Method and apparatus for controlling the trajectory of a vehicle traveling in a lane and related vehicles.

Abstract

The present invention relates to a method for controlling in real time the path of a motor vehicle traveling in a lane, the method comprising the steps of detecting a corner in the lane and then, when the motor vehicle enters said corner, determining a first amount and a second amount of a plurality of successive sampling increments based on state variables characterizing movement of the motor vehicle; determining a first stored value and a second stored value, the first stored value depending on a first amount determined at a current sampling increment and a first amount determined for at least one of the preceding sampling increments, and the second stored value depending on a second amount determined at a current sampling increment and a second amount determined for at least one of the preceding sampling increments; The method includes the steps of storing the first and second stored values ​​determined for each sampling increment in a memory, then determining an understeer gradient value according to the first and second stored values ​​stored in the memory when the motor vehicle exits the corner, and determining a command for the motor vehicle based on the understeer gradient value thus determined.

Description

[Technical Field]

[0001] The present invention generally relates to controlling the path of an automated vehicle, particularly to keeping the automated vehicle within its lane during turns.

[0002] More specifically, the present invention relates to a method and device for controlling the path of an automated vehicle traveling in a lane.

[0003] The present invention relates equally to an automated vehicle that includes this type of route control device. [Background technology]

[0004] Autonomous and semi-autonomous automated vehicles are designed to travel on public roads without driver intervention. For this purpose, automated vehicles are equipped with a set of digital sensors that enable the collection of data characterizing the status of the automated vehicle and its environment. Autonomous vehicles are also given software that enables the analysis of that data. This software uses algorithms to then generate commands to steer the automated vehicle. In particular, this software is designed to generate control methods that control the assisted steering system, such as keeping the automated vehicle centered in its lane. This type of control method is conventionally called Lane Centering Assist (LCA) method.

[0005] The control method specifically allows for adjustment of the turning angle of the steerable wheels of the automated vehicle. This turning angle depends on the curvature of the lane, the speed of the automated vehicle, and a parameter known as the understeer gradient, which quantifies the behavior of the automated vehicle when turning. The understeer gradient is 1 m / s 2 This can be defined as the angle that should be given to the wheels of an automatic vehicle subjected to lateral acceleration.

[0006] This parameter is not measurable. Moreover, the range of variation for this parameter is quite large, which can lead to poor setting points for the turning angle and, therefore, significant off-centering of the automatic vehicle when turning.

[0007] Patent application FR3104106 describes a method for determining the understeer gradient. This method is based on changing the amount of oversteer relative to the nominal value as soon as oversteer is observed.

[0008] However, changes encountered along the vehicle's path, such as changes in the vehicle's load, which significantly affect the understeer gradient (and therefore the turning angle), cannot be taken into account instantaneously while the vehicle is moving. [Overview of the project]

[0009] This invention proposes to improve the control of an automated vehicle's path when turning by taking into account turns encountered during the vehicle's movement in real time.

[0010] More specifically, according to the present invention, a method for controlling the path of an automated vehicle traveling in a lane, The steps include detecting a turn in the lane, and then, when the automated vehicle enters the turn, A step of determining a first and second amount of a plurality of consecutive sampling increments based on state variables that characterize the movement of an automated vehicle, A step of determining a first stored value and a second stored value, wherein the first stored value is a function of a first amount determined for the current sampling increment and a first amount determined for at least one of the preceding sampling increments, and the second stored value is a function of a second amount determined for the current sampling increment and a second amount determined for at least one of the preceding sampling increments. The steps include storing the first and second stored values ​​determined for each sampling increment in memory, and then, when the automatic vehicle exits the turn, The steps include determining the value of the understeer gradient as a function of the first stored value and the second stored value stored in memory, The steps include determining a command for the automated vehicle based on the determined understeer gradient value, and A method including this is proposed.

[0011] Therefore, thanks to the present invention, the understeer gradient value is determined in real time at regular intervals while the automated vehicle is moving. More specifically, the understeer gradient value is determined and updated turn by turn throughout the entire movement of the automated vehicle in the lane. The automated vehicle path control setpoint is also adjusted advantageously in real time. This makes it possible to bring the automated vehicle as close as possible to the ideal path centered on the lane (in straight lines as well as in turns) without, in that case, leading to abrupt changes in the path. This makes it possible to ensure the comfort of the occupants of the automated vehicle by preventing jolts when the path control changes.

[0012] The present invention is particularly advantageous in the case of heavy-duty or utility vehicles whose load distribution may fluctuate during movement (e.g., during delivery). The understeer gradient is therefore adjusted throughout the entire movement of the automated vehicle, taking these load changes into account, without external intervention.

[0013] Other advantageous and non-limiting features of the control methods according to the present invention, taken individually or in all technically possible combinations, are as follows: The first stored value is a function of the sum of a first quantity determined for the current sampling increment and a first quantity determined for at least one of the preceding sampling increments, and the second stored value is a function of the sum of a second quantity determined for the current sampling increment and a second quantity determined for at least one of the preceding sampling increments. The state variables that characterize the movement of an automated vehicle are functions of the rotation angle components of the vehicle's wheels, the curvature of the lane, the vehicle's speed, or the vehicle's wheelbase. Furthermore, prior to the turn detection step, there is a step to initialize the understeer gradient value based on a predetermined value. The step of determining the value of the understeer gradient is performed for each turn the automated vehicle makes. Furthermore, there is a step of correcting the value of the understeer gradient in order to determine an intermediate value of the understeer gradient, wherein the intermediate value of the understeer gradient is determined based on a weighting between the determined value of the understeer gradient and a predetermined value. The understeer gradient value is determined based on the ratio between the first stored value and the second stored value. Also, a) A step of determining a first acceleration value and another second acceleration value of an automated vehicle, b) A step of determining the difference between a first acceleration value and another second acceleration value, c) If the determined difference is greater than a predetermined threshold, the step of further correcting the value of the understeer gradient based on a correction value which is a function of the determined difference is given. Turn detection depends on parameters that characterize the movement of the automated vehicle, at least some of which are selected from the angle of the automated vehicle's front wheels, the yaw velocity of the automated vehicle, the lateral offset between the center of gravity of the automated vehicle and the ideal path, the lateral acceleration of the automated vehicle, or the speed of the automated vehicle's movement. The first and second quantities are determined using recursive least squares as functions of state variables that characterize the movement of the automated vehicle. The step of determining a command for an automated vehicle includes a substep of determining the components of the rotation angle of the automated vehicle's wheels.

[0014] The present invention also relates to an automated vehicle comprising a powertrain, a steering system, and a device for real-time path control as described above, adapted to control the steering system.

[0015] The present invention also relates to a device for controlling the path of an automated vehicle traveling in a lane, comprising a computer and a memory having a database with a finite number of locations, wherein the computer To detect a turn in the lane, and then, when the automated vehicle enters the turn, Determining the first and second quantities of multiple consecutive sampling increments based on state variables that characterize the movement of an automated vehicle, Determining a first stored value and a second stored value, wherein the first stored value is a function of a first amount determined for the current sampling increment and a first amount determined for at least one of the preceding sampling increments, and the second stored value is a function of a second amount determined for the current sampling increment and a second amount determined for at least one of the preceding sampling increments. The first and second stored values ​​determined for each sampling increment are stored in memory, and then, when the automatic vehicle exits the turn, The value of the understeer gradient is determined as a function of the first stored value and the second stored value stored in memory, The command for the automated vehicle is determined based on the value of the determined understeer gradient. Regarding devices designed to perform certain tasks.

[0016] The present invention also relates to an automated vehicle comprising a powertrain, a steering system, and a previously introduced device for real-time path control adapted to control the steering system.

[0017] Of course, the various features, variations, and embodiments of the present invention can be associated with each other in various combinations, provided that they are not incompatible or mutually exclusive.

[0018] The following description, with reference to the attached drawings provided as non-limiting examples, will clearly explain what the present invention comprises and how it can be realized. [Brief explanation of the drawing]

[0019] [Figure 1] This is a schematic diagram of a part of an automated vehicle. [Figure 2] This is a representation of the "bicycle" model applied to autonomous vehicles traveling within a lane. [Figure 3] This diagram shows an example of a method according to the present invention for controlling the route of an automated vehicle, in the form of a flowchart. [Figure 4] This figure shows a schematic diagram of a closed-loop method for controlling the route of an automated vehicle. [Modes for carrying out the invention]

[0020] Figure 1 shows an automated vehicle 1 (hereinafter also referred to as "vehicle 1"). Conventionally, this automated vehicle 1 has four wheels 3, a chassis that supports the drivetrain (i.e., the engine and means for transmitting engine torque to the drive wheels), a steering system (for example, equipped with a steering column), bodywork elements, and passenger compartment elements.

[0021] As shown in Figure 1, vehicle 1 also includes a control unit 5. The control unit 5 provides commands and controls various components of vehicle 1. For example, the control unit 5 can receive information from various digital sensors present in vehicle 1, such as a speed sensor or a sensor that measures the rotation angle of the front wheels of vehicle 1.

[0022] The control unit 5 can equally control actuators coupled to the steering column of the vehicle 1, for example, by communicating control setpoints to the actuators. The control unit 5 includes a path control device 10 for this purpose. The path control device 10 is adapted to generate control setpoints. For example, in the case of an autonomous or semi-autonomous vehicle, the path control device 10 allows path control setpoints to be generated to orient the vehicle 1 or to keep the vehicle 1 in a lane, particularly during turns in that lane.

[0023] Here, the control device 10 includes a computer 12 and a memory 14. The memory 14 contains a database. The computer 12 stores in its memory an application consisting of a computer program that, when executed by the processor, includes instructions that enable the computer 12 to execute the application in the manner described below.

[0024] Here, the path of vehicle 1 is modeled by a so-called "bicycle" model. Figure 2 shows a representation of the "bicycle" model applied to vehicle 1 traveling in a lane. In the context of this model, vehicle 1 is modeled by a frame and two wheels (in the case of a bicycle), namely a steerable front wheel 3a and an unsteerable rear wheel 3b.

[0025] The formula introduced below is a determinant.

