Method and device for determining the total mass of a vehicle

Abstract

The present invention relates to a method (100) for determining a total mass of a vehicle, comprising steps of: - receiving (102) a value representative of a torque applied to the wheels of the vehicle, - receiving (102) a value representative of a longitudinal acceleration of the vehicle, the determination method (100) being characterised in that it further comprises steps of: - introducing (110) the values representative of a torque and a longitudinal acceleration into a model for estimating a value representative of a longitudinal force applied to the vehicle, and - determining (112) the total mass of the vehicle on the basis of the value representative of the longitudinal force estimated by the estimation model and the value representative of the longitudinal acceleration.

Description

[0001]DESCRIPTION Title of the invention: Method and device for determining the total mass of a vehicle. The present invention relates to the field of the automotive industry, and more specifically concerns a method for determining the mass of a vehicle, carried out in the vehicle during its use. The total mass of a vehicle, comprising the sprung mass and the unsprung mass, is used by Advanced Driver Assistance Systems (ADAS) implemented in the vehicle. It should be noted here that the unsprung mass constitutes the mass of the vehicle components that precisely follow the undulations of the road on which the vehicle is traveling, i.e., in particular the wheels, rims, tires, and shock absorbers of the vehicle, while the sprung mass of the vehicle constitutes the mass of the vehicle components located above the shock absorbers.The unsprung mass is therefore almost constant throughout the vehicle's use, while the sprung mass depends on the vehicle's load and the number of occupants, and can therefore vary considerably. Currently, advanced driver assistance systems use a fixed mass stored in the vehicle's main computer, regardless of the load. The vehicle's chassis actuators, particularly the braking and stability control systems, therefore do not have the exact total mass of the vehicle. Consequently, the commands these systems apply to the chassis and wheels deviate from the intended braking or stability target, which can impact stability and passenger comfort.Accurate determination of the vehicle's total mass is therefore crucial for designing optimal, self-adaptive vehicle control while minimizing the energy used by this control system, thus enabling rapid convergence of commands towards the set objective. Furthermore, precise knowledge of the vehicle's total mass improves route planning, particularly for electric vehicles, whose range is significantly impacted by their mass. In fact, the European Commission mandates real-time determination of the vehicle's total mass for vehicle manufacturers. Existing solutions for determining a vehicle's mass during operation utilize measured or estimated data concerning the vehicle's longitudinal dynamics, including its longitudinal acceleration, longitudinal speed, and rolling resistance coefficients.However, the corresponding algorithms require knowledge of all the forces involved in this longitudinal dynamic. Most of these forces, particularly aerodynamic ones, are not directly measurable, necessitating the use of observers, Kalman filters, and / or derivations of equations formalizing this longitudinal dynamic. These derivations assume that the forces are constant and, consequently, that their derivatives are zero, which is not entirely accurate since resistive forces vary depending on the vehicle's speed, wind speed, air density, tire characteristics, and the road surface, among other factors. Therefore, these solutions do not provide the total mass of the vehicle with sufficient accuracy.The present invention aims to remedy at least in part the aforementioned drawbacks by providing a method and a device for determining the total mass of a vehicle, which notably eliminates the need to calculate the resistive forces applied to the vehicle.To this end, the invention proposes a method for determining the total mass of a vehicle, comprising the steps of: - receiving a representative value of a torque applied to the wheels of the vehicle, - receiving a representative value of a longitudinal acceleration of the vehicle, the method of determination being characterized in that it further comprises the steps of: - injecting the representative values ​​of a torque and a longitudinal acceleration into a model for estimating a representative value of a longitudinal force applied to the vehicle, and - determining the total mass of the vehicle as a function of the representative value of the longitudinal force estimated by the estimation model and the representative value of the longitudinal acceleration.It is understood that the estimation model provides a representative value for the longitudinal force output and that the mass is determined based on a ratio between the representative value of the longitudinal force and the representative value of the longitudinal acceleration. This estimation model does not require measurements of resistive forces; for example, it is based on a neural network, a regression model, or an identification system. Thanks to the invention, a precise total mass of the vehicle is thus obtained, which allows the braking and trajectory control systems to perform very well in terms of vehicle stability and passenger comfort. In one embodiment of the invention, the estimation model is derived from a supervised learning phase, implemented prior to the execution of the determination process.This supervised learning phase allows the system to learn the resistive forces involved in the vehicle's longitudinal dynamics and deduce them from the vehicle's traction or propulsion force. Once this learning phase is complete, the estimation model therefore does not need to perform any calculations or measurements to account for these resistive forces. Compared to the prior art, the invention thus requires significantly fewer computing resources within the vehicle. Furthermore, this supervised learning phase corrects uncertainties related to measurements or estimates of the torque applied to the vehicle's wheels and the vehicle's longitudinal acceleration. The estimation model is, for example, a neural network with at least the representative values ​​of the torque and the representative values ​​of the longitudinal acceleration as inputs, and the representative value of the longitudinal force as output.The neural network, for example, has four neural layers, which is sufficient to obtain good accuracy in the representative value of the longitudinal force. The estimation model can also include additional input parameters, such as the vehicle's lateral acceleration, wind speed, or the gradient of the road on which the vehicle is traveling. In this variant, the estimation model is refined but requires a longer supervised learning phase and means of determining values ​​corresponding to these additional parameters within the vehicle, such as sensors or observers. The determination step includes, for example, calculating a gross total mass equal to the ratio between the estimated longitudinal force and the longitudinal acceleration.According to an advantageous and optional feature of the invention, the determination step further includes a time-smoothing step for the calculated gross total mass. Since the determination of the vehicle's total mass is performed in real time, the representative values ​​of the torque applied to the vehicle's wheels and the longitudinal acceleration can be distorted by sudden changes in vehicle speed, resulting in fluctuations in the calculated gross total mass in real time. The smoothing step uses, for example, a Kalman filter or a recursive least squares method. It makes it possible to obtain an accurate total vehicle mass even during sudden changes in vehicle speed.According to another advantageous and optional feature of the invention, the method for determining the total mass of a vehicle according to the invention includes a step that inhibits the determination step when at least one of the following conditions is met: - the slope on which the vehicle is traveling is determined to be greater than a predetermined angle; - the ratio in absolute value between the longitudinal acceleration of the vehicle and its lateral acceleration is determined to be less than a predetermined threshold; and - the deceleration of the vehicle is determined to be greater than a predetermined braking threshold. This inhibition step ensures that certain prerequisites are met before determining the total mass of the vehicle, these prerequisites allowing the estimation model to be as accurate as possible.Indeed, the estimation model is designed under driving conditions that should closely match those of the vehicle when the process of determining its total mass is carried out. In particular, when the vehicle is traveling on a steep incline, the estimation model may provide the longitudinal force with less accuracy.The invention also relates to a device for determining the total mass of a vehicle, comprising: - means for receiving a value representative of a torque applied to the vehicle's wheels, - means for receiving a value representative of the vehicle's longitudinal acceleration, the determination device being characterized in that it further comprises: - a model for estimating a value representative of a longitudinal force applied to the vehicle, capable of receiving as input the representative values ​​of a torque and a longitudinal acceleration, and - means for determining the total mass of the vehicle as a function of the representative value of the longitudinal force estimated by the estimation model, and the representative value of the longitudinal acceleration. The receiving means are, for example, a communication port connected to a computer bus of the vehicle.The invention makes it possible to use the signals already available on this computer bus, without requiring additional external data. In this device for determining the total mass of a vehicle according to the invention, the estimation model is, for example, a trained neural network having as inputs at least the representative value of the torque and the representative value of the longitudinal acceleration, and as output the representative value of the longitudinal force. Furthermore, in this device for determining the total mass of a vehicle according to the invention, the determination means include, for example, means for calculating a gross total mass equal to the ratio between the estimated longitudinal force and the longitudinal acceleration, and means for time-smoothing the gross total mass resulting from the calculation means.The invention also relates to a vehicle comprising means for determining the torque applied to the vehicle's wheels, means for determining the vehicle's longitudinal acceleration, and a device for determining the vehicle's total mass according to the invention, capable of receiving data from the means for determining the vehicle's torque and longitudinal acceleration. These means for determining the vehicle's torque and longitudinal acceleration are, for example, sensors or estimators installed in the vehicle. The determination device according to the invention and the vehicle according to the invention offer advantages similar to those of the determination method according to the invention.Other features and advantages of the invention will become apparent from the following description, on the one hand, and from several illustrative and