written as: [Formula 62] TIFF2025527378000046.tif6170

[0216] The performance of the controller K depends on the curvature of the avoidance trajectory T. Specifically, the smaller the curvature (i.e., the larger the radius of curvature at each point on the trajectory), the easier it is to follow the avoidance trajectory T.

[0217] The proposed concept, which optimizes using linear matrix inequalities, requires fixing the term σ (which is inversely proportional to the curvature) to find the controller.

[0218] In other words, the resulting controller is optimal for only one curvature considered. However, if the vehicle has to follow a trajectory with a larger curvature, this controller will not be as robust, leading to the risk of instability. Conversely, if the vehicle is made to follow a trajectory with a smaller curvature, the controller will not perform as well (there will be delays or overshoots in following).

[0219] Then, upon triggering of the AES function, an avoidance trajectory T0 adapted to the situation is calculated over a fixed time horizon (the next 3 seconds).

[0220] As a preference, a controller K can then be synthesized to match the curvature of the avoidance trajectory T0.

[0221] More specifically, the above inequality allows for the synthesis of several controllers K(ρ), each of which is in one of the above intervals (more specifically, the average of the above intervals). TIFF2025527378000047.tif5170). These controllers are then associated with K(ρ, It can be written in the form of TIFF2025527378000048.tif5170).

[0222] Then, the term σ calculated for the avoidance trajectory T0 is closest to It is now possible to select a composite controller for the value TIFF2025527378000049.tif5170.

[0223] In practice, as shown in Figure 8, the calculation module B1 can determine the avoidance trajectory T0 and calculate the value for the parameter σ. After that, the switch B2 selects the controller K(ρ, You will be able to select the file format (TIFF2025527378000050.tif5170).

[0224] Therefore, the controller will depend not only on the vehicle speed V but also on the curvature of the avoidance trajectory T0.

[0225] At this stage, α ψ It should also be briefly noted that there are several types of controller K that can be obtained, depending on the chosen values ​​of α. ψ It is possible to change the performance of the controller K by adjusting the value of .

[0226] Therefore, the adjustment coefficient α ψ It is possible to obtain a controller K that minimizes the position tracking error when the adjustment coefficient α ψ When the value of is large, it is possible to obtain a controller K that minimizes the heading tracking error. At the start of avoidance, the adjustment coefficient α ψ is chosen to have a small value (less than 20) to ensure that the vehicle follows the avoidance trajectory T0 correctly. Meanwhile, at the end of the avoidance maneuver (after passing the obstacle), the adjustment coefficient αψ is chosen to have a large value (greater than 20) to ensure that the vehicle is indeed repositioned parallel to the road.

[0227] Thus, for example, if the driver wants to regain control at the beginning of the avoidance maneuver, by going further around the obstacle than intended (within the meaning of the avoidance trajectory T0), the adjustment factor α at the end of the avoidance maneuver ψ A large value of means that the setpoint will not unnecessarily bring the vehicle back onto the avoidance trajectory when this trajectory is far too far (as this would otherwise be unsettling for the driver).

[0228] Now that the computational hypotheses have been clearly established, the method executed by the motor vehicle's computer 13 to implement the present invention can be described.

[0229] Here, the computer 13 is programmed to carry out the method recursively, i.e. iteratively in a loop.

[0230] To do this, in a first step, the computer 13 verifies that the autonomous obstacle avoidance function (AES) can be activated and that an obstacle avoidance trajectory has been planned by unit B1.

[0231] The computer 13 then calculates a controller K(ρ, Calculate the parameter σ to select the image (TIFF2025527378000051.tif5170).

[0232] At this stage, it may be preferred that the computer suspend the process and not activate the AES function if the parameter σ exceeds a predetermined threshold, since it is assumed that in such an eventuality the process is unable to achieve completely safe obstacle avoidance.

[0233] If not, the AES function is activated.

[0234] The computer 13 then tries to define certain control settings for the conventional steering system and other control settings for the differential braking system that will allow this avoidance trajectory T0 to be followed as closely as possible.

[0235] To do this, start by calculating or measuring the following parameters: - measured steering angle δ, -Estimated yaw moment M z , - the derivative with respect to time of the measured steering angle δ -Transverse velocity V y , - yaw velocity r, - Opposing neck angle Ψ L , -Skidding L , - Curvature of the avoidance trajectory γ ref , - the saturated steering angle setpoint δ obtained in the previous iteration ref , - the saturated yaw moment setpoint M obtained in the previous iteration z_ref

[0236] The computer 13 then calculates the selected adjustment factor α ψ and the matrix K obtained from data corresponding to the curvature of the avoidance trajectory T0. ρ1 , K ρ2 , K ρ3 Get the formula.

[0237] The controller K is then calculated as a function of the vehicle speed V by predetermining the values ​​of the coefficients α1, α2, α3.

[0238] Next, the controller K determines the unsaturated steering angle set value δ Κ and saturated steering angle setting value δ ref and the unsaturated yaw moment setting value M zΚ and saturated yaw moment setting value M z_refIt is possible to obtain values ​​for and .

[0239] The matrix K of the controller K is calculated taking into account the saturation function so that the settings fit perfectly to the selected saturation model. ρ1 , K ρ2 , K ρ3 It will be noted that is synthesized.

[0240] Finally, a saturated steering angle setpoint δ is used to turn the wheels of the motor vehicle 10. ref to the power steering actuator. Similarly, to brake the wheels of the motor vehicle 10, a saturated yaw moment setpoint M z_ref is transmitted to the actuator of the differential braking system.

[0241] The method is then repeated in a loop over the entire avoidance trajectory T0.

[0242] The invention is in no way limited to the embodiments described and shown, and a person skilled in the art will know how to modify the embodiments in any way in accordance with the invention.

[0243] Therefore, the method can be applied to other types of fields where a specific trajectory should be followed, for example in aeronautics or robotics (especially when the robot is small and one of the robot's commands needs to be saturated).

Claims

1. 1. A method for autonomous control of at least one actuator of a self-propelled device (10) capable of influencing the trajectory of said self-propelled device (10), comprising: - the trajectory of the self-propelled device (10) and parameters relating to the speed (V) of the self-propelled device (10) (V ｙ , r, Ψ Ｌ , y Ｌ , δ, M ｚ , γ ｒｅｆ ) and - using a computer (13), the parameters (V ｙ , r, Ψ Ｌ , y Ｌ , δ, M ｚ , γ ｒｅｆ ) as a function of the control setting (M ｚ＿ｒｅｆ , δ ｒｅｆ ) using a controller (K); Including, The controller (K) used varies as a function of the speed (V) of the vehicle and the parameter (V ｙ , r, Ψ Ｌ , y Ｌ , δ, M ｚ , γ ｒｅｆ ) to be applied to Ｍｚ ｐ , K δ ｐ ) and at least one saturation compensation gain (K) to be applied to the control setpoint value determined in the previous iteration. Ｍｚ ａｗ , K δ ａｗ ) and several components including

2. 2. The control method according to claim 1, wherein the self-propelled device (10) is a vehicle having wheels (11, 12), a power steering actuator, and a differential braking actuator, and the controller (K) includes several components for determining control settings for the power steering actuator and control settings for the differential braking actuator.

3. Variables (α1, α2, α3) that depend on the velocity (V) and a local controller (K) that does not depend on the velocity (V) ρ１ , K ρ２ , K ρ３ 3. The control method according to claim 1, wherein the controller (K) can be written in the form of a sum of multiple products of

4. Each local controller (K ρ１ , K ρ２ , K ρ３ 4. A control method according to claim 3, wherein ρ is determined for determined values ​​of a vector (ρ) of two variation (scheduling) parameters, preferably one of said variation (scheduling) parameters being equal to the speed V of the vehicle and preferably the other of said variation (scheduling) parameters being equal to the reciprocal of said speed.

5. 5. Control method according to any one of claims 1 to 4, wherein the controller (K) is subject to at least one saturation function modelled for each non-linear sector.

6. the saturation function follows a set point amplitude limiter model; ψ １ (+) = sat η (K(r).x)-K(r).x In the above formula, K is the controller and sat η 6. The control method according to claim 5, wherein x is a limiting function and x is a state vector of the self-propelled device (10) including said parameters.

7. The saturation function is determined by the control setting (δ ｒｅｆ ) according to a model that limits the variation of ψ ２ (+) = sat ｖ (A) １ x+B １ sat η (K(r).+))-(A １ x+B １ sat η K (r).+) In the above formula, K is the controller and sat ｖ 7. A control method according to claim 5 or 6, wherein: is an amplitude limiting function; A1 and B1 are predetermined matrices; and x is a state vector of the self-propelled device (10) including said parameters.

8. The controller (K) determines that one output (z) to be minimized is the tracking error (e ｙＬ ) and heading angle error (Ψ Ｌ 8. A control method according to any one of claims 1 to 7, characterized in that it is based on a modelling of the device (10) that is a function of

9. 9. A control method according to claim 1, wherein the self-propelled device (10) is a vehicle, a trajectory (T0) to be followed by the vehicle is planned, a parameter (σ) relating to the curvature of the trajectory (T0) is calculated, and the calculation step is executed on the condition that the parameter is included within a predetermined interval.

10. 10. The control method according to claim 1, wherein the self-propelled device (10) is a vehicle having wheels (11, 12), a power steering actuator, and a differential braking actuator, and the control setting value for the power steering actuator is calculated as a function of the control setting value for the differential braking actuator determined in a previous iteration and / or the control setting value for the differential braking actuator is calculated as a function of the control setting value for the power steering actuator determined in a previous iteration.

11. A self-propelled device (10) comprising at least one actuator capable of influencing the trajectory of the device (3) and a computer (13) for controlling the actuator, characterized in that the computer (13) is programmed to carry out the method according to any one of claims 1 to 10.

Images