Patent Application
20250115231
The invention relates to model-based predictive control of a motor vehicle. In particular, a method for model-based predictive control of a motor vehicle is claimed.
Current intelligent cruise controls (so-called “predictive green ACCs”) for motor vehicles are able to take route topology into account, but the driving strategy and thus the longitudinal control is rule-based. This rule-based implementation normally results in less than optimal solutions with regard to fuel consumption, comfort, and driving time. As the drive systems become more complex, these rules become more complex and require more computing power. An optimal operation of a vehicle (e.g. with regard to fuel consumption, comfort, and driving time) can only be achieved with familiarity with the route. A driver of the motor vehicle must therefore drive proactively, but only has limited knowledge of the route ahead, and no knowledge of upcoming driving obstacles.
The object of the present invention is to create a method for controlling a motor vehicle, that takes the problems described above into account. This problem is solved by the subject matter of the independent claims. Advantageous embodiments are the subject matter of the dependent claims, the following description, and the drawings.
The present invention uses the model predictive control (MPC) approach. There are three steps used in such model-based predictive control. In the first step, a virtual driving horizon (prediction horizon) is created from available maps and sensor data. The prediction horizon forms the solution space for a trajectory planner and controller in which the longitudinal trajectory of the motor vehicle, e.g. the speed or torque trajectory, is generated. In a second step, an iterative online generation and control of a longitudinal trajectory takes place, optimizing the trajectory with regard to performance goals in accordance with the MPC approach. In a third step, the trajectory is implemented automatically. The present invention contains a modification of the second step in this process, such that certain computing time requirements can be met for mass production. For this, the present invention creates an architecture with which the second step can be carried out and the computing times required for mass production can be met.
In particular, the present invention results in a functional architecture comprising two model-based predictive controllers (MPC solvers) that work sequentially, which are then supplemented by a downstream post processing unit in one embodiment. The two MPC solvers can be referred to as a high level solver (HLS) and a tracker. The high level solver module assumes the long-term planning of the longitudinal trajectory and uses the MPC approach for this. The rough long-term planning of the trajectory is based on the route. This allows for a correct, optimal dealing with inanimate horizon objects (inclines, speed limits, other traffic signs, e.g. “Stop” and “Yield,” and curves). The length of the travel horizon can range from 50 meters to 5,000 meters. Dynamic horizon objects can also be fundamentally taken into account, e.g. traffic lights. This is only roughly, due to the long computing times. The trajectory adjusted for dynamic objects may need to be overwritten by a quicker system. The high level solver therefore does not send a direct trajectory preference to the vehicle. Instead, the desired trajectory is further processed in the quicker tracker.
In this regard, a method for model-based predictive control of a motor vehicle is created. In one step, an MPC algorithm is carried out, which comprises a high level solver module, in which, when the high level solver module is applied to an upcoming route segment, a high level longitudinal trajectory is calculated, according to which the motor vehicle is to travel within a route-based high level prediction horizon. The high level solver module solves a non-linear problem in particular, and works with continuous compensation variables for discrete states (e.g. gears). This approach has less effect on the solution space than when discrete operating states are taken into consideration. This has advantages with regard to obtaining optimal results.
The high level longitudinal trajectory calculated by the high level solver module is sent to a tracker solver module in the MPC algorithm as an input value. The tracker solver module calculates a tracker longitudinal trajectory based on the high level longitudinal trajectory calculated by the high level solver module for a time period in the future, according to which the motor vehicle is to travel in a time-based tracker prediction horizon. The tracker prediction horizon is shorter than the high level prediction horizon, and therefore only covers a portion of the high level prediction horizon, e.g. an initial segment thereof. The tracker solver module is distinguished by quick computing times and reliability. When current solutions from the high level solver module are unavailable, the tracker solver module can supply torques, for example, based on the most recent solutions from the high level solver module.
Because the high level solver module assumes the rough long-term planning, it can be expected that this process requires significant computing time, regardless of the hardware. It is therefore advantageous to deal with dynamic horizon objects (other vehicles or road users) in the tracker solver module, with a higher computing frequency. To enable this, the tracker solver module works with a significantly shorter forecast than the high level solver module. Unlike the high level solver module, the tracker solver module is time-based. This results in a fundamental compatibility to the arbitration system.
The tracker solver module generates a desired trajectory, which can theoretically correspond to the desired trajectory from the high level solver module, but is time-based and high resolution. The longitudinal trajectory of the tracker solver module also differs from that of the high level solver module in that it has a higher cycling rate, which is enabled by the shorter forecast time. In this regard, the tracker longitudinal trajectory is calculated with a higher resolution than the high level longitudinal trajectory in one embodiment.
It is particularly advantageous if the length of the prediction horizon is between 50 meters and 5,000 meters. The speed trajectory of the vehicle can still be optimized online even if the forecast over the route is relatively long or far, e.g. with a forecast of a few kilometers. This requires a lot of computing. It is advantageous to calculate the trajectory for a relatively short forecast with the tracker solver module, e.g. for the region immediately in front of the motor vehicle. This allows for particularly short reaction times. In this regard, the length of the high level prediction horizon is set to between 50 meters and 5,000 meters in one embodiment. The tracker longitudinal trajectory can be calculated by the tracker solver module for a few seconds (or less) of the initial segment of the high level prediction horizon immediately in front of the motor vehicle.
