Method for planning a target trajectory for an autonomous vehicle

Abstract

The present invention relates to a method for planning a target trajectory for an autonomous vehicle (1), wherein for each object (2, 3) detected during a planning period, a quality value (Q) representing a measure of the reliability of the detection of the object (2, 3) is determined. For each object (2, 3), a minimum allowable acceleration (aQ) that is allowed when the vehicle (1) brakes towards the object (2, 3) is determined depending on the quality value (Q) of the object. A trajectory candidate and an object cost determined for the trajectory candidate are evaluated depending on the acceleration (ax(t)) set by each trajectory candidate and depending on the minimum allowable acceleration (aQ) when braking towards the object (2, 3). The selection of the target trajectory is made from a set of trajectory candidates depending on the trajectory cost, and additionally taking into account the evaluation of the trajectory candidates and the object costs determined for those trajectory candidates.

Description

[Technical Field]

[0001] The present invention relates to a method for planning a target trajectory of an automated vehicle according to the preamble of claim 1. [Background technology]

[0002] German Patent Application Publication No. 102020108857 discloses a method for planning a target trajectory to be automatically followed by a vehicle. The planning is based on determining a set of distinct candidates for the target trajectory and selecting one candidate from the set of determined candidates. The selection is based on a predetermined cost function. If changes in the boundary conditions to be observed and the driving task to be performed are identified, the cost functions for the individual trajectory sections of the candidates are adapted to the changed boundary conditions and driving task, thereby providing a pre-control of the selection, so that trajectory sections that are more suitable for observing the changed boundary conditions and performing the changed driving task are assigned a lower cost than other trajectory sections. The target trajectory includes, as a data set, both information on the position path to be followed by the vehicle when traveling along the target trajectory and information on the acceleration and driving speed of the vehicle when traveling along the target trajectory. Furthermore, a set of trajectories including the selected target trajectory is discretized (multiple trajectories including the selected target trajectory are separated), whereby a predetermined set of time-ordered trajectory support points is determined in the forward region, and a set of trajectories that pass through different trajectory support points in a time-order manner is determined. The set of trajectories that pass through the trajectory support points in a time-order manner forms a trajectory group to be considered as candidates for selecting the target trajectory. A cost is determined for each trajectory in the trajectory group using a predetermined cost function, whereby a total cost of the trajectory segment is determined using a weighted sum of the costs of the trajectory segments determined for various boundary conditions. The cost of a certain trajectory is determined using the sum of the total costs of that trajectory segment. Subsequently, the trajectory with the lowest cost is selected from the trajectory group as the target trajectory.

[0003] From DE 10 2020200183 A1, a method for controlling a vehicle is known, in which a probabilistic free space map is created around the vehicle, including static and dynamic objects, and the trajectory of the vehicle is planned taking into account the probabilistic free space map and optimized using a cost function.

[0004] From DE 10 2015 016 544 A1, a method is known for determining an avoidance trajectory of a vehicle for driving around an obstacle, in which the avoidance trajectory is optimized with respect to predetermined criteria, which include an upper limit for the acceleration of the vehicle, a lower limit for the distance to the obstacle, and a transverse (lateral) speed at the end of the avoidance trajectory.

[0005] German Patent Application Publication No. DE 10 2016 218 121 A1 discloses a method for planning a trajectory, in which a movement model of the vehicle, collision-related objects, and physical limitations of the vehicle are used to plan the trajectory, whereby a number of possible trajectories are evaluated using a cost function, and the trajectory with the smallest cost is then selected. Summary of the Invention [Problem to be solved by the invention]

[0006] It is an object of the present invention to provide a novel method for planning a target trajectory for an automated vehicle. [Means for solving the problem]

[0007] This object is achieved according to the invention by a method having the features of claim 1.

[0008] Advantageous embodiments of the invention are the subject matter of the dependent claims.

[0009] In a method for planning a target trajectory for an automated vehicle, particularly a highly automated or autonomous vehicle, a set of candidate trajectories is established for a predetermined planning horizon, with each candidate trajectory defining a path the vehicle should follow if that candidate trajectory is selected as the target trajectory and defining an acceleration at which the vehicle should follow that path. Objects are detected within the planning horizon, and a trajectory cost is assigned to each candidate trajectory using a predetermined cost function, where the cost function includes an object cost that depends on the detected object. The object cost of an object for a candidate trajectory increases as the distance between the object and the candidate trajectory decreases. A target trajectory is selected from the set of candidate trajectories depending on the trajectory costs.

