Methods for controlling a vehicle

Abstract

Method for controlling a vehicle (100), comprising: by one or more controllers (110, 206): Determine (202) an intended route of the vehicle (100); Determining (208) locations of virtual attraction nodes (A1T, A2T, A3T, A4T, A5T, A6T,414a-414f) on the intended route; Determining (208) locations of virtual repulsion nodes ( R 1 R , R 2 R , R 3 R , R 4 R , R 1 L , R 2 L , R 3 L , R 4 L , R 5 L , R 6 L ,410 a − 410 f ,412 a − 412 d ) along one or more boundaries that are set back from the intended route; Calculating (208, 210) virtual attractive forces (γ) A ) between one or more positions of the vehicle (100) and the virtual attraction nodes (A 1 T, A 2 T, A 3 T, A 4 T, A 5 T, A 6 T,414a-414f); Calculating (208, 210) virtual repulsive forces (γ) R ) between one or more positions of the vehicle (100) and the virtual repulsion nodes ( R 1 R , R 2 R , R 3 R , R 4 R , R 1 L , R 2 L , R 3 L , R 4 L , R 5 L , R 6 L ,410 a − 410 f ,412 a − 412 d ) ; Steering a trajectory of the vehicle (100) at least partially using the virtual gravitational forces (γ) A ) and the virtual repulsive forces (γ R ); Controlling (204) the trajectory of the vehicle (100) using a comfort mode and a precision mode, wherein both the comfort mode and the precision mode use control points, with the precision mode using a larger number of control points than the comfort mode; and Use (204) of precision mode in areas of an upcoming steering maneuver of the vehicle (100), an upcoming road junction, an upcoming road ramp, a construction zone or an area with relatively high traffic volume.

Description

Introduction

[0001] This description refers to the field of vehicle trajectory control.

[0002] A system for reliable trajectory control in autonomous or semi-autonomous vehicles can be advantageous.

[0003] DE 10 2016 205 442 A1 describes a method for optimizing the path planning of a vehicle, comprising a search algorithm that calculates a path to a given destination of the vehicle based on static information used as attractive or repulsive forces, wherein the search algorithm additionally integrates data on dynamic information as attractive forces into the planning.

[0004] DE 10 2021 212 648 A1 describes an autonomous parking route search system based on a cognitive sensor.The parking route search system comprises a cognitive sensor configured to obtain cognitive information for determining the presence and location of an obstacle by scanning the space around a vehicle being driven within a parking space; a global path search module configured to generate a global path for the interior of the parking space as a node map by searching for a free space in which the vehicle can be driven, using the cognitive information and by placing a node in that free space; and an optimal local path generation module configured to generate an optimal local path from the vehicle's current location to a destination by connecting the node placed in the free space as the vehicle moves within the parking space.

[0005] German patent DE 10 2013 207 572 A1 describes a vehicle behavior control device installed in a vehicle to control its steering. A virtual road contour, representing the outline of a drivable road area along which the vehicle will travel, is determined based on data from a position sensor. The virtual road contour is reconstructed as a sequence of road segments, such as straight segments and right- and left-curved segments. If the current road segment where the vehicle is located is right- or left-curved, the next road segment the vehicle will subsequently be on is provisionally set to a straight segment until the vehicle reaches it.The steering angle is controlled in such a way that the distance to a boundary of a virtual road is adjusted to the correct distance from the vehicle to a boundary of a virtual road located in front of the vehicle. Description

[0006] According to the invention, a method for controlling a vehicle by one or more controllers includes determining the intended path of the vehicle. The method further includes determining the positions of virtual attraction nodes on the intended path. Additionally, the method includes determining the positions of virtual repulsion nodes along one or more boundaries that are spaced apart from the intended path or located at obstacles near the intended path. The method further includes calculating virtual attraction forces between one or more positions of the vehicle and the virtual attraction nodes. Moreover, the method includes calculating virtual repulsion forces between one or more positions of the vehicle and the virtual repulsion nodes.The method additionally includes controlling the vehicle's trajectory, at least partially, using virtual attractive and repulsive forces. The method also includes controlling the vehicle's trajectory using a comfort mode and a precision mode, both of which utilize control points. The precision mode uses a greater number of control points than the comfort mode. Furthermore, the method includes using the precision mode in areas of an upcoming steering maneuver, an approaching intersection, an approaching road ramp, a construction zone, or an area with relatively high traffic volume.

