Improved method for allocating actuator commands of a motor vehicle

Abstract

Improved method for allocating actuator commands of a motor vehicle The invention relates to a method for allocating actuator commands of a motor vehicle comprising the use of an active set for determining an optimal vector of actuator commands. Figure for abstract: Fig. 2

Description

Title of the invention: Improved method for allocating actuator commands of a motor vehicle Technical field

[0001] The present invention relates to the field of controlling motor vehicle equipment.

[0002] It relates more specifically to a method for controlling vehicle movement, or "vehicle motion control" (VMC) in English. The vehicle may, for example, have four steered wheels and implement a plurality of actuators acting in particular on the steering angle and on the differential braking of the wheels.

[0003] The invention also relates to a control system configured to implement the control method as well as a motor vehicle comprising such a control system. Prior art

[0004] It is known to implement motor vehicles in which all four wheels are steered (4RD or 4WS for "four-wheel steering" in English). The steering of the two rear wheels is typically implemented by a single actuator but can also be implemented by two separate actuators, each acting on one of the two rear wheels.

[0005] A motor vehicle with four steering wheels can be steered more easily and has better stability and maneuverability compared to a vehicle with two steering wheels in which only the front wheels can be steered.

[0006] Thesis [1] describes a vehicle motion control system, the vehicle having four steered wheels and a differential braking actuator for each of the four wheels.

[0007] The described control system can in particular be implemented to optimize the controls making it possible to distribute a desired yaw moment on the differential braking actuators of the four wheels and on the actuator controlling the steering of the rear wheels.

[0008] [Fig. 1] shows the block diagram of such a control system for controlling the yaw rate of a motor vehicle. As illustrated, this system operates in a closed loop.

[0009] First, a reference model 1 is used to determine the desired yaw rate of the vehicle ïPref, i.e. the time derivative of the desired yaw angle ^ref. The desired yaw rate depends in particular on the driver's actions, for example on the steering wheel or pedals of the vehicle, and / or actions of a control unit of a partially or fully autonomous vehicle. Without this being limiting, the reference model 1 can be defined by a bicycle model known as such, by a closed-loop regulator or by any means making it possible to define a yaw setpoint representative of the desired behavior.

[0010] A high-level software controller 2 determines a yaw moment Mz corresponding to the desired yaw rate. Then, a command allocation unit 4 implements a command allocation method to determine the commands to be optimized taking into account the limitations of the chassis 3.

[0011] The commands to be optimized by the allocation method include the braking force (or longitudinal force) at the left front wheel Fxfi, the braking force at the right front wheel Fxfr, the braking force at the right rear wheel Fxn> the braking force at the left rear wheel Fxrl and the lateral force at the rear wheels Fyr linked to the steering of the rear wheels.

[0012] These commands are transmitted to corresponding low-level software controllers 5, 6, 7, 8, 9 which translate these force commands respectively into braking accelerations rb.fi, rb.fr, rbji, rbjIT for each of the four front left, front right, rear left, rear right wheels respectively, and into the steering angle of the rear wheels ôr.

[0013] The vehicle actuators and the vehicle 10 then implement the commands determined by the control system. Vehicle sensors are used to measure the movement of the vehicle, including the yaw rate 7' of the vehicle which is compared to the desired yaw rate of the reference model.

[0014] The high-level controller 2 then determines a new yaw moment Mz to be reproduced as a function of the desired yaw rate and the measured yaw rate and the command allocation process is repeated.

[0015] Thus, the closed-loop system as described establishes an optimal allocation of the controls on the differential braking actuators each acting on one of the four wheels of the vehicle and on the rear wheel steering actuator, called the four-wheel steering actuator or 4RD actuator.

[0016] The commands sent to the different actuators can be calculated by a command allocation method known as such, for example implementing a constrained optimization algorithm known as such. An example of a suitable command allocation algorithm is an active set algorithm as described in the thesis [2]. This algorithm optimizes the commands according to a matrix linking the actuator commands to the setpoints given by the closed-loop regulator.

[0017] This matrix, called the control effectiveness matrix matrix" in English, is determined by the equations of the vehicle's physics. However, the constraints imposed on the algorithm can be chosen so that the instructions calculated by it respect predetermined criteria of performance, service or even safety.