[0010] According to the present invention, for each detected object, a quality value is determined, which represents a measure of the reliability of the object's detection. For each object, a minimum allowable acceleration allowed when a vehicle brakes towards the object is determined depending on the object's quality value. Trajectory candidates and the object costs determined for these trajectory candidates are evaluated depending on the acceleration set by each trajectory candidate and depending on the minimum allowable acceleration when braking towards the object. The selection of the target trajectory is additionally performed taking into account the evaluation of the trajectory candidates and the object costs determined for these trajectory candidates.

[0011] This method allows, for example, in traffic scenarios where multiple objects are present, to take into account the quality of object detection, i.e., the probability of presence, and to adapt the maximum braking intervention accordingly when automatically controlling the vehicle. This ensures that stronger braking intervention than permitted is not performed solely due to an object with poor detection quality. However, if a stronger braking intervention is required based on an object with higher detection quality, performing this stronger braking intervention remains an option. It is also possible to drive around objects with poor detection quality.

[0012] In one possible embodiment of the method, the trajectory candidate with the lowest trajectory cost is selected from the set of trajectory candidates as the target trajectory, and thus the trajectory in which the detected objects have the least impact on the vehicle and / or the trajectory in which the least number of objects are present is selected as the target trajectory.

[0013] In another possible embodiment of the method, categorization is performed during the evaluation, and in this categorization, an unfiltered category and a filtered category are distinguished. All object costs and corresponding (relevant, suitable) trajectory candidates are assigned to the unfiltered category. That is, no trajectory candidates are filtered. Only object costs and corresponding trajectory candidates whose minimum acceleration value set by each trajectory candidate is greater than the minimum allowable acceleration of each object are assigned to the filtered category. This allows only trajectory candidates whose acceleration falls within the allowable range determined by the minimum allowable acceleration to be assigned to the filtered category. That is, trajectory candidates that result in unacceptable braking are not included in the filtered category. Taking categorization into account can further improve the reliability of the method.

[0014] In another possible embodiment of the method, when selecting a target trajectory, the trajectory candidate with the lowest trajectory cost is selected from the unfiltered category and the filtered category, respectively. For the two selected trajectory candidates, the minimum value of the acceleration set by each trajectory candidate is calculated, and the candidate with the larger minimum value is selected as the target trajectory. This means that when there are multiple possible trajectories, a trajectory that applies less strong braking toward the object while maintaining the same level of safety is selected as the target trajectory, thereby improving the comfort of the vehicle occupants while maintaining the same level of safety.

[0015] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. [Brief explanation of the drawings]

[0016] [Figure 1] 1 shows a schematic representation of the course of acceleration of a vehicle over time and the corresponding course of longitudinal (longitudinal) position of the vehicle over time as the vehicle follows a certain path of a certain track. [Figure 2] 2 shows a schematic representation of the course of the acceleration of a vehicle over time and the corresponding course of the longitudinal position of the vehicle over time as the vehicle follows different trajectory paths; [Figure 3] 10 shows a schematic representation of the minimum allowable acceleration, which depends on the reliability of the object detection. [Figure 4] 1 shows a schematic diagram of a method for determining object costs. [Figure 5] 1 illustrates a schematic diagram of a method for selecting a target trajectory from trajectories of different categories. [Figure 6] 1 shows a schematic plan view of a traffic scenario with one vehicle and several objects. DETAILED DESCRIPTION OF THE INVENTION

[0017] In all the drawings, the same reference numerals are used to designate corresponding parts.

[0018] Figure 1 shows the time course of the acceleration ax(t) of the vehicle 1 depending on time t (as a function of time t) and the longitudinal position x(t) of the vehicle 1 corresponding to that acceleration when the automated vehicle 1, in particular a highly automated vehicle or an autonomous vehicle, shown in detail in Figure 6, follows the path of the trajectories T1 to Tn shown in Figure 5.

[0019] Each of the tracks T1 to Tn is represented by its longitudinal position x(t), transverse position, and their derivatives, i.e., the velocities in the longitudinal (vertical) and transverse (horizontal) directions, and the accelerations ax(t) in the longitudinal direction and the acceleration in the transverse direction up to the planning period (planning area, planning range, planning point) P. Here, the longitudinal position x(t) is derived by integrating the predefined acceleration ax(t). The acceleration ax(t) is defined such that the integrated longitudinal position x(t) satisfies the following determination conditions. x1(t) < x2(t) (1) And min(ax1(t)) < min(ax2(t)), provided that t = [0, P] (2)

[0020] Here, x1(t) = the longitudinal position x(t) when following track T1 x2(t) = the longitudinal position x(t) when following track T2 ax1(t) = the acceleration in the longitudinal direction when following track T1, ax2(t) = the acceleration in the longitudinal direction when following track T2 That is.