non-limiting examples given by reference to the accompanying schematic drawings, on the other hand, in which: [Fig. 1] represents a vehicle according to the invention, comprising sensors and / or estimators as well as a device for determining the vehicle's mass according to the invention, in one embodiment of the invention, and [Fig. 2] represents steps of a method for determining the mass of the vehicle of Figure 1 according to the invention, in one embodiment of the invention. According to one embodiment of the invention, a vehicle 2 represented in Figure 1 comprises a set 20 of sensors and / or estimators whose outputs are sent to a computer bus of the vehicle 2, also called a CAN bus (for "Controller Area Network").This assembly 20 may include, in particular, an estimator of the slope α, in degrees, of the road on which vehicle 2 is traveling, a sensor for the longitudinal speed θ of vehicle 2, a sensor for the lateral acceleration θ of vehicle 2, a sensor for the longitudinal acceleration θ of vehicle 2, and a sensor for the torque θ applied to the wheels of vehicle 2. Vehicle 2 also includes a device 1 according to the invention for determining the total mass M of vehicle 2, implementing a method 100 (shown in Figure 2) for determining the total mass M of vehicle 2 according to the invention, both in software and hardware. The determination device 1 is, for example, the main computer of vehicle 2.It includes in particular means for receiving 10 the values ​​provided by the set 20 of sensors and / or estimator on the CAN bus, at least one processor, a random access memory and a read-only memory in which is stored a computer program whose instructions, when executed on the processor, enable the implementation of the determination process 100. The read-only memory includes in particular an estimation model 12 of a representative value of a longitudinal force ^. ^ applied to vehicle 2, function of a representative value of a torque ^ ^ applied to the wheels of vehicle 2 and with a value representative of a longitudinal acceleration ^ ^ of vehicle 2. In this embodiment of the invention, the representative value of the torque ^^applied to the wheels of vehicle 2 is given by the torque sensor ^ ^and expresses, for example, this torque in Newton-meters. Furthermore, the representative value of a longitudinal acceleration ^ ^ The value of vehicle 2 is given by the longitudinal acceleration sensor ^ ^ of vehicle 2 and is, for example, expressed in meters per second squared. Estimation model 12 implements a four-layer neural network. In variants, the neural network has more or fewer than four layers. The neural network has two inputs, one of which is intended to receive the pair ^ ^ applied to the wheels of vehicle 2 and the other being intended to receive the longitudinal acceleration ^ ^ of vehicle 2. It provides as output a value representative of a longitudinal force ^ ^ applied to vehicle 2, which here is a longitudinal force ^ ^applied to vehicle 2, expressed in Newtons. The estimation model 12 is derived from a supervised learning phase carried out on a test bench or through a road test campaign under real-world conditions, on a test vehicle of the same type as vehicle 2. Once this learning phase is completed, the estimation model 12 is fixed and stored in the read-only memory of the determination device 1. The estimation model 12 is, of course, obtained before vehicle 2 leaves the factory and is valid for all vehicles of the same model and engine. During the supervised learning phase, the test vehicle is subjected to different driving conditions and loaded with varying load levels in order to cover a whole range of total mass values ​​for the test vehicle.The various driving conditions encompass a range of longitudinal acceleration values ​​for the test vehicle and a range of torque values ​​applied to the vehicle's wheels, which are successively applied as input to the neural network. Each different pair of these torque and longitudinal acceleration values ​​constitutes an iteration of the supervised learning phase, during which this pair of values ​​is applied as input to the neural network, along with a ground truth value for the resulting longitudinal force on the vehicle. The ground truth value is obtained by multiplying the applied longitudinal acceleration value by the total mass of the test vehicle during this iteration. This actual mass can be measured statically on a test bench or determined based on the vehicle's unladen weight and its load (passenger weight, cargo weight, fuel quantity, etc.).During each iteration, the longitudinal acceleration value and the torque value applied to the wheels of the test vehicle, which are input to the neural network, are taken from the vehicle's CAN bus. These values ​​are generally noisy measurements because they are instantaneous, and their accuracy depends on the location of the corresponding sensors in the vehicle. The acceleration value... ^The signal is optionally filtered to attenuate measurement noise during the calculation of the longitudinal force to be applied at the output of the neural network. It should be noted that through this supervised learning phase, the estimation model eliminates errors related to the estimation or measurement of the torque applied to the wheels of the test vehicle and the vehicle's longitudinal acceleration. This is because, even if the torque applied to the wheels and the longitudinal acceleration available on the CAN bus are not precise, the longitudinal force necessary to determine the total mass of the vehicle is obtained through learning. This is why, during the supervised learning phase, a precise measurement of the vehicle's total mass is used to calculate the longitudinal force applied at the output of the neural network. The supervised learning phase therefore serves to