[0018] The inventors have noticed that the command allocation algorithm as described in thesis [2] is not optimal insofar as it is likely to loop on a non-optimal intermediate value of the command vector. This blocking phenomenon can generate a high unnecessary computational load because the command allocation method remains blocked on the same command values ​​until reaching a maximum number of iterations allocated to it, without this computational load improving the quality of the commands allocated to the actuators. In addition, this can generate commands that are not optimal and which can therefore negatively impact the trajectory of the vehicle.

[0019] There is therefore a need to further improve existing order allocation methods.

[0020] The aim of the invention is to meet at least part of this need. Summary of the invention

[0021] To do this, the invention relates in one of its aspects to a method for allocating actuator commands of a motor vehicle, the method comprising: a / the selection of an initial vector u(Q) which satisfies the equality constraints Bu = v and the inequality constraints Cu > U where C — ( ) and U — ( _|), and the selection of an active set W containing a subset of the inequality constraints, li being a control vector of the vehicle actuators, I being the identity matrix and and M being vectors representing the respectively minimum and maximum predetermined constraints of the actuator controls, B being the control efficiency matrix and v being a setpoint representative of a desired behavior of the vehicle; b / the determination of a sequence of vectors {}, where each successive vector u(n + 1) is obtained by: b 1 / the determination of a disturbance vector p which minimizes the cost function || A( + ​​p) - b || and which verifies Bp — 0, the components of p which correspond to the components of u^n) whose inequality constraints are part of the active set W being zero s, A being a weighted control efficiency matrix obtained from B and b being a weighted setpoint obtained from v; b2 / the verification that u(n) + p satisfies the inequality constraints; b3 / if u(n) + p satisfies the inequality constraints, we define h(h + 1) = u(n) + p and we determine the Lagrange multipliers p, and X such that AT( Au -h) = ( BT Cq ) ( ^ ) °ù Co is a matrix that contains the columns of C corresponding to the inequality constraints present in the active set W, and we check that all the Lagrange multipliers X are positive or null, if all the Lagrange multipliers X are positive or null we allocate the control vector u( n + 1) to the actuators; if at least one Lagrange multiplier X is negative, we check that u(n+ 1) is identical to a backup vector u<'pt, if u(n + 1) is identical to we reset the active set W; if at least one Lagrange multiplier X is negative and if u ( u + 1 ) is different from uopt, we remove from the active set W the inequality constraint corresponding to the smallest Lagrange multiplier X and we assign to UaP{ the value u(n + 1); b3 / if u(n) + p does not satisfy the inequality constraints, we define U ( n + 1 ) = it ( il ) + ap where a is the largest number between 0 and 1 such that ll(n+ 1) satisfies the inequality constraints so that a component of the control vector u(n+ 1) reaches a bound of its inequality constraints, and we add to the active set W the inequality constraint corresponding to this component.

[0022] Preferably, A is equal to B and b is equal to v.

[0023] Alternatively, A is equal to WaB and b is equal to WaV, where Wa is a matrix positive definite representative of a weighting of the actuators.

[0024] According to a preferred embodiment, the setpoint v is representative of a yaw moment of the vehicle, the vehicle comprising at least one rear wheel steering actuator and, for each of the four wheels of the vehicle, a differential braking actuator, the control vector u comprising the controls of these actuators.

[0025] The invention also relates to a control system for a motor vehicle comprising a command allocation unit configured to allocate commands to the actuators of the vehicle from a command request representative of a desired yaw moment of the vehicle, the command allocation unit being configured to implement the method as described above.

[0026] The invention finally relates to a motor vehicle comprising a plurality of actuators and a control system as described above. Brief description of the drawings

[0027] [Fig-1] [Fig.l] is a block diagram of a vehicle control system.

[0028] [Fig.2] [Fig.2] illustrates a method of allocating orders according to the invention.

[0029] [Fig.3] [Fig.3] is a graph representing the result of a simulation comparing the evolution of the braking torque on the left front wheel obtained by a control allocation method according to the prior art with that obtained by a method according to the invention.

[0030] [Fig.4] [Fig.4] is a graph representing the result of a simulation comparing the evolution of the braking torque on the right front wheel obtained by a control allocation method according to the prior art with that obtained by a method according to the invention.

[0031] [Fig.5] [Fig.5] is a graph representing the result of a simulation comparing the evolution of the braking torque on the left rear wheel obtained by a control allocation method according to the prior art with that obtained by a method according to the invention.

[0032] [Fig.6] [Fig.6] is a graph representing the result of a simulation comparing the evolution of the braking torque on the right rear wheel obtained by a control allocation method according to the prior art with that obtained by a method according to the invention.