[0007] The method can also involve assigning different weights to different virtual attraction nodes, where the different weights represent different virtual attractive forces generated by the different virtual attraction nodes. Furthermore, the method can involve assigning different weights to different virtual repulsion nodes, where the different weights represent different virtual repulsion forces generated by the different virtual repulsion nodes. At least some of the virtual repulsion nodes can be constructed as circles, with the radii of the circles representing the weights of at least some virtual repulsion nodes. At least some of the virtual repulsion nodes constructed as circles can be located at obstacles that the vehicle must avoid.

[0008] The procedure may further include using the virtual attractive forces and the virtual repulsive forces to determine costs, with the control of a vehicle trajectory further involving the use of the costs in a model predictive control that controls the vehicle's trajectory.

[0009] A second method for controlling a vehicle by one or more controllers comprises determining an intended path for the vehicle. The method further comprises determining the positions of first virtual repulsion nodes along a first boundary spaced from the intended path, and determining the positions of second virtual repulsion nodes along a second boundary spaced from the intended path. The method also comprises using at least some of the first virtual repulsion nodes to generate a first virtual repulsion curve and using at least some of the second virtual repulsion nodes to generate a second virtual repulsion curve.Furthermore, the method includes calculating first virtual repulsion forces between one or more positions of the vehicle and the first virtual repulsion curve, and calculating second virtual repulsion forces between one or more positions of the vehicle and the second virtual repulsion curve. In addition, the method includes steering the vehicle, at least partially, using the first virtual repulsion forces and the second virtual repulsion forces.

[0010] The second method may further include determining the positions of virtual attraction nodes located along, or generally along, the intended travel path, using at least some of the virtual attraction nodes to generate a virtual attraction curve, and generating virtual attractive forces between one or more positions of the vehicle and the virtual attraction curve. The method may also include steering the vehicle, at least partially, using the virtual attractive forces, the first virtual repulsive forces, and the second virtual repulsive forces. The virtual attraction and repulsive curves may be Bézier curves.

[0011] The second method for controlling a vehicle can include steering the vehicle using a comfort mode and a precision mode, where in precision mode the first virtual repulsion curve and the second virtual repulsion curve are closer together than in comfort mode. Precision mode can be used in areas where the vehicle is about to make a steering maneuver, such as an upcoming road intersection, an upcoming road ramp, a construction site, or in areas with relatively heavy traffic.

[0012] A vehicle comprises one or more steered wheels and one or more electronic controls suitable for controlling the steered wheels.The one or more controllers are programmed together with the following instructions: determining an intended path for the vehicle; determining the positions of virtual attraction nodes along the intended path; determining the positions of virtual repulsion nodes along one or more boundaries that are spaced from the intended path or located at obstacles near the intended path; calculating virtual attraction forces between one or more positions of the vehicle and the virtual attraction nodes; calculating virtual repulsion forces between one or more positions of the vehicle and the virtual repulsion nodes; and controlling the steered wheels, at least partially, using the virtual attraction forces and the virtual repulsion forces.

[0013] The vehicle can also include one or more controllers that are collectively programmed with an instruction to assign different weights to different virtual attraction nodes, where the different weights represent different virtual attractive forces generated by the different virtual attraction nodes. The vehicle can also include one or more controllers that are collectively programmed to assign different weights to different virtual repulsion nodes, where the different weights represent different virtual repulsion forces generated by the different virtual repulsion nodes.

[0014] In the vehicle, at least some of the virtual repulsion nodes can be arranged on circles, the radii of the circles representing the weights of the at least some of the virtual repulsion nodes. The vehicle can also include one or more controllers that are jointly programmed to use the virtual attractive forces and the virtual repulsive forces to determine costs, wherein the instruction to control the steered wheels further includes an instruction to use the costs in a model predictive controller that controls the steered wheels of the vehicle.

[0015] Asymmetric error constraints can be provided (e.g., a comfort trajectory on one side of an intended vehicle path and a precision trajectory on the other side). Path planning commands can be provided that, in addition to a nominal desired trajectory, include asymmetric error constraints, comfort and precision commands, and transition commands between the comfort and precision commands. The control system can follow a desired vehicle trajectory and adjust its behavior to accommodate the additional planner commands.

[0016] The above summary does not represent every embodiment or aspect of the present description. The features and advantages of the present disclosure mentioned above, as well as other possible features and advantages, will be readily apparent from the following detailed description of embodiments and best ways of carrying out the disclosure when considered in conjunction with the accompanying drawings and claims. Furthermore, this description expressly includes combinations and sub-combinations of the elements and features described above and below. Brief description of the drawings Fig. Figure 1 illustrates a vehicle with a steering system. Fig. Figure 2 shows part of the vehicle's steering system. Fig. Figure 3 illustrates a method for controlling the vehicle's trajectory. Fig. 3A shows in detail some boundary conditions that are in the Fig. The 3 control methods shown are used. Fig. Figure 4 illustrates the control of the vehicle trajectory when the vehicle passes through a fork or junction in the roadway. Fig. Figure 5 further illustrates the control of the vehicle trajectory when the vehicle passes through a fork or junction on the roadway. Fig. Figure 6 shows an example of a comfort mode, a transition mode, and a precision mode for controlling the vehicle's trajectory. Detailed description

[0017] The present description can be implemented in many different forms. Representative examples of the description are shown in the drawings and are described here in detail as non-restrictive examples of the disclosed principles. For this purpose, elements and restrictions described in the sections "Summary," "Introduction," "Description," and "Detailed Description," but not expressly set forth in the claims, should not be included in the claims, either individually or collectively, either by implication, by inference, or otherwise.

[0018] For the purposes of this description, the use of the singular includes the plural and vice versa, unless expressly excluded; the terms "and" and "or" apply in both the subjunctive and disjunctive moods; "every" and "all" mean "everyone and all"; and the words "including," "containing," "comprehensive," "with," and the like mean "including without limitation." Furthermore, words of approximation such as "about," "almost," "essentially," "generally," "approximately," etc., may be used here to mean "at, close to, or almost at" or "within 0-5% of" or "within acceptable manufacturing tolerances," or logical combinations thereof.

[0019] See first Fig. 1. A vehicle 100 is depicted. Vehicle 100 can be an autonomous or semi-autonomous vehicle in which steering and route planning are carried out at least partially without human intervention or with limited human intervention. Vehicle 100 can be any type of vehicle, such as a car, truck, van, SUV, or other vehicle. As shown in Fig. As shown in Figure 2, the vehicle 100 has a steering control system 102.

[0020] The steering control system 102 may contain one or more electronic control units (ECUs), such as ECU 110. ECU 110 may be a microprocessor-based control unit and should be understood as having the electronic resources (microcontroller, software, memory, inputs, outputs, circuits, and the like) to perform the functions of ECU 110 described in this description. The functions described in this description may also be distributed among one or more electronic control units in the vehicle 100, which may be interconnected by data buses and / or hardwiring and therefore share data and computing power.

[0021] Since the ECU 110 and other control units in the vehicle 100 may be microprocessor-controlled devices, they can operate based on instructions; such instructions may contain one or more programmed software commands. Furthermore, some or all of these instructions may contain additional instructions.

[0022] The inputs to the ECU 110 can include one or more steering angle sensors 111. The steering angle sensor(s) 111 can provide the angles of the steered wheels of the vehicle 100. The vehicle 100 can have one or more steered wheels. The vehicle 100 can have two steered wheels. The vehicle 100 can also have more than two steered wheels, e.g., four steered wheels.

[0023] The inputs to the ECU 110 can also include the output of a navigation system 114. The navigation system 114 can contain map data to know the configurations and locations of roads, including their boundaries, lane markings, and intersections. The navigation system 114 can also contain a desired destination or location for the vehicle 100 and one or more suggested routes for the vehicle 100 to reach the destination.

[0024] Furthermore, the inputs of the ECU 110 can include perception sensors such as one or more cameras 112, LIDAR (Light Detection and Ranging) remote sensing 116, radar 118, and GPS sensors 120. Such perception sensors can provide the ECU 110 with information about the vehicle's 100 environment. This environment can include the vehicle's 100 current position. The environment can include visual or otherwise perceptible cues of other vehicles or obstacles near the vehicle 100. The environment can also include lane markings, guardrails, road edges, intersections, obstacles, and other features of the roads on which the vehicle 100 can travel.

[0025] An output from ECU 110 can include one or more actuators 122 for controlling the steered wheels of vehicle 100. As such, ECU 100 is able to control the steered wheels of vehicle 100 to steer or control the trajectory of vehicle 100. Vehicle 100 can have one or more steered wheels. Vehicle 100 can have two steered wheels. Vehicle 100 can have four steered wheels.

[0026] Fig. Figure 3 additionally shows a block diagram of a system and a procedure for controlling the trajectory of a vehicle according to this exemplary description. Block 202 is a perception block. The position and surroundings of the vehicle 100 can be detected there, for example, by perception sensors (camera(s) 112, LIDAR remote sensing 116, radar 118, and GPS 120) and / or by the navigation system 114. The output of block 202 can be a nominal or intended path of the vehicle 100 to reach its destination. In block 204, reference trajectories for reaching the destination can be calculated. Several reference trajectories can be calculated, e.g., comfort, transition, and precision trajectories. As will be described in more detail later, a comfort trajectory can be used when precise control of the steering or trajectory of the vehicle 100 is not considered necessary.This could be, for example, along a generally straight stretch of road without nearby intersections, exit or entrance ramps, or obstacles. (When a comfort trajectory is used, this description may refer to the vehicle's control system being in comfort mode.) On the other hand, a precision trajectory can be used when precise control is more important. This might be the case, for example, in road sections with intersections, exit or entrance ramps, lane restrictions, or nearby obstacles. (When a precision trajectory is used, this description may refer to the vehicle's control system being in precision mode.)A transition trajectory can be used for situations where the road surface transitions between areas where a comfort trajectory and a precision trajectory are desirable. (When a transition trajectory is used, this description may refer to the vehicle 100's steering system as being in transition mode.) In comfort mode, steering can be performed with fewer control points, relaxed limits, and / or a smaller bandwidth than in precision mode to make the steering of the vehicle 100 less "harsh."

[0027] In block 214, a model of the “system” to be controlled, namely vehicle 100 and its steering system, is provided to the control unit 206.

[0028] A selected trajectory from block 204 can be forwarded to a model predictive controller 206. The model predictive controller 206 can be located in the ECU 110, in another controller of the vehicle 100, or in several controllers that can be interconnected. The model predictive controller 206 provides a control input to the actuators 122 so that the vehicle 100 follows the desired trajectory. The controller 206 tests various possible input sequences (in this case, steering inputs provided via the actuators 122) to find the one that causes the output to approximate the reference trajectory with the lowest cost V over a prediction horizon with 'p' prediction steps into the future. The model predictive controller 206 then commands the first element of the optimal input sequence (u) k) to control the vehicle. The model predictive control system 206 then performs the calculation again in the next time interval for the digital control system. The calculation of the costs V is described in more detail below.

[0029] The model predictive control system 206 also utilizes other inputs, the generation of which will now be discussed. First, after the planning target and reference trajectories have been provided in block 204, obstacle and road geometry as well as directional error constraints can be provided from block 204 to block 208. In block 208, planning nodes and Bézier curves can be applied by the ECU 100.

[0030] It will now be on Fig. 4. Referenced. The planning nodes used in Block 208 can be either virtual attracting or virtual repelling nodes. In Fig. 4. It is assumed that the vehicle 100 changes from a straight lane 300 via a fork 301 to another lane 302. Lane 300 may have a first boundary, e.g., the left edge 300a, and a second boundary, e.g., the right edge 300b. Lane 302 may have a first boundary, e.g., the left edge 302a, and a second boundary, e.g., the right edge 302b. One or more virtual attraction nodes may be placed along the path (for example, along the center 300c of lane 300 and turning into the center 302c of lane 302) on which the vehicle 100 is to travel. Such virtual attraction nodes may include: virtual attraction node A1T virtual attraction node A2T virtual

[0031] Attraction node A3T virtual attraction node A4T virtual attraction node A5T and virtual attraction node A6T One purpose of the virtual attraction nodes is to create virtual “fields” that can be used in calculating the cost function V, which is used in model predictive control 206 by Fig. As shown in Figure 3, these virtual attraction nodes effectively exert an "attractive" force on vehicle 100. This attractive force helps ensure that vehicle 100 follows its intended path. The virtual attraction nodes can be located on the intended path of vehicle 100 or anywhere along it.