[0021] In addition to the path that vehicle 1 should follow, for the tracks T1 to Tn, the acceleration profile (acceleration curve) when vehicle 1 follows that path is also set. Here, the acceleration ax(t) is the temporal acceleration profile (temporal acceleration curve) when following the tracks T1 to Tn. The acceleration ax(t) represents a negative acceleration when vehicle 1 decelerates. The acceleration ax(t) has a minimum value min(ax(t)).

[0022] Examples of the different temporal profiles of the accelerations ax1(t), ax2(t) of vehicle 1, and the corresponding temporal profiles of the longitudinal positions x1(t), x2(t) of vehicle 1, as well as the minimum values min(ax1(t) ) , min(ax2(t) ) are shown in Fig. 2.

[0023] In a method for planning a target trajectory for an autonomous vehicle 1, for example, a plurality of trajectories T1 to Tn are set in the form of a trajectory group. The number of trajectories T1 to Tn is, for example, on the order of 1000.

[0024] Here, the following conditions are applicable. x1(t) < x2(t), provided that min(ax1(t)) < min(ax2(t)) (3)

[0025] This condition means that, when min(ax1(t)) < min(ax2(t)), for any two trajectories T1 and T2 of the trajectory group, the trajectories T1 to Tn of the trajectory group are determined such that x1(t) < x2(t) holds. This is a peripheral condition applied when setting the trajectories T1 to Tn.

[0026] That is, in order to plan the target trajectory, a set of trajectory candidates for a predetermined planning period P is set, and each trajectory candidate sets the path that the vehicle 1 should follow when the trajectory candidate is selected as the target trajectory, and also sets the acceleration ax(t) when the vehicle 1 should follow that path.

[0027] In order to plan the target trajectory, objects 2 and 3 during the planning period P, which are shown in more detail in FIG. 6, are detected, and a trajectory cost is assigned to each trajectory candidate using a predetermined cost function. Here, the cost function includes an object cost that depends on the detected objects 2 and 3, and the object costs of objects 2 and 3 for a certain trajectory candidate increase as the distance between objects 2 and 3 and that trajectory candidate decreases. Subsequently, the target trajectory is selected from the set of trajectory candidates depending on the trajectory cost.

[0028] At this time, for each detected object 2 and 3, a measure of the reliability of the detection of objects 2 and 3, that is, a quality value Q representing the probability of the existence of objects 2 and 3 around the vehicle 1 is determined.

[0029] Furthermore, for each object 2 and 3, depending on the quality value Q of that object, the minimum allowable acceleration aQ that is allowed when the vehicle 1 brakes toward the corresponding object 2 and 3 is determined.

[0030] 3 shows such a minimum permissible acceleration aQ, which depends on the reliability of the detection of the objects 2, 3, i.e. on the quality value Q of the objects 2, 3. When the vehicle brakes, the minimum permissible acceleration aQ has a negative value. Since a negative acceleration is a deceleration, the minimum permissible acceleration aQ simultaneously determines the maximum permissible deceleration.

[0031] The minimum allowable acceleration aQ permitted when braking towards the objects 2, 3 is determined taking into account the following conditions: min(ax(t))≧aQ (4)

[0032] Here, min(ax(t)) is the minimum acceleration on the trajectories T1 to Tn, and when braking, min(ax(t)) also becomes a negative value.

[0033] That is, equation (4) checks whether the braking acceleration aQ remains within a tolerance range, which is determined by the quality value Q of the corresponding object 2, 3.

[0034] It is important here that for each value of the minimum permissible acceleration aQ there exists at least one predetermined trajectory T1 to Tn, where the following applies: min(ax(t))=aQ (5)

[0035] In Fig. 3, both the minimum allowable acceleration aQ and the quality value are divided into low, medium and high regions. It is clear that braking with maximum force is permitted only if the objects 2, 3 have a high quality value Q (represented by the unhatched region), whereas such braking is not permitted for objects 2, 3 with medium and low quality values ​​Q (represented by the hatched region).