In another embodiment, signal post processing (SPP) is used. Signal post processing involves subsequent processing of signals. In this case, partial results calculated in advance at different times are linked to one another such that they are in a temporal relationship to one another. This results in a time interval in which all of the information is available. The temporal basis is that of the tracker in particular. Moreover, the SPP and tracker are implemented in the same software module, such that they have the same temporal basis. In this regard, the tracker longitudinal trajectory calculated in the tracker solver module is sent to a post processing unit as an input in another embodiment, where it is processed to obtain a control signal. The motor vehicle can subsequently be controlled on the basis of this control signal. The post processing unit can be downstream of the model-based predictive control. The post processing unit converts the tracker longitudinal trajectory into desired torques and desired forces, in particular without using model-based predictive control.
The MPC algorithm comprises a longitudinal dynamics model and a high level cost function that is dedicated to the high level solver module, and the high level longitudinal trajectory is calculated taking the longitudinal dynamics model into account while minimizing the high level cost function. In a similar manner, the tracker solver module can have a dedicated tracker cost function, and the tracker longitudinal trajectory can be calculated taking the longitudinal dynamics model into account while minimizing the tracker cost function. The output from the tracker solver module can have a direct impact on driving comfort through the minimization of the tracker-specific cost function.
The high level solver module can deliver desired speed patterns, a charging state of the vehicle battery, drive forces, drive torques, braking forces, or braking torques as outputs. In this regard, the high level longitudinal trajectory comprises a speed trajectory in one embodiment, according to which the motor vehicle is to travel within the high level prediction horizon.
The high level longitudinal trajectory can also comprise the course of the charging state of a battery used to store energy for an electric machine in the motor vehicle with which the vehicle is powered. The charging state (or State of Charge: SoC) is the current amount of energy in the battery in relation to its maximum charge.
The high level and/or tracker longitudinal trajectories can also comprise a braking force trajectory for the brakes in the motor vehicle, according to which the brakes are to provide braking forces (weaker to none) within the high level prediction horizon, or within the tracker prediction horizon. Instead of the aforementioned drive forces and braking forces, drive torques and braking torques can be determined by the longitudinal trajectory. The torque trajectories relate to the torques for at least one wheel on the motor vehicle, and comprise both positive and negative torques that are to be provided by the electric machine, internal combustion engine, and brakes in the motor vehicle.
The tracker longitudinal trajectory can comprise a torque trajectory for at least one drive assembly (e.g. an electric machine) in the motor vehicle, according to which the at least one drive assembly is to provide drive torques within the tracker prediction horizon.
Exemplary embodiments of the invention shall be explained in greater detail below in reference to the schematic drawings, in which the same or similar elements have the same reference symbols. Therein:
The motor vehicle 1 also has a drive train 7, which can comprise an electric machine 8, which can be operated as a motor and a generator, a battery 9, a transmission 10, and brakes 19. The electric machine 8 can drive the wheels of the motor vehicle 1 via the transmission 10 when used as a motor. The electricity needed for this comes from the battery 9, in particular via power electronics 18. The battery 9 can also be charged by the electric machine 8 via the power electronics 18, when the electric machine is used as a generator (recuperation). The battery 9 can also be charged at an external charging station.
A computer program 11 can be stored in the memory 4. The computer program 11 can be executed on the processor 3, for which reason the processor 3 and memory 4 are connected to one another by the communication interface 5. When the computer program 11 is executed on the processor 3, the processor 3 fulfills the functions described in conjunction with the drawings, or executes the steps in the method.
The computer program 11 contains an MPC algorithm 13, which contains a high level solver module 13.1. The MPC algorithm 13 also contains a longitudinal dynamics model 14 for the motor vehicle 1. The high level solver module 13.1 can access the longitudinal dynamics model 14. The MPC algorithm 13 also contains a high level cost function 15.1 that is to be minimized, which is dedicated to the high level solver module 13.1. The high level solver module 13.1 is designed to propose an optimized operation of the motor vehicle 1, e.g. with regard to speed limits and stops, as well as traffic lights and inclines.
The longitudinal dynamics model 14 comprises a loss model 27 for the motor vehicle 1. The lost model 27 describes the operating behavior of components that are relevant with regard to efficiency, e.g. the electric machine 8, internal combustion engine 17, and brakes 19. Overall losses in the motor vehicle 1 can be derived therefrom. The processor 3 executes the MPC algorithm 13 and predicts the behavior of the motor vehicle over the course of a dynamic, route-based high level prediction horizon 24 (
The optimized high level longitudinal trajectory 31 is calculated for an upcoming route segment taking the longitudinal dynamics model 14 into account, in which the high level cost function 15.1 is minimized. The high level solver module 13.1 assumes the rough, long-term planning for the longitudinal trajectory 31 and uses the MPC approach for this. The rough, long-term planning of the high level longitudinal trajectory 31 is based on the route. This allows for a correct, optimal dealing with inanimate horizon objects (inclines, speed limits, and other traffic signs, e.g. “Stop” or “Yield” signs, curves, traffic lights). The high level longitudinal trajectory 31 comprises a speed trajectory 31.1 in this exemplary embodiment, according to which the motor vehicle 1 is to travel within the high level prediction horizon 24. Points along the route that the motor vehicle 1 passes are assigned optimized speeds for this. The high level longitudinal trajectory 31 can also comprise an optimized course 31.2 of a charging state for the battery 9. The high level longitudinal trajectory 31 can also comprise a braking force trajectory 31.3 for the brakes 19, according to which the brakes 19 are to provide braking forces within the high level prediction horizon 24.
In addition to the high level solver module 13.1, the MPC algorithm 13 also comprises a tracker solver module 13.2 with a dedicated tracker cost function 15.2. The tracker solver module 13.2 can access the longitudinal dynamics model 14. The high level longitudinal trajectory 31 calculated by the high level solver module 13.1 is also sent to the tracker solver module 13.2 as an input, e.g. by the processor 3, through the communication interface 5. The tracker solver module 13.2 is designed to propose an optimized operation of the motor vehicle 1, in particular with regard to maintaining distance to other objects, and preventing collisions.
The processor 3 executes the tracker solver module 13.2. The tracker solver module 13.2 contains instructions, or program code, with which the processor 3 calculates a tracker longitudinal trajectory 32 based on the high level longitudinal trajectory 31 calculated by the high level solver module 13.1 for a future time interval, according to which the motor vehicle 1 is to travel within a time-based tracker prediction horizon 30 (e.g. of 2 seconds). The tracker prediction horizon is significantly shorter than the high level prediction horizon. By way of example, the tracker prediction horizon 30 can cover those points along the route in the high level speed trajectory 31.1 that the motor vehicle 1 will pass in the next two seconds (at speeds in accordance with the speed trajectory 31.1). In this manner, the tracker prediction horizon 30 only covers an initial segment of the high level prediction horizon 24, as is illustrated by way of example in
The detection unit 6 can measure current state variables for the motor vehicle 1, record the corresponding data, and send this to the high level solver module 13.1, the tracker solver module 13.2, and to a post processing unit 16, which is described below. Furthermore, information regarding stationary objects and/or route data from an electronic map in a navigation system 20 for the motor vehicle 1 can be updated cyclically for a forecast horizon, or prediction horizon (e.g. 500 meters) in front of the motor vehicle 1, and sent to the modules 13.1, 13.2 and 16. The route data can contain information relating to elevation gains, curves, and speed limits, as well as traffic lights and stops. Furthermore, a maximum lateral acceleration can be calculated for the motor vehicle 1 when travelling through a curve. The detection unit 6 can also locate the motor vehicle 1, in particular from a signal generated by a GNSS sensor 12, in order to precisely locate the vehicle in the electronic map. The detection unit 6 can also contain an environment sensor 33 for detecting objects in the vehicle's 1 environment, e.g. a radar sensor, camera system, and/or lidar sensor. Dynamic objects within the environment of the vehicle 1 can also be detected therewith, e.g. other vehicles or pedestrians. The processor 3 can access information regarding these objects via the communication interface 5 for example. This information can be entered in the longitudinal model 14 of the motor vehicle 1, in particular as limits or constraints for the calculation of the high level longitudinal trajectory 31 and/or the tracker longitudinal trajectory 32.
The output of the optimization by the MPC algorithm 13 relates to optimal speeds 31.1 of the motor vehicle 1, and rotational torques 32.1 of the electric machine 8, or the internal combustion engine 17, or braking forces 32.1 of the brakes 19, and charging states 31.2 of the battery 9 for certain times and locations that are calculated within the prediction horizons 24, 30. The torque trajectory 32.2 and braking force trajectory 31.1 proposed by the tracker solver module 13.2 are sent to a post processing unit 16 in the present invention, which shall be explained in greater detail below in reference to
The post processing unit 16 is designed to calculate drive and braking forces, and recuperation limits. Furthermore, a requirement for stopping the motor vehicle 1 is also to be calculated, and an exit flag check is to be carried out. The torque trajectory 32.2 and braking force trajectory 31.1 proposed by the tracker solver module 13.2 are processed in the post processing unit 16 to obtain a control signal 34, which contains values with which the motor vehicle 1 can be controlled, e.g. by controlling the actuators in the motor vehicle 1. The control signal 34 can contain drive forces (stronger to none), braking forces (weaker to none), limit values for recuperation torques, or a stop signal (“flag vehicle stop”) for the motor vehicle 1. The data from the detection unit 6, the navigation system 20, and/or the high level longitudinal trajectory 31 can be used to calculate the control signal 34. Furthermore, the tracker solver module 13.2 and the post processing unit 16 can be queried simultaneously, such that both modules 13.2, 16 start their calculations at the same time.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2021/074380 | 9/3/2021 | WO |