[0032] The planning nodes used in Block 208 can also be virtual repulsion nodes. Such virtual repulsion nodes can include: virtual repulsion node R1R, virtual repulsion node R2R, virtual repulsion node R3R and virtual repulsion node R4R which are each located at a certain distance from the intended path of vehicle 100. In this case, they are located to the right of the intended path and can be arranged along the right edges of the respective carriageways 300 and 302. Alternatively, the virtual repulsion nodes can also be arranged along a carriageway boundary, e.g., a lane marking between two carriageways.

[0033] Additional virtual repulsion nodes can also be provided, such as virtual repulsion nodes. R1L virtual repulsion node R2L virtual repulsion node R5L and virtual repulsion node R6L which are each located at a certain distance from the intended path of vehicle 100. In this case, they are located to the left of the intended path and can be arranged along the left edges of the respective carriageways 300 and 302. Alternatively, the virtual repulsion nodes can also be arranged along the edge of a lane boundary on the respective carriageway.

[0034] The ECU 110 can also draw Bézier curves defined by the respective virtual attraction and repulsion nodes. For example, Bézier curve 312, which generally runs along the left edge 300a of lane 300 and merges into the left edge 302a of lane 302, can be drawn by the Fig. 4 given parameter equation (1) for B Lcan be described. Furthermore, the Bezier curve 310, which generally runs along the right edge 300b of carriageway 300 and merges into the right edge 302b of carriageway 302, can be described by the in Fig. 4 given parameter equation (2) for B R can be described. It is therefore evident that the Bezier curve 310 and the Bezier curve 312 can be constructed from virtual repulsion nodes.

[0035] The Bezier curve 314 can also be drawn by the ECU 110. The Bezier curve 314 can generally run along the center 300c of lane 300 (which may lie on or generally on the target trajectory for vehicle 100) and merge into the center 302c of lane 302 (which may also lie on or generally on the target trajectory for vehicle 100 if vehicle 100 passes through the fork 301 between lane 300 and lane 302). The Bezier curve 314 can be determined by the parametric equation (3) for B T in Fig. 4 will be described.

[0036] The perception sensors on vehicle 100 can be found in block 202 ( Fig. 3) also detect that an obstacle or a potential obstacle (here the second vehicle 340) is located in front of vehicle 100. ECU 110 can create a virtual repulsion node. R3L Place it on or near the second vehicle 340, specifically in this example near the right rear corner of the second vehicle 340. The virtual repulsion node R3L can be constructed as circle 342. The virtual repulsion node R3L The circle 342 can be oriented towards the vehicle at 100 degrees. The circle 342 can have a radius that represents the strength of the repelling node, which can reflect the importance of the potential obstacle in question. That is, if an obstacle or potential obstacle is considered more significant, the virtual repulsion node can be 100 degrees. R3L The obstacle can be placed further away (in this case on a larger circle 342) from the obstacle to help vehicle 100 maintain a greater distance from it. The radius of the circle can also reflect the uncertainty about the obstacle's position; if the obstacle is a moving vehicle, a larger radius can be assigned due to the uncertainty about its current position. A smaller radius can be used if the obstacle is fixed, as the obstacle's position is more certain.

[0037] A virtual repulsion node R4L It can be placed at vertex 344 of the intersection between lanes 300 and 302. Here too, vertex 344 can be a particularly important point that vehicle 100 wants to avoid. The virtual repulsion node R4L can be constructed as a circle 346 with a radius representing the strength of the repulsion node, which may be proportional to the importance of the location that vehicle 100 wishes to avoid, or otherwise directly related to it. The virtual repulsion node R4L can be oriented towards the vehicle 100 or generally towards it.

[0038] In addition, it will be noted Fig. Reference is made to Figure 5. There, a virtual repulsion node 410a, a virtual repulsion node 410b, a virtual repulsion node 410c, a virtual repulsion node 410d, a virtual repulsion node 410e, and a virtual repulsion node 410f are shown. (These virtual repulsion nodes can be collectively referred to as virtual repulsion nodes 410a-410f). Also shown are the virtual repulsion nodes 412a, 412b, 412c, and 412d. (These virtual repulsion nodes can be collectively referred to as virtual repulsion nodes 412a-412d). Furthermore, the virtual attraction node 414a, the virtual attraction node 414b, the virtual attraction node 414c, the virtual attraction node 414d, the virtual attraction node 414e and the virtual attraction node 414f are shown.(These virtual attraction nodes can be collectively referred to as virtual attraction nodes 414a-414f). The virtually attractive and virtually repulsive nodes can be weighted differently, depending at least in part on the importance of their positions along the planned path of vehicle 100. For example, virtual attraction node 414c is located immediately at the intersection of lane 300 and lane 302, which can be a particularly important point for guiding and controlling vehicle 100. Virtual attraction node 414c is shown with a larger radius in the figure, indicating that it has a greater relative weight when calculating the virtual attraction force exerted on vehicle 100. Virtual attraction node 410b may have a greater weight because it is located at the beginning of the intersection between lane 300 and lane 302.The virtual repulsion node 410d can also have a greater weight, as it is located at the vertex 344; if an appropriate distance to the vertex 344 is maintained, crossing the roadway can be avoided when the vehicle travels 100 from roadway 300 to roadway 302.

[0039] The virtual repulsion nodes 410a-410f and the Bézier curve 411 they form can generate virtual repulsion fields, as shown by the arrows emanating from the virtual repulsion nodes 410a-410f and the Bézier curve 411 they form. Virtual repulsion nodes 412a-412d and the Bézier curve 413 they form can also generate virtual repulsion fields, as illustrated by the arrows emanating from the virtual repulsion nodes 412a-412d and the Bézier curve 413 they form. On the other hand, virtual attraction nodes 414a-414f and the Bezier curve 415 formed by them can generate virtual attraction fields, as shown by the arrows pointing towards the virtual attraction nodes 414a-414f and the Bezier curve 415 formed by them.

[0040] The output of Block 208 ( Fig. 3) can be a net “field” (the “net potential field”) representing the attractive or repulsive “potential” of the attracting and repulsive nodes and the attractive and repulsive Bézier curves. The potential is a measure of the virtual force generated by the attracting and repulsive nodes on the vehicle 100. The weighted “cost” with respect to this potential (the “potential field cost”) is calculated in Block 210 for use in model predictive control 206.

[0041] The model predictive control 206 can be a discrete digital control that operates with a control interval or period "k". The control has a curve 250 that represents the reference trajectory for the vehicle 100. The model predictive control 206 calculates a cost function V as a summation of a number of cost terms, such as: V=12∑k=1p{(y−yref)2Wy+(u−uref)2Wu+(uk−uk−1)2WΔu+ε2Wε+γ2Wγ}, where The variable(s) are based on the movement of vehicle 100 (i.e., the "outputs"); y ref Reference states are / are that relate to the target path of vehicle 100; u the inputs for the system for controlling the vehicle trajectory is / are; u ref the pre-calculated input setpoint(s) for model predictive control is / are, i.e., the input that would lead to outputs that follow the setpoints based on forward calculations; ε is a step function that indicates whether "soft" constraints would be violated by a particular control action; γ is the virtual net strength of the virtual attraction nodes, the virtual repulsion nodes, and the Bezier curves; W y the weighting assigned to the outputs; W u the weight assigned to the inputs; W Δuthe weighting that is assigned to the rate at which the inputs change; W ε the weight that is assigned to the violation of “soft” restrictions; Wγ is the weight assigned to the potential field of the virtual attraction nodes, the virtual repulsion nodes, and the Bezier curves; and p is the prediction horizon.

[0042] In the equation for the cost function V, y and y can be ref The matrices are 4 × 1 matrices, where the variables in the matrices are the lateral position, lateral velocity, rotational position, and rotational rate of the vehicle 100. u can be a scalar, namely the steering angle input command that is fed to the actuator(s) 122.

[0043] The term (u k - u) k-1 2This reflects the rate at which the steering angle changes. This term is included in the cost function V to illustrate that fewer "jerks" in the steering of vehicle 100 are desirable.

[0044] “Soft” constraints can be restrictions whose violation may be detrimental, such as approaching a predetermined distance to a road edge or boundary without actually crossing it. “Soft” constraints are conditions that should preferably, but not necessarily, not be violated and are therefore included in the calculation of the cost function V. If no control solution can be found that does not violate a “hard” constraint, the system can revert the steering from autonomous to manual. Another example of the relationship between a soft and a hard constraint is that a hard constraint states that the vehicle must not approach a specific obstacle to a predetermined distance (e.g.,five feet); a related soft requirement may be that the vehicle may not approach the obstacle to within a larger specified distance (e.g. ten feet).

[0045] The various W terms in the equation above are weighting factors, which can be constants. They can be assigned based on the system calibration to achieve the desired system performance.

[0046] In model predictive control 206, curve 250 shows the reference trajectory of vehicle 100. Curve 252 is the actual output, reflecting the trajectory of vehicle 100 up to time "k", and curve 254 is the predicted output, reflecting the predicted trajectory of vehicle 100. (Arrow 251 and arrow 253 denote the time before and after time "k", respectively.) Curve 256 is the past control input, and curve 258 is the predicted control input that causes vehicle 100 to converge to the reference trajectory. After calculating the minimum cost V over the forecast horizon and the inputs (curve 258) that would result in the minimum cost according to the prediction, model predictive control 206 applies the first input u. k+1The model predictive control 206 then recalculates the minimum cost V over an incremental forecast horizon and reapplies the first input from this calculation, and the process continues.

[0047] Model predictive control operates with constraints. The boundary conditions 212, within which model predictive control 206 can operate, are defined in Fig. 3A is shown in detail. There, stands n u for the length of the tax horizon; stands n p for the length of the forecast horizon; place u min and u max represents the minimum or maximum input limit; stand y min and y max for the minimum and maximum performance requirements; ε represents the step function mentioned above with respect to soft constraints; and A represents min and A max represent programmable parameters.

[0048] The calculation of the total field strength based on virtually attracting Bézier curves (i.e., Bézier curves constructed based on virtually attracting nodes) and virtually repulsive Bézier curves (i.e., Bézier curves constructed based on virtually repulsive nodes) can be performed as follows. A virtual attractive force γ A can be calculated as γ A = y BT - y where y BTThe position of the virtual attracting Bézier curve is and y is the position of vehicle 100 in the xy-coordinate set (where y is a lateral direction relative to vehicle 100 and x is a longitudinal direction defined by vehicle 100). That is, the virtual attractive force is greater when vehicle 100 is farther from the virtual attracting Bézier curve. (The force can be scaled based on the strength of the respective attraction nodes that form the basis of the Bézier curve.) A virtual repulsive force γ R On the other hand, this can be: γR={1yBL−y,y<01yBR−y,y≥0

[0049] This means the virtual repulsive force can be greater when the vehicle is 100 units closer to a virtual repulsive Bézier curve. The force can be scaled based on the strength of the respective repulsive nodes that form the base of the Bézier curve.

[0050] Thus, the total costs based on the potential fields for any longitudinal position of the vehicle 100 can be as follows: γ=γAγR={yBT−yyBL−y,y<0yBT−yyBR−y,y≥0

[0051] The costs y to be assigned to the virtual attraction and repulsion nodes can be calculated similarly according to the lateral distance y of the vehicle 100 from the virtual attraction and repulsion nodes. (The virtual attraction and repulsion forces can be scaled based on the strength of the respective virtual attraction and repulsion nodes.) These costs can be added in a similar manner. All costs y can then be summed to determine the total costs to be assigned to the virtual attraction and repulsion nodes and the virtual attraction and repulsion Bézier curves used in the calculation of V by the model predictive control 206.

[0052] It will now be on Fig.Reference is made to Figure 6. There, the use of a comfort mode, a transition mode, and a precision mode is illustrated. In this example, vehicle 100 may be on lane 500 and intend to remain on lane 500 instead of turning onto lane 502 via a junction 501. While vehicle 100 is traveling on a section 504 of lane 500 before junction 501, it can be controlled in comfort mode. The positions of the virtual repulsion nodes and the resulting Bézier curves 506 and 508 can be relatively far apart, as precision control and its associated relatively harsh driving style may not be advantageous on a section of track without intersections or obstacles. However, as vehicle 100 approaches junction 501, it may become more important for it not to veer onto lane 502.Therefore, a transition mode can be activated for road segment 509 in which the Bézier curve 510 on the right (i.e., towards the fork) is closer to the intended path of vehicle 100 than the Bézier curve 512 on the left. Consequently, the controller (i.e., the track planner) can prescribe asymmetric error constraints or asymmetric error limits. In this case, the controller prescribes a comfort trajectory on the left and a precision trajectory on the right. Furthermore, in precision mode, the Bézier curves 516 and 518 can be relatively close together (i.e., closer to each other) in track segment 514. Finally, the behavior can revert to the standard comfort mode.At this point, Bezier curve 520 and Bezier curve 522 can again be relatively farther apart and enter comfort mode, since the vehicle 100 is no longer on a section of road where particularly precise control could be advantageous.

[0053] Additionally or alternatively, a precision mode can be implemented by increasing the number of control points to which the vehicle's trajectory can be regulated. In the case of model predictive control, this can be achieved by increasing the bandwidth of the relevant digital control system. Increased precision in driving maneuvers or in areas that would benefit from the additional precision may come at the cost of greater harshness resulting from the more precise control. A comfort mode can be implemented by using fewer control points, and a transition mode by a number of control points between the precision and comfort modes.

[0054] A precision mode or precision trajectory can be used, for example, in areas where an upcoming steering maneuver of the vehicle, an upcoming road intersection, an upcoming road ramp, nearby obstacles (such as other vehicles) or a construction zone may make precise control of the vehicle's trajectory advantageous.

Claims

[1] Method for controlling a vehicle (100), comprising: by one or more controllers (110, 206): Determine (202) an intended route of the vehicle (100); Determining (208) locations of virtual attraction nodes (A1T,A2T,A3T,A4T,A5T,A6T,414a-414f) on the intended route; Determining (208) locations of virtual repulsion nodes (R1R,R2R,R3R,R4R,R1L,R2L,R3L,R4L,R5L,R6L,410a−410f,412a−412d) along one or more boundaries that are set back from the intended route; Calculating (208, 210) virtual attractive forces (γ) A ) between one or more positions of the vehicle (100) and the virtual attraction nodes (A1T,A2T,A3T,A4T,A5T,A6T,414a-414f); Calculating (208, 210) virtual repulsive forces (γ) R) between one or more positions of the vehicle (100) and the virtual repulsion nodes (R1R,R2R,R3R,R4R,R1L,R2L,R3L,R4L,R5L,R6L,410a−410f,412a−412d); Steering a trajectory of the vehicle (100) at least partially using the virtual gravitational forces (γ) A ) and the virtual repulsive forces (γ R ); Controlling (204) the trajectory of the vehicle (100) using a comfort mode and a precision mode, wherein both the comfort mode and the precision mode use control points, with the precision mode using a larger number of control points than the comfort mode; and Use (204) of precision mode in areas of an upcoming steering maneuver of the vehicle (100), an upcoming road junction, an upcoming road ramp, a construction zone or an area with relatively high traffic volume. [2] The method of claim 1, further comprising assigning (208) different weights to different virtual attraction nodes (A1T,A2T,A3T,A4T,A5T,A6T,414a−414f), where the different weights result in different virtual attractive forces (γ) A ) represent the various virtual attraction nodes (A1T,A2T,A3T,A4T,A5T,A6T,414a−414f) be generated. [3] Method according to claim 2, further comprising assigning (208) different weights to different virtual repulsion nodes (R1R,R2R,R3R,R4R,R1L,R2L,R3L,R4L,R5L,R6L,410a−410f,412a−412d) where the different weights result in different virtual repulsive forces (γ) R ) represent the various virtual repulsion nodes (R1R,R2R,R3R,R4R,R1L,R2L,R3L,R4L,R5L,R6L,410a−410f,412a−412d) be generated. [4] Method according to claim 3, wherein at least some of the virtual repulsion nodes (R1R,R2R,R3R,R4R,R1L,R2L,R3L,R4L,R5L,R6L,410a−410f,412a−412d) are constructed as circles (342, 346), where the radii of the circles are the weights of at least some virtual repulsion nodes (R1R,R2R,R3R,R4R,R1L,R2L,R3L,R4L,R5L,R6L,410a−410f,412a−412d) represent. [5] Method according to claim 4, wherein at least some of the virtual repulsion nodes (R1R,R2R,R3R,R4R,R1L,R2L,R3L,R4L,R5L,R6L,410a−410f,412a−412d), which are constructed as circles (342, 346), are located at obstacles (340) which the vehicle (100) is to avoid. [6] The method of claim 1, further comprising: Using (210) the virtual attractive forces (γ A ) and the virtual repulsive forces (γ R ) to determine costs (V, γ); and wherein controlling a trajectory of the vehicle (100) further includes using the costs (V, γ) in a model predictive control (206) that controls the trajectory of the vehicle (100).

Images