[0033] [Fig.7] [Fig.7] is a graph representing the result of a simulation comparing the evolution of the steering angle of the rear wheels obtained by a control allocation method according to the prior art with that obtained by a method according to the invention.

[0034] [Fig.8] [Fig.8] is a graph representing the result of a simulation comparing the evolution of the trajectory of the vehicle obtained by a method of allocating controls according to the prior art with that obtained by a method according to the invention. Detailed description

[0035] [Fig. 1] has been described in the preamble and will therefore not be commented on below. The method for allocating orders according to the invention can advantageously be implemented by the control system illustrated in [Fig. 1].

[0036] [Fig.2] illustrates a method of allocating orders according to the invention.

[0037] The objective of the command allocation method is to determine the command vector u which allows the optimal control of the actuators of the motor vehicle as a function of an instruction representative of a desired behavior and of the characteristics of the vehicle and the actuators. A component of the command vector u represents a command value of one of the actuators of the system which it is sought to control.

[0038] In a preferred application of the invention, the command allocation method is used to control the actuators of a motor vehicle so that the vehicle reproduces a yaw moment setpoint. The actuators implemented may in particular comprise a rear wheel steering actuator and an ac- differential braking donor for each of the four wheels of the vehicle. The vector u then comprises at least five components, each corresponding to a command value of one of these actuators.

[0039] The method begins at step 110, in which a starting vector if ( 0) is selected which satisfies the equality constraints Bu = v and the inequality constraints Cu>U where C = ( 5 ) ct U — ( ) • These equality and inequality constraints are predetermined. They are preferably updated at each new instance of the command allocation method and depend in particular on the physical characteristics of the actuators and the vehicle and the trajectory of the vehicle.

[0040] The control instruction v is a vector which corresponds to the desired behavior of the motor vehicle, for example its yaw moment or the total braking that one wishes to obtain.

[0041] B is the control effectiveness matrix of the vehicle. As is known from the prior art, the control effectiveness matrix represents the degree of effectiveness of each actuator in carrying out a certain control request, which depends on the characteristics of the vehicle and the actuators. The control effectiveness matrix is ​​a dynamic matrix that is recalculated continuously, or at least at each calculation time step of the software.

[0042] We also select an active set W which contains a subset of the inequality constraints Cu > U.

[0043] Preferably, if an instance of the order allocation method has been implemented previously, the starting point of the new instance of the method is taken to be the optimal vector u(n 4 1) and the active set W as determined at the end of the previous instance of the method.

[0044] From these initial conditions, a sequence of vectors is iteratively determined until an optimal control vector u is obtained. Each successive vector u^n+ 1) is obtained by the following steps, implemented at the n-th iteration of the method.

[0045] In step 120, the optimal perturbation vector p is determined which minimizes the cost function || A ( w(n) 4- p) - b || and which verifies Bp — 0.

[0046] p is a vector of the same dimension as u(n) ■ The components of p which correspond to the components of u(n} whose corresponding inequality constraints are part of the active set W are equal to 0. This amounts to considering the inequality constraints present in the active set as equality constraints.

[0047] For example, if the component of the control vector corresponding to the rear wheel steering actuator has reached a value which corresponds to

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] the maximum amplitude of rotation of the rear wheels due to the physical limitations of the actuator and if the corresponding maximum amplitude constraint is in the active set, then the value of the corresponding component of the vector p is 0. If this component of the control vector has not reached one of its inequality constraints, the corresponding component of the vector p is determined during the minimization of the cost function. Thus, if for example the control vector u has five components of which the first and third have reached their maximum value and the fifth its minimum value, the first, third and fifth components of the vector p will be zero and we will work on the second and fourth components of p. Thus, we only work advantageously on a subset of the actuator commands that have not reached one of their predetermined inequality constraints. This makes it possible in particular to limit the computational load. Preferably, the 2-norm || a || 2 = ara is used for minimizing the cost function. Any known minimization method can be used to determine the vector p. A is a weighted control efficiency matrix obtained from B and b is a weighted setpoint obtained from v. For example, A and b can be respectively defined by A = WaB and b = W„V where Wa is a positive definite weighting matrix that allows greater importance to be given to the commands of certain actuators relative to others. This allows in particular to favor one solution among others when several local optima are present. In the absence of weighting, we have A = B and b = v. A and b can also be defined by A = ( 'j and j. up is a preferred control vector, which is preferably equal to the zero vector. This corresponds to zero controls and minimal control energy. V is a weighting parameter, typically large in value, e.g., greater than or equal to 5, 10, or 100, to emphasize minimizing Bu - v over minimizing u " up. W is a control weighting matrix. Once the vector p is determined, we determine in step 130 whether u(n) + p satisfies the inequality constraints C(u(n) + p) > U- If so, we define in step 140 u(n + 1) = u(n ) + p and we determine the Lagrange multipliers q and X such that Ar( Aiin + 1)- b) = (Br Cj) ( , as for example described in thesis [2].