[0026] The variables considered in this model are as follows: The yaw velocity of vehicle 1, denoted by dψ / dt, corresponds to the rotational speed of vehicle 1 around the vertical axis passing through the center of gravity G of vehicle 1. The azimuth angle, denoted by ψ, corresponding to the angle between the longitudinal axis of the vehicle 1 and the tangent to the path Ideal path I d Linked to the distance of the center of gravity G of the vehicle 1 from The lateral speed of the vehicle 1, denoted by TIFF0007848305000001.tif7170 The center of gravity G of the vehicle 1 and the ideal path I d The lateral offset, denoted by y, corresponding to the offset between The rotational speed of the front wheel 3a with respect to the vertical axis, denoted by dδ / dt The angle of the front wheel 3a, denoted by δ, that is, the angle between the front wheel 3a of the vehicle 1 and the longitudinal axis, and The ideal path I on which the vehicle 1 should be d The position error integral corresponding to the time integral of the offset of the center of gravity G of the vehicle 1 with respect to the ideal path I. This error integral is expressed by the following equation ∫-ydt

[0027] The vehicle 1 is thus represented by what is generally referred to as the state vector (hereinafter "state data X") defined by the following equation TIFF0007848305000002.tif43170

[0028] According to the "bicycle" model, the equation of the path of the vehicle 1 is given by the following equation TIFF0007848305000003.tif9170Where δ req (unit radian, hereinafter denoted as rad) is the angle setpoint (and thus the control setpoint) of the front wheel 3a for the vehicle 1 to remain on the ideal path I in the lane d Or approach the ideal path I d And is the angle setpoint (and thus the control setpoint) of the front wheel 3a for the vehicle 1 to remain on the ideal path I in the lane ρ (unit m -1 ) is the curvature of the lane (or the curvature of the path in the "bicycle" model), B ρ Represents interference data (especially linked to the curvature of the lane), A represents the dynamic relationship with the state data X

[0029] Here, matrix A represents the coefficients c (expressed in Newtons / rad) of the cornering stiffness of the front and rear wheels of vehicle 1. f and c r The distance I between the vehicle's center of gravity G and the front drivetrain, and between the vehicle's center of gravity G and the rear drivetrain 1. f and I r These distances (shown in Figure 2) depend on the mass m (in kg) of vehicle 1 and the longitudinal velocity v (in m / s) of vehicle 1 (hereinafter also referred to as the vehicle's speed).

[0030] The coefficient c of the wheel's cornering stiffness f and c r This is a concept well known to those skilled in the art. For example, the coefficient c of the cornering stiffness of the front wheel. f Therefore, equation F f =2.c f. α f This makes it possible to write F in the formula. f α is the lateral sliding force relative to the front wheel, f This is the rotation angle of the front wheel.

[0031] In the context of the "bicycle" model, the measured value Y1 is also expressed as a function of state data X by the relation Y1=CX, where C is the data containing measured values ​​from various digital sensors included in the vehicle 1.

[0032] In the remainder of this invention, The lateral acceleration of vehicle 1, which corresponds to the normal component of the acceleration of vehicle 1 in the reference frame coupled to vehicle 1 (and is therefore perpendicular to the path), and The transverse acceleration of vehicle 1 corresponds to the acceleration acting on vehicle 1 in a manner perpendicular to the direction of movement of vehicle 1 with respect to a reference frame connected to the ground. It is also defined.

[0033] This "bicycle" model is then used in a control method for the route of vehicle 1, which is stored in the control unit 5. For example, this control method allows vehicle 1 to be kept in the center of the lane when traveling in a straight line or through part of a turn.

[0034] Figure 4 shows a schematic diagram of the closed-loop function of this control method.

[0035] In Figure 4, X ref This corresponds to the ideal path of vehicle 1 within its lane. In practice, this is often the path that passes through the center of the lane in which vehicle 1 is traveling. This ideal path is the path that the control unit 5 requires vehicle 1 to achieve (or maintain).

[0036] For this purpose, the control method takes the form of a looped process. According to Figure 4, the state of vehicle 1, in particular the path of vehicle 1, is given by element 22. This element 22 is actually connected to control unit 5, which sets the rotation angle δ for the front wheels. req The route of vehicle 1 is controlled in a manner that satisfies the equation (for state data X and measurement data Y1) from the "bicycle" model described above, which conforms to the above.

[0037] The functional schematic diagram shown in Figure 4 also illustrates the presence of an observation element 26. This element 26 enables the supply of estimates of the state of vehicle 1. In practice, element 26 is connected to various digital sensors in vehicle 1 and thus receives all measurements related to vehicle 1.

[0038] Element 26 also receives information transmitted by the control unit 5 regarding routing from element 22.

[0039] Element 26 is then the estimated data X est The estimated path of vehicle 1 is generated using the observed data L. For this purpose, element 26 combines the measured values ​​of vehicle 1 and the variables necessary for defining the control method, which are estimated from those measured values.P Generates observational data L. P This is a function of the vehicle's speed.

[0040] In other words, observational data L P This is determined from the movement speed of the vehicle involved (vehicle 1). Estimated data X est In that case, the following equation is satisfied. TIFF0007848305000004.tif9170, L P This is the gain value associated with the observer element 26.

[0041] As shown in Figure 4, the estimated data X est Next, the ideal path X ref This is compared with the estimated path. The difference between the estimated path and the ideal path is handled by element 20. This element 20 is a new control setpoint, for example, the rotation angle δ of the front wheel. req Component δ FBK It is adapted to generate new control setpoints related to the front wheel rotation angle δ. For this purpose, element 20 employs adjustment data Ks. req Component δ FBK The new control setpoints for the estimated path X are est and the ideal path X ref It is obtained by multiplying the difference between the two by the adjustment data Ks. The new control setpoint is therefore a function of the adjustment data Ks. This adjustment data Ks is actually represented in matrix form.