calibrate the torque and acceleration measurements from the CAN bus.Returning to vehicle 2 shown in Figure 1, the determination device 1 also includes, stored in its read-only memory, a module 14 for determining the total mass M of vehicle 2, capable of dividing the longitudinal force ^^ given at the output of the estimation model 12 by the longitudinal acceleration ^^ provided by the receiving means 10, and a module 16 for inhibiting the use of the estimation model 12 when suitable driving conditions are not met. The method 100 for determining the total mass M of vehicle 2 according to the invention, performed during the use of vehicle 2, is now described in relation to Figure 2. A first step 102 of the determination method 100 comprises, by the receiving means 10, receiving the value of the torque ^. ^ applied to the wheels of vehicle 2 as transmitted on the CAN bus, the reception of the longitudinal acceleration value ^ ^of vehicle 2 as transmitted on the CAN bus, and advantageously the reception of the slope α of the road on which vehicle 2 is traveling, of the longitudinal speed ^ ^ of vehicle 2 and lateral acceleration ^ ^of vehicle 2, as transmitted on the CAN bus. A second, optional step 104 of the determination method 100 is the verification, by the inhibition module 16, of the vehicle's driving conditions. In this second step 104, the inhibition module 16 verifies, in particular, that vehicle 2 is not in conditions less favorable to the implementation of the method according to the invention, such as a phase of heavy braking, significant lateral dynamics, or a steep incline, before allowing the determination of the total mass M of vehicle 2. Alternatively, additional conditions are verified before allowing the determination of the total mass M of vehicle 2, such as a small variation in engine torque or the non-use of an emergency braking device of vehicle 2.When driving conditions are favorable for the use of the determination of the total mass M of vehicle 2, the inhibition module 16 gives the value 1 to an indicator S of authorization of the determination of the total mass M of vehicle 2, otherwise the inhibition module 16 gives the value 0 to this indicator S. In other words, the indicator S remains at the value of zero as long as: - the slope α on which the vehicle 2 is traveling is greater in absolute value than a predetermined angle, for example fixed at 5 degrees: | α| < 5, and / or - a ratio in absolute value between the longitudinal acceleration ^^ of the vehicle 2 and a lateral acceleration ^. ^ of vehicle 2, is below a predetermined threshold, for example fixé - a deceleration of vehicle 2 is greater than a predetermined braking threshold, for example: |^ -^(^) − ^^(^ − 1)| > 0.015m.s1 during the deceleration phase, with t the instant of measurement of the longitudinal velocity ^ ^The inhibition module 16 then reads, in a third step 106 of the determination process 100, the value of the indicator S, and as long as the indicator S is zero (branch N), the inhibition module 16 inhibits, in a step 108, the determination of the total mass M of the vehicle 2. Otherwise, when the value of the indicator S is one (branch Y), the determination process 100 continues with a step 110 of injecting the torque value ^ ^ and longitudinal acceleration ^ ^ received during the first step 102, as input to the estimation model 12, which provides as output a value of a longitudinal force ^ ^ applied to vehicle 2. It should be noted that when the optional second step 104 and third step 106 are not implemented, the first step 102 is followed directly by step 110 of injecting the torque value ^ ^ and longitudinal acceleration ^ ^as input to the estimation model 12. The determination module 14 then determines, during a step 112 of determining the total mass M of the vehicle 2, a total gross mass m(t): m(t) = ^ ^ / ^ ^ Since this total gross mass m(t) fluctuates slightly over time t due to abrupt changes in vehicle speed, this determination step 112 may include a time-smoothing step 1120 of the total gross mass m(t) in order to obtain the total mass M of vehicle 2. For this, the recursive least squares method applied to the modèle :^^(^) = ^(^) ∗ ^^(^) (1) Starting from the initial values ​​^(^ = 0) and ^(^ = 0), with for example ^(^ = 0) equal to the mass of the vehicle unladen and ^(^ = 0) = 0.002, F(t) denoting an intermediate variable distinct from the longitudinal force ^^(^) as a function of time t, the application of recursive least squares allows us to obtain the estimate of the total mass M = ^(^) by following the recursive equations (2) and (3) below: ^(^ − 1) + ^(^ − 1)^^(^)^^^(^) − ^(^ − 1)^^(^)^ (2) The total mass M = ^(^) obtained is almost constant after convergence of this time filter, and is read into the main computer to be provided to the vehicle's driver assistance systems. As an example, for a vehicle with a total mass of 2000 kg, the invention allows an estimation of the vehicle's total mass with an accuracy of 5%, in less than ten seconds, and before the vehicle reaches 10 m / s. -1In other words, even at startup with a significant speed variation, the recursive least squares algorithm converges very quickly. Alternatively, the determination step 112 does not include a time-smoothing step 1120; for example, the determination step 112 is implemented only during phases where the vehicle's speed is quasi-static, so as to obtain a nearly constant gross total mass m(t), taken as the total mass M of the vehicle. Of course, the invention is not limited to the examples just described, and numerous modifications can be made to these examples without departing from the scope of the invention. In particular, the characteristics of the different embodiments can be combined to implement the invention, provided that these embodiments are not incompatible with each other.