[0055] Co is a matrix that contains the columns of C corresponding to the inequality constraints present in the active set W.

[0056] It is then determined in step 150 whether all the Lagrange multipliers X are positive or negative.

[0057] If this is the case, the commands w(n + 1) are allocated to the actuators in step 160 and the command allocation method concludes. These commands it(n + 1) constitute the optimal commands of the actuators.

[0058] If at least one Lagrange multiplier X is negative, it is determined in step 170 whether u(n+ 1) is identical to the backup vector u°pt.

[0059] In practice, the backup vector ^pt records the value of u (n + 1) at the end of step 190 of the n-th iteration of the method in order to compare it to the value of u(n+ 1) at step 170 of the previous iteration of the method having ended in step 170.

[0060] If u(n + 1) is equal to u^pt, this means that the order allocation method is blocked in the situation where the same inequality constraint is removed at each iteration of the active set W and where the next iteration concludes with a calculation of this same constraint which gives the same value as in the previous iteration. This forms an infinite loop which prevents the method from determining the optimal orders while mobilizing a significant computational load. The active set W is then reinitialized at step 180 and the next iteration of the method is moved to by returning to step 120.

[0061] Advantageously, this allows the process to be taken out of the infinite loop. Thus, it is possible in particular to rework the controls of the actuators whose inequality constraint limits had been reached.

[0062] Preferably, the active set W is reset to a value different from the initial value selected in step 110.

[0063] If on the one hand at least one Lagrange multiplier X is negative and if on the other hand u(n + 1) is different from or ^p1 is not defined, the smallest Lagrange multiplier X is identified in step 190 and the inequality constraint corresponding to this multiplier is removed from the active set W. This makes it possible to rework on the least optimal actuator control. The value m(h + l) is then assigned to ^p1. The next iteration of the method is then moved to step 120.

[0064] If u ( n) + p does not satisfy the inequality constraints, we define in step 200 U ( n + 1 ) = u ( n ) 4- ap where a is the largest number between 0 and 1 such that 1) satisfies the inequality constraints, and we add to the active set W the inequality constraint corresponding to the component of the control vector u(n+ 1) having reached a bound of its inequality constraints. In other words, we determine the inequality constraint which constrains the value of a. This implies that the component of the control vector which corresponds to this inequality constraint has reached the bound of the latter. We then move on to the next iteration n+1 of the process by returning to step 120.

[0065] Figures 3 to 8 illustrate the results of a simulation comparing a method for allocating actuator commands of a motor vehicle according to the state of the art with the method according to the invention.

[0066] In this simulation, we seek to optimize the controls of the differential braking actuators of the wheels and the steering actuator of the rear wheels so as to reproduce a yaw moment setpoint.

[0067] The simulation implements a representative vehicle model allowing the use of a closed-loop control allocation method. This is the MADA model (advanced modeling of the dynamics of an automobile) developed by the Renault company.

[0068] The simulation studies the response of the vehicle on a circle of 15 m radius at constant longitudinal acceleration.

[0069] Figures 3 to 6 show the temporal evolution of the differential braking torque of each of the front left wheels bfll, bf!2, front right bfrl, bfr2, rear left brll, brr2, rear right brrl, brr2 respectively obtained by a method according to the prior art and by the method according to the invention.

[0070] [Fig.7] shows the temporal evolution of the steering angle of the rear wheels drl, dr2 respectively obtained by a method according to the prior art and by the method according to the invention.

[0071] [Fig.8] shows the evolution of the trajectory of the vehicle t1, t2 respectively obtained by a method according to the prior art and by the method according to the invention.

[0072] It is clear from these figures that the method according to the prior art and the method according to the invention apply braking in the same way to the front left and rear left wheels between t = 45 s and t = 55 s. However, after 55 s, the method according to the prior art remains stuck on an infinite loop: it constrains the braking to a zero value and maintains this constraint indefinitely, so that new braking control values ​​cannot be calculated.