[0042] The adjustment data Ks is a function of the vehicle 1's speed. In other words, the control method shown in Figure 4 uses different values ​​of the adjustment data Ks, each of which is associated with the vehicle 1's moving speed.

[0043] The values ​​of the adjustment data Ks associated with each of the relevant travel speeds are determined when vehicle 1 is designed. These values ​​are therefore fixed before vehicle 1 is put into use. The front wheel rotation angle δ is generated from the adjustment data Ks. req Component δ FBK The control setpoints related to this are therefore called predictions.

[0044] Figure 4 also shows the presence of the predictor element 24. This element 24, in particular, predicts the rotation angle δ of the front wheels necessary to follow that lane. req Component δ FFD By evaluating this, it becomes possible to take the curvature of the lane into account.

[0045] Vehicle 1 is at the center of the turn ( Using the formula introduced to describe the “bicycle” model in the perpetual regime (TIFF0007848305000005.tif7170=0, y=0, and dδ / dt=0), the front wheel rotation angle δ determined by predictor element 24 is given by the formula introduced to describe the “bicycle” model in the perpetual regime. req Component δ FFD The settings related to this should be written in the following format. TIFF0007848305000006.tif8170In formula, L (unit: m) is the wheelbase of vehicle 1. ∇ SV This is the understeer gradient specific to vehicle 1, defined by the following equation: TIFF0007848305000007.tif15170, M f and M r The values ​​(in kg) represent the weights applied to the front drivetrain and rear drivetrain of vehicle 1, respectively.

[0046] The predictor 24 is also connected to various digital sensors in the vehicle 1 and thus receives all measurements related to the vehicle 1.

[0047] As shown in Figure 4, in order for vehicle 1 to move in a turn with a known curvature ρ, the angle to be applied to the steering wheel (and therefore the angle setting point δ to be transmitted to the wheel) is req ) ultimately consists of two components (δ) determined by the observer element 26 and the predictor element 24, respectively. FBK and δ FFD It depends on ). δ req =δ FBK +δFFD

[0048] The present invention therefore relates to the angle to be applied to the steering wheel (and thus the angle setting point δ to be transmitted to the wheel) in order for the vehicle 1 to move in a turn having a known curvature ρ. req The purpose is to determine ).

[0049] The computer 12 of the control device 10 (and more generally the control unit 5) is adapted to perform a method for controlling the route of the automatic vehicle 1.

[0050] The method performed by the computer 12 is adapted to control the path of the automated vehicle 1 in a lane, particularly during turns, in real time. Here, the term "real time" means that the path of the automated vehicle 1 can be controlled at regular intervals as the vehicle 1 moves within the lane.

[0051] For this purpose, computer 12 employs a method that includes several steps, as described below.

[0052] The series of steps employed in the context of this method are shown in the form of a flowchart in Figure 3.

[0053] As shown in Figure 3, this method begins in step E2, when the automated vehicle 1 departs and starts moving within the lane. At this point, the autonomous lane-following function is considered to be activated.

[0054] When activating this function, set the rotation angle δ req To determine the predetermined value ∇ SV_init Understeer gradient value from ∇ SV This includes the initialization step E4. This predetermined value ∇ SV_init This is, for example, a default value stored in memory 14. SV_initFor example, the weight M applied to the front drivetrain and rear drivetrain of vehicle 1, respectively. f and M r And the corresponding stiffness coefficient c f and c r It depends on the rotation angle δ. req The setpoint generated by the control unit 5 is this predetermined value ∇ SV_init Determined based on a predetermined value ∇ SV_init More specifically, the required rotation angle δ req Component δ FFD This enables the determination of the required rotation angle δ. In parallel with this, the observation element 26 determines the required rotation angle δ req The other component δ FBK The rotation angle setpoint δ at startup is estimated. req This is therefore obtained by summing these two components. This starting setpoint is then communicated to the steering system of the automatic vehicle 1.

[0055] The method then proceeds to steps E6-E60. These steps E6-E60 are executed in a loop while vehicle 1 is moving. More specifically, these steps are performed for each successive sampling increment δt of multiple sampling increments δt of the time that automatic vehicle 1 has been moving. This sampling increment δt is, for example, on the order of 10 milliseconds.

[0056] In the relevant sampling increment δt, during step E6, the computer 12 detects whether the lane includes a turn.

[0057] To detect the presence of a turn in a lane, the computer 12 verifies at least the following conditions regarding the characteristic parameters of the movement of the automated vehicle 1. These parameters characterizing its movement are, for example, the angle of the front wheels, the yaw velocity of the vehicle 1, the lateral velocity of the vehicle 1, and the lateral acceleration or lateral offset of the vehicle 1. Alternatively, it may be based on data from map and navigation software.

[0058] Here, a turn is detected, in particular, when the angle of the front wheels, the yaw velocity of vehicle 1, and the lateral velocity of vehicle 1 have the same sign. Another condition for detecting a turn concerns the absolute value of the lateral acceleration between a minimum threshold and a maximum threshold. The minimum threshold is, for example, 0.84 m / s². 2 The order of magnitude is such that the maximum threshold is, for example, 1.5 m / s. 2 This is the order.

[0059] A turn is also detected when the lateral offset is smaller than a predetermined value, for example, less than 1 meter.

[0060] A turn occurs when the time derivative of the yaw velocity is less than a predetermined value over a certain time period, for example, 0.05 rad / s over 1 second. 2 It can also be detected when it is smaller than that.

[0061] This turn detection is only employed when the vehicle 1's speed is greater than the minimum speed threshold for vehicle 1, and when its movement at low speeds only slightly represents the typical movement behavior of the automated vehicle 1 in its lane.

[0062] If no turn is detected in step E6, i.e., if vehicle 1 is traveling on a straight section of the lane, the method proceeds to E8. During this step, understeer gradient ∇ SV_δt The value of is equal to a certain value. This certain value is, for example, a predetermined value ∇ stored in memory 14. SV_init Alternatively, this constant value may be the value of the understeer gradient determined for the preceding sampling increment and stored in the database of memory 14 (this determination is described below).

[0063] As shown in Figure 3, the method then includes step E10, during which the rotation angle setting point δ (using the formula introduced above) is set. req Therefore, in order to determine the control setpoint for the path of the automatic vehicle 1, the understeer gradient ∇ determined in step E8 SV_δtThe value of is used. Predictor element 24, more specifically, the required rotation angle δ req Component δ FFD To determine the understeer gradient ∇ determined in step E8 SV_δt The value of is used. In parallel with this, the observation element 26 determines the required rotation angle δ req The other component δ FBK Estimate the rotation angle setting point δ. req This is therefore obtained by summing these two components. This set point is then communicated to the control system of the automatic vehicle 1.

[0064] The sampling increment is then incremented to perform the steps of the method for the next sampling increment (as described above, the method is performed at regular intervals while vehicle 1 is moving in the lane). The method then returns to step E6.

[0065] If, in step E6, the computer 12 detects that vehicle 1 is in the middle of a turn, then vehicle 1 will therefore turn the detected turn, and the method proceeds to step E20.

[0066] During this step, the computer 12 determines that while the automatic vehicle 1 is driving in the lane, it has been running for a predetermined duration τ since the engine was started. app It is evaluated whether the vehicle traveled in one (or more) turns during that time. In other words, here the computer 12 evaluates whether the vehicle traveled in total for at least this predetermined duration τ which constitutes the learning period for the method. app During this predetermined duration τ, it is determined whether the vehicle has traveled through a turn (in one or more turns). app This is, for example, on the order of 50 seconds, greater than 30 seconds.

[0067] If this is not the case, the method proceeds to step E22, during step E22, the computer 12 determines for the relevant sampling increment a first quantity Φ(δt) associated with the understeer gradient. T.The value of Y(δt) and the second quantity Φ(δt) T .Determine the value of Φ(δt).

[0068] Equation (Equation 3) can be rewritten in the following form, more specifically, involving the state variables Φ and Y that characterize the movement of the vehicle 1 in the lane of the vehicle 1. Y(δt)=Φ(δt).Θ(δt) Wherein Θ(δt)=∇ SV , Y(δt)=δ req -ρL and Φ(δt)=ρv 2 <​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​

[0072] During this step, the value of the rotation angle δ used is... req This is a value obtained in an open loop and measured by sensors related to the automatic vehicle 1 with a sampling increment δt.

[0073] The first quantity Φ(δt) T .Y(δt) and the second quantity Φ(δt) T Φ(δt) is then determined by recursive least squares based on the instantaneous values ​​of the matrix Y(δt) and Φ(δt) for the sampling increment δt.

[0074] As shown in Figure 3, the method proceeds to step E24. During this step, the computer 12 stores the first stored value Φ T .Y and the second stored value Φ T Store .Φ in the database in memory 14.

[0075] The first stored value Φ T .Y is the first quantity Φ(δt) determined for the sampling increment δt (in step E22). T It is a function of Y(δt), but is also a function of the first quantity determined for the preceding sampling increment. The second stored value Φ T The same applies to Φ, and the second stored value Φ T Φ is the second quantity Φ(δt) determined for the sampling increment δt (in step E22). T It is a function of Φ(δt) and is also a function of a second quantity determined with respect to the preceding sampling increment.

[0076] For example, the first (and second) stored value Φ T .Y(Φ each) T Φ) corresponds to the sum of the first (and each of the second) quantities determined for all sampling increments up to the current sampling increment.

[0077] In this case, the computer 12 actually has a first stored value stored for the preceding sampling increment (which itself is obtained from the sum of the preceding first values ​​stored) and a first quantity Φ(δt) determined for the current sampling increment δt. T Store the result of the sum with .Y(δt).

[0078] Alternatively, the stored first (and second) value Φ T .Y may correspond to the average of the first (and each of the second) quantities determined for all sampling increments up to the current sampling increment.

[0079] For example, when starting the automatic vehicle 1, the first quantity Φ(δt) T .Y(δt) and the second quantity Φ(δt) T Φ(δt) is likely to be 0. The first stored value Φ determined in the first sampling increment during the turn. T .Y and the second stored value Φ T Φ, therefore, directly depends on the instantaneous values ​​of matrices Y(δt) and Φ(δt) determined for this first sampling increment during the turn.