Claims

CLAIMS 1- A method for determining (100) the total mass (M) of a vehicle (2), comprising the steps of: - receiving (102) a representative value of a torque (φ) applied to the wheels of the vehicle (2), - receiving (102) a representative value of a longitudinal acceleration (φ) of the vehicle (2), the method for determining (100) being characterized in that it further comprises the steps de : - injection (110) of representative values ​​of a torque (^^) and a longitudinal acceleration (^ ^ ) in an estimation model (12) of a representative value of a longitudinal force (^ ^ ) applied to vehicle (2), and determination (112) of the total mass (M) of vehicle (2) as a function of the representative value of the longitudinal force (^ ^ ) estimated by the estimation model (12) and the representative value of the longitudinal acceleration (^ ^2. A method for determining (100) the total mass (M) of a vehicle (2) according to claim 1, wherein the estimation model (12) is derived from a supervised learning phase, implemented prior to the execution of the determination method (100).

3. A method for determining (100) the total mass (M) of a vehicle (2) according to claim 2, wherein the estimation model (12) is a neural network having as inputs at least the representative value of the torque (^ ^ ) and the representative value of the longitudinal acceleration (^ ^ ), and for output the representative value of the longitudinal force (^ ^ ).4- Method for determining (100) the total mass (M) of a vehicle (2) according to any one of claims 1 to 3, wherein the determination step (112) comprises calculating a gross total mass equal to the ratio between the longitudinal force (^ ^ estimated ) and longitudinal acceleration (^ ^5. A method for determining (100) the total mass (M) of a vehicle (2) according to claim 4, wherein the determination step (112) comprises a time-smoothing step (1120) of the calculated gross total mass.

6. A method for determining (100) the total mass (M) of a vehicle (2) according to any one of claims 1 to 5, comprising an inhibition step (108) of the determination step (112) when at least one of the following conditions is met: - a slope on which the vehicle (2) is traveling is determined to be greater than a predetermined angle, - a ratio in absolute value between the longitudinal acceleration (θ) of the vehicle (2) and a lateral acceleration (θ) ^) of the vehicle (2), is determined to be less than a predetermined threshold, and - a deceleration of the vehicle (2) is determined to be greater than a predetermined braking threshold. 7- Device for determining (1) a total mass (M) of a vehicle (2), comprising: - means for receiving (10) a value representative of a torque (^^) applied to the wheels of the vehicle (2), - means for receiving (10) a value representative of a longitudinal acceleration (^ ^ ) of the vehicle (2), the determination device (1) being characterized in that it further comprises: - an estimation model (12) of a representative value of a longitudinal force (^^) applied to the vehicle (2), capable of receiving as input the representative values ​​of a torque (^ ^ ) and a longitudinal acceleration (^ ^ ), and means of determining (14) the total mass (M) of the vehicle (2) as a function of the representative value of the longitudinal force (^ ^) estimated by the estimation model (12), and the representative value of the longitudinal acceleration (^ ^ 8. Device for determining (1) the total mass (M) of a vehicle (2) according to claim 7, wherein the estimation model (12) is a neural network trained having as inputs at least the representative value of the pair (^ ^ ) and the representative value of the longitudinal acceleration (^ ^ ), and for output the representative value of the longitudinal force (^ ^ 9- Device for determining (1) the total mass (M) of a vehicle (2) according to claim 7 or 8, wherein the means for determining comprise means for calculating a gross total mass equal to the ratio between the longitudinal force (^ ^ estimated ) and longitudinal acceleration (^ ^), and means for time smoothing of the total gross mass resulting from the calculation means. 10- Vehicle (2) comprising means for determining (20) a torque (^^) applied to the wheels of the vehicle (2), means for determining (20) a longitudinal acceleration (^ ^ ) of the vehicle (2), and a device for determining (1) a total mass (M) of the vehicle (2) according to any one of claims 7 to 9, capable of receiving data from the means for determining the torque (20) ^ ) and longitudinal acceleration (^ ^ ) of the vehicle (2).

Images