[0073] In comparison, the method according to the invention detects this blocking and cancels the corresponding braking constraint. This makes it possible to request braking on the left wheels after 55 s, when the steering of the rear wheels has reached its minimum value, equal to -3.5° in this simulation.

[0074] As a result, trajectory tracking is significantly improved with the method according to the invention, as visible in [Fig.8].

[0075] Thus, this simulation illustrates the ability of the method according to the invention to improve the allocation and optimization of commands by avoiding untimely blocking of constraints on the actuator commands.

[0076] Other variants and improvements may be provided without departing from the scope of the invention. The method according to the invention may be applied to the allocation of actuator commands of a motor vehicle in general without being limited to the case of reproducing a yaw moment setpoint. In the context of reproducing a yaw moment setpoint, the method according to the invention may take into account other actuators in addition to the differential braking actuators and the rear wheel steering actuator, in particular one or more power steering and / or electric steering actuators and / or one or more single-wheel drive actuators. Although the simulation example is described with reference to the use of a single rear wheel steering actuator, the invention also applies to the case of a vehicle comprising two rear wheel steering actuators acting separately on each of the two rear wheels.List of cited documents.

[0077] [1] “Optimal Coordination of Chassis Systems for Vehicle Motion Control. Automatic Control Engineering », Kissai, M. (2019), doctoral thesis, Université Paris Saclay

[0078] [2] “Backstepping and control allocation with applications to flight control”, Harkegârd, O. (2003), doctoral thesis, Linköpings University

Claims

1. Claims Method for allocating actuator commands of a motor vehicle, the method comprising: a / the selection (110) of a starting vector w(0) which satisfies the equality constraints Bu = v and the inequality constraints Cu > U where C = ( ) and U = ( ), and the selection of an active set W containing a subset of the inequality constraints, u being a control vector of the vehicle actuators, I being the identity matrix and and w being vectors representing the respectively minimum and maximum predetermined constraints of the actuator controls, B being the control efficiency matrix and v being a setpoint representative of a desired behavior of the vehicle; b / the determination of a sequence of vectors {, where each successive vector u(n + 1) is obtained by: b 1 / the determination (120) of a disturbance vector p which minimizes the cost function || A( u(n) + p) - b || and which verifies Bp = 0, the components of p which correspond to the components of u(n) whose inequality constraints are part of the active set W being zero, A being a weighted control efficiency matrix obtained from B and b being a weighted instruction obtained from v; b2 / verification (130) that u(n) + p satisfies the inequality constraints; b3 / if uin) + p satisfies the inequality constraints, we define (140) U ( n + 1 ) = u ( n ) + p and we determine the Lagrange multipliers q and X such that A7 ( Au- b) — (BT Cq ) ( ) where Co is a matrix which contains the columns of C corresponding to the inequality constraints present in the active set W, and we check (150) that all the Lagrange multipliers X are positive or negative, if all the Lagrange multipliers X are positive or negative we allocate (160) the control vector u(n+ 1) to the actuators; if at least one Lagrange multiplier X is negative, we verify (170) that u (n + 1) is identical to a backup vector if u(n + 1) is identical to Ut>pf we reset (180) the active set W if at least one Lagrange multiplier / . is negative and if u(n + 1) is different from U(>p^ we remove (190) from the active set W the inequality constraint corresponding to the smallest Lagrange multiplier X and we assign the value u(n + 1); b3 / if u(n) + p does not satisfy the inequality constraints, we define (200) u(n+ 1) = u(n) + ap where a is the largest number between 0 and 1 such that u(n + 1) satisfies the inequality constraints so that a component of the control vector u(n+ 1) reaches a bound of its inequality constraints, and we add to the active set W the inequality constraint corresponding to this component.

2.

3. A method according to claim 1, A being equal to B and b being equal to v. A method according to claim 1, A being equal to WaB and b being equal to W^V, Wa being a positive definite matrix representative of a weighting of the actuators.

4. Method according to one of the preceding claims, the setpoint v being representative of a yaw moment of the vehicle, the vehicle comprising at least one rear wheel steering actuator and, for each of the four wheels of the vehicle, a differential braking actuator, the control vector u comprising the controls of these actuators.

5. A control system for a motor vehicle (10) comprising a command allocation unit (4) configured to allocate commands to the actuators of the vehicle from a command request representative of a desired yaw moment of the vehicle, the command allocation unit being configured to implement the method according to one of the preceding claims.

6. A motor vehicle (10) comprising a plurality of actuators and a control system according to the preceding claim.

Images