[0080] The method then proceeds to step E26, during which the computer 12 determines whether the automatic vehicle 1 has exited the turn detected in step E6.

[0081] If this is not the case, i.e., if the automatic vehicle 1 is still in the turn detected in step E6, the method returns to step E20.

[0082] On the other hand, if Auto Vehicle 1 has exited the turn it was negotiating, the method proceeds to step E28. This, therefore, means that Auto Vehicle 1 is now traveling in a straight line.

[0083] During this step, the computer 12 processes component δ FFD (Therefore, rotation angle δ)req The understeer gradient ∇ used to determine the set point SV_δt_act The value is updated. Therefore, it should be noted that the understeer gradient value is updated only when the automatic vehicle 1 is traveling in a straight line (and thus between two consecutive turns). The understeer gradient value is advantageously updated with each turn while the automatic vehicle 1 is moving. This makes it possible to prevent abrupt changes in the path control setpoint of the automatic vehicle 1, especially during turns, and thus ensure the comfort of the occupants of vehicle 1.

[0084] Here, the understeer gradient value is updated ∇ SV_δt_act The first stored value Φ is determined in step E24, and therefore determined during the turn in which vehicle 1 just exited. T .Y and the second stored value Φ T It is a function of Φ. The updated value of the understeer gradient is ∇ SV_δt_act More specifically, the first stored value Φ T .Y and the second stored value Φ T It is determined as the ratio between Φ and Φ. TIFF0007848305000009.tif14170

[0085] However, in step E20, it is determined that the travel time of vehicle 1 during one or more turns has not reached the predetermined duration τapp, so it is considered that the learning period of this method has not ended. The value of the understeer gradient ∇ determined in step E28 SV_δt_act This is not considered optimal and therefore must be corrected.

[0086] For this purpose, step E30 is set to the midpoint of the understeer gradient value ∇ SV_δt_int To determine the understeer gradient value ∇ determined in step E28 SV_δt_act This is a step to correct for the understeer gradient. The midpoint value of this understeer gradient is ∇ SV_δt_int This is the value of the understeer gradient determined in step E28 ∇SV_δt_act And the predetermined value ∇ used in initialization step E4 SV_init It is determined based on a weighting between the following. In other words, if little turn data is obtained, an adjustment factor is applied to limit the understeer gradient value estimation error. A given duration τ app This adjustment during the learning period allows for linear and gradual convergence of the understeer gradient value, in order to generate the most constant and fluid possible control point (where there is no swaying felt by the occupants of vehicle 1).

[0087] The method then proceeds to step E32, which determines a first acceleration value and another second acceleration value for vehicle 1 for the relevant sampling increment, where, for example, acceleration is the lateral acceleration of vehicle 1 and another acceleration is the lateral acceleration of vehicle 1. The computer 12 then determines the difference between the first acceleration and the second acceleration.

[0088] In step E34, this difference is compared to a predetermined acceleration threshold. This predetermined acceleration threshold makes it possible to identify possible understeer gradient estimation errors, such as those that may be observed when a heavy object is loaded onto the vehicle 1, or in the case of a so-called sharp curve (where lateral acceleration will be high). Here, this predetermined acceleration threshold takes the form of a map, for example. This map shows, for example, that the difference between the first acceleration and the second acceleration is approximately 0.2 m / s². 2 If it is less than a predetermined threshold, the midpoint of the understeer gradient ∇ SV_δt_int This indicates that no correction is applied. Final value of understeer gradient ∇ SV_δt_fin Therefore, the median value of the understeer gradient is ∇ SV_δt_int Equivalent to (step E36a).

[0089] However, the difference between the first acceleration and the second acceleration is approximately 0.2 m / s². 2 If this value is greater than the predetermined threshold, the median value of the understeer gradient ∇ SV_δt_intThis is corrected by a correction value added to this intermediate value (step E36b). This correction value is given, for example, by the map described above. For example, if the difference between the first acceleration and the second acceleration is 1 m / s² 2 If it is greater than 10, the understeer gradient correction value is 1.7 × 10 -3 rad.s 2 The order of magnitude is / m. Final value of understeer gradient ∇ SV_δt_fin Therefore, the median value of the understeer gradient is ∇ SV_δt_int This becomes equal to the sum of the above correction value that is added to it.

[0090] Computer 12 then calculates the final value of the understeer gradient (with or without correction) ∇ SV_δt_fin Using (using the formula introduced above), the rotation angle setpoint δ req Therefore, the route control setpoint for automatic vehicle 1 is determined (step E38).

[0091] In a similar manner to that described in step E10 above, the predictor element 24, more specifically, the required rotation angle δ req Ingredients FFD To determine this, the final value of the understeer gradient obtained in step E36a or E36b is ∇ SV_δt_fin The value of is used. In parallel with this, the observation element 26 determines the required rotation angle δ req Other components δ FBK We estimate the rotation angle δ. req The setpoint is therefore obtained by summing these two components. This setpoint is then communicated to the control system of the automatic vehicle 1.

[0092] The sampling increment is then incremented to perform the steps of the method for the next sampling increment (as described above, the method is performed at regular intervals as vehicle 1 moves in the lane). The method then returns to step E6.

[0093] In step E20, while driving in the lane, the automatic vehicle 1 has a predetermined duration τ app If computer 12 determines that the vehicle has traveled through one (or more) turns during a duration equal to E40, the method proceeds to E40.

[0094] During step E40, the computer 12 calculates the total duration of travel during one or more turns since the last update of the database. tot Determine whether it has reached the target. Here, this total duration τ tot It is greater than 50 seconds.

[0095] Total duration τ tot For example, a predetermined duration τ corresponding to the learning period. app It is proportional to a predetermined duration τ. app If it is 50 seconds, the total duration τ tot For example, this is 100 seconds. In another example, a given duration τ app If it is 30 seconds, the total duration τ tot It is 70 seconds.

[0096] Since the last database update, the total duration of running during a turn τ tot If the value has not been reached, the method proceeds to steps E42 and E44, similar to steps E22 and E24 described above, respectively. Following step E44, the computer 12 then retrieves the first stored value Φ T .Y and the second stored value Φ T Φ is stored in the database of memory 14, and these values ​​are obtained from measurements taken for the current sampling increment δt.

[0097] Similar to the case of E26 described above, in step E46, the computer 12 determines whether the automatic vehicle 1 has exited the turn detected in step E6.

[0098] If this is not the case, i.e., if the automatic vehicle 1 is still negotiating the turn detected in step E6, the method returns to step E20.

[0099] On the other hand, if Auto Vehicle 1 has exited the turn it was negotiating, the method proceeds to step E48. This therefore means that Auto Vehicle 1 is now traveling in a straight line.

[0100] During step E48, the computer 12 processes component δ in a manner similar to that of step E28 described above. FFD The understeer gradient value used to determine the understeer gradient ∇ SV_δt_act Update the value.

[0101] As shown in Figure 3, when this value of the understeer gradient is updated, the method, in the same manner as described above for steps E32, E34, E36a, and E36b, respectively, retrieves the understeer gradient value ∇ obtained in step E48. SV_δt_act Steps E50, E52, E54a, and E54b are followed, allowing the determination of the final value of the understeer gradient from the updated value.

[0102] Next, in step E56, the computer 12 calculates this final value ∇ of the understeer gradient (corrected or uncorrected by the correction value) (in the same manner as in step E38 described above). SV_δt_fin Using (using the formula introduced above), the rotation angle setpoint δ req Therefore, the route control setpoint for the automated vehicle 1 is determined.

[0103] The sampling increment is then incremented to perform the steps of the method for the next sampling increment (as described above, the method is performed at regular intervals when vehicle 1 is moving in the lane). The method then returns to step E6.

[0104] In step E40, the total duration of travel during a turn since the last database update is τ tot If the condition is met, the method proceeds to step E60, during which the database is updated.

[0105] At the start of step E60, the database retrieves the first stored value Φ determined for the previous sampling increment. T .Y and the second stored value Φ T Remember .Φ.

[0106] During step E60, the computer 12 therefore retrieves the first stored value Φ T .Y and the second stored value Φ T Update each of .Φ.

[0107] In reality, computer 12, on the one hand, the first stored value Φ T .Y(each the second stored value Φ) T Determine the first intermediate value (and the second intermediate value, respectively) that is proportional to Φ). The coefficient of the proportional relationship to be applied is, for example, a given duration τ. app and total duration τ tot It is a function of the ratio of [the two factors].

[0108] For example, a predetermined duration τ app This is equal to 50 seconds, and the total duration τ tot In a situation where is equal to 100 seconds, the coefficient of the proportional relationship applied is 1 / 2. The first intermediate value (and the second intermediate value, respectively) is therefore the first stored value Φ T .Y = 1 / 2 (each representing the second stored value Φ) T It is equal to Φ (1 / 2).

[0109] In step E60, the first stored value Φ T .Y and the second stored value Φ T Φ is therefore updated respectively by the first intermediate value and the second intermediate value (by overwriting them). Label "First stored value Φ" T.Y" and "Second stored value Φ" T The .Φ is therefore held after step E60.

[0110] As shown in Figure 3, the method then returns to step E40.

Claims

1. A method for controlling the path of an automated vehicle (1) traveling in a lane, The steps include detecting a turn in the lane, and then, when the automatic vehicle (1) enters the turn, Based on the state variables (Φ, Y) that characterize the movement of the aforementioned automatic vehicle (1), a first quantity (Φ(δt) of a plurality of consecutive sampling increments (δt) Ｔ Y(δt) and the second quantity (Φ(δt) Ｔ The step of determining Φ(δt), The step of determining a first stored value (Φ Ｔ .Y) and a second stored value (Φ Ｔ .Φ), wherein the first stored value (Φ Ｔ .Y) is a function of the first quantity (Φ(δt) Ｔ .Y(δt)) determined for the current sampling increment and the first quantity determined for at least one of the preceding sampling increments, and the second stored value (Φ Ｔ .Φ) is a function of the second quantity (Φ(δt) Ｔ .Φ(δt)) determined for the current sampling increment and the second quantity determined for at least one of the preceding sampling increments; the step of determining a first stored value (Φ Ｔ .Y) and a second stored value (Φ Ｔ .Φ); The first stored value (Φ) determined for each sampling increment (δt) Ｔ . Y) and the second stored value (Φ Ｔ The steps are to save Φ) to memory, and then when the automatic vehicle (1) exits the turn, The first stored value (Φ) stored in memory Ｔ . Y) and the second stored value (Φ Ｔ The understeer gradient (∇) is a function of Φ. ＳＶ＿δｔ＿ａｃｔ ) The step of determining the value, The determined understeer gradient (∇ ＳＶ＿δｔ＿ａｃｔ The steps of determining a command for the automatic vehicle (1) based on the aforementioned value of ) A method that includes this.

2. The first stored value (Φ Ｔ Y) is the first amount (Φ(δt) determined for the current sampling increment. Ｔ . A function of the sum of Y(δt) and a first quantity determined for at least one of the preceding sampling increments, wherein the second stored value (Φ Ｔ Φ) is the second amount (Φ(δt) determined for the current sampling increment. Ｔ The method according to claim 1, wherein Φ(δt) is a function of the sum of a second amount determined for at least one of the preceding sampling increments.

3. The state variables (Φ, Y) that characterize the movement of the automatic vehicle (1) are the components (δ) of the rotation angle of the wheels of the automatic vehicle (1). ＦＦＤ The method according to claim 1 or 2, wherein the function is the curvature of the lane, the speed (v) of the automatic vehicle (1), or the wheelbase (L) of the automatic vehicle (1).

4. The method according to claim 1 or 2, wherein the step of determining the value of the understeer gradient is performed for each turn negotiated by the automatic vehicle (1).

5. The midpoint of the understeer gradient (∇ ＳＶ＿δｔ＿ｉｎｔ The process further includes the step of correcting the value of the understeer gradient in order to determine the intermediate value (∇) of the understeer gradient. ＳＶ＿δｔ＿ｉｎｔ ) is the determined understeer gradient (∇ ＳＶ＿δｔ＿ａｃｔ The method according to claim 1 or 2, which is determined based on a weighting between the aforementioned value and a predetermined value.

6. The aforementioned understeer gradient (∇ ＳＶ＿δｔ＿ａｃｔ The aforementioned value of the first stored value (Φ Ｔ . Y) and the second stored value (Φ Ｔ The method according to claim 1 or 2, determined based on the ratio between Φ and Φ.

7. The steps include determining a first acceleration value and another second acceleration value of the automatic vehicle, A step of determining the difference between the first acceleration value and the other second acceleration value, If the determined difference is greater than a predetermined threshold, the step of further correcting the value of the understeer gradient based on a correction value which is a function of the determined difference. The method according to claim 1 or 2, further comprising:

8. The method according to claim 1 or 2, wherein the detection of the turn depends on parameters characterizing the movement of the automatic vehicle (1), and at least some of the parameters are selected from the angle of the front wheels of the automatic vehicle (1), the yaw speed of the automatic vehicle (1), the lateral offset between the center of gravity (G) of the automatic vehicle (1) and the ideal path, the lateral acceleration of the automatic vehicle (1), or the moving speed of the automatic vehicle (1).

9. The first quantity (Φ(δt) Ｔ Y(δt) and the second quantity (Φ(δt) Ｔ The method according to claim 1 or 2, wherein Φ(δt) is determined using a recursive least squares method as a function of state variables (Φ, Y) that characterize the movement of the automatic vehicle (1).

10. The step of determining the command for the automatic vehicle (1) is the component of the rotation angle of the wheels of the automatic vehicle (1) (δ ＦＦＤ The method according to claim 1 or 2, further comprising a substep of determining ).

11. A device (10) for controlling the path of an automated vehicle (1) traveling in a lane, comprising a computer (12) and a memory (14) having a database with a finite number of locations, wherein the computer (12) The process involves detecting a turn in the lane, and then, when the automatic vehicle (1) enters the turn, Based on the state variables (Φ, Y) that characterize the movement of the aforementioned automatic vehicle (1), a first quantity (Φ(δt) of a plurality of consecutive sampling increments (δt) Ｔ Y(δt) and the second quantity (Φ(δt) Ｔ . Determine Φ(δt) and The first stored value (Φ Ｔ . Y) and the second stored value (Φ Ｔ The first stored value (Φ) is determined. Ｔ Y) is the first amount (Φ(δt) determined for the current sampling increment. Ｔ . A function of Y(δt) and a first quantity determined for at least one of the preceding sampling increments, wherein the second stored value (Φ Ｔ Φ) is the second amount (Φ(δt) determined for the current sampling increment. Ｔ The first stored value (Φ(δt)) is a function of Φ(δt) and a second quantity determined for at least one of the preceding sampling increments. Ｔ . Y) and the second stored value (Φ Ｔ . To determine Φ, The first stored value (Φ) determined for each sampling increment (δt) Ｔ . Y) and the second stored value (Φ Ｔ . Storing Φ) in memory, and then, when the automatic vehicle (1) exits the turn, The first stored value (Φ) stored in memory Ｔ . Y) and the second stored value (Φ Ｔ The understeer gradient (∇) is a function of Φ. ＳＶ＿δｔ＿ａｃｔ ) Determining the value, The determined understeer gradient (∇ ＳＶ＿δｔ＿ａｃｔ To determine a command for the automatic vehicle (1) based on the aforementioned value of ) A device (10) designed to perform the following actions.

12. An automated vehicle (1) comprising a powertrain, a steering system, and a device (10) for real-time route control according to claim 11, adapted to control the steering system.

Images