[0036] In order to avoid problems when planning a target trajectory in the case where a collision between the vehicle 1 and objects 2, 3 having a low quality value Q can only be avoided on a target trajectory that uses stronger braking than is permitted, the trajectory candidates and the object costs determined for these trajectory candidates are further evaluated depending on the acceleration ax(t) set by each trajectory candidate and depending on the minimum permitted acceleration aQ when braking towards the objects 2, 3, and the selection of the target trajectory is additionally carried out taking into account the evaluation of the trajectory candidates and the object costs determined for these trajectory candidates.

[0037] Here, the object cost of the object is evaluated for each trajectory in the trajectory set, where, as already explained, the object cost of objects 2, 3 relative to a trajectory candidate increases as the longitudinal and transverse distances between objects 2, 3 and the trajectory candidate decrease.

[0038] If there are multiple objects 2, 3, each of these objects 2, 3 may contribute to the object cost of the trajectory candidate. In such a case, it is examined which of the objects 2, 3 is most relevant, i.e., contributes the most, in terms of object cost, and the determination of the object cost for the trajectory candidate is made solely based on the most relevant object 2, 3. In other words, the less relevant objects 2, 3 are not taken into account when determining the object cost.

[0039] When evaluating the trajectory candidates and their determined object costs, a categorization is performed that distinguishes between an unfiltered category A and a filtered category F, both of which are shown in Figure 5. All object costs and corresponding trajectory candidates are assigned to the unfiltered category A. In contrast, only object costs and corresponding trajectory candidates whose minimum acceleration value min(ax(t)) set by each trajectory candidate is greater than the minimum allowable acceleration aQ of each object 2, 3 are assigned to the filtered category F.

[0040] Due to the quality values ​​of objects 2 and 3, the maximum velocity towards those objects is aQ=-5m / s 2 is allowed to brake with an acceleration of -2 m / s and the corresponding trajectories T1 to Tn have an acceleration min(ax(t)) of -2 m / s 2 If the brake is required to reduce the speed to a value of -2 m / s, the condition min(ax(t))>aQ is met. 2 >-5m / s 2 The corresponding trajectories T1 to Tn and object costs are assigned to both category A and category F.

[0041] Due to the quality values ​​of objects 2 and 3, the maximum velocity towards those objects is aQ=-5m / s 2 is allowed to brake with an acceleration of -6m / s, and in the corresponding trajectories T1 to Tn, the acceleration min(ax(t)) is -6m / s 2 If a Q requires braking down to a value of , then the condition min(ax(t))>aQ is not met. In this case, the corresponding trajectories T1-Tn and object costs are assigned to the unfiltered category A but not to the filtered category F.

[0042] FIG. 4 shows a possible embodiment of the method steps for determining object costs.

[0043] In the first branch V1, it is checked whether all the trajectory candidates of the trajectory family have already been evaluated. In the positive case, represented by the Yes branch J1, the method ends.

[0044] In the negative case, represented by the No branch N1, a second branch V2 checks whether all objects 2, 3 of the trajectory candidate have been evaluated. In the positive case, represented by the Yes branch J2, in method step S1 the next trajectory candidate is selected and the method starts again with this trajectory candidate.

[0045] In the negative case, represented by the No branch N2, a determination of the object costs for the corresponding objects 2, 3 of the corresponding trajectory candidates is performed in a further method step S2.

[0046] Subsequently, in the third branch V3, the condition according to equation (4) is checked to determine whether the minimum acceleration in the corresponding trajectory T1 to Tn is greater than the minimum allowable acceleration aQ.

[0047] In the positive case, represented by the Yes branch J3, in a further method step S3 the object costs for the corresponding trajectories T1 to Tn in the unfiltered category A and in the filtered category F are updated.

[0048] In the negative case, represented by the No branch N3, in a further method step S4, the object costs for the corresponding trajectories T1 to Tn are updated only in the unfiltered category A.

[0049] After execution of method step S3 or method step S4, in a further method step S5 the next object 2, 3 of the respective trajectory candidate is selected and the examination of this object 2, 3 is continued in branch V2.

[0050] That is, an object cost is calculated for each trajectory T1-Tn in the group of trajectories and for each object 2, 3 included in the set of multiple objects 2, 3. If the condition min(ax(t))>aQ is met for each trajectory T1-Tn and each object 2, 3, the object cost is assigned to both categories A, F; otherwise, it is assigned only to the unfiltered category A.

[0051] When all object costs for all possible combinations of trajectories T1 to Tn have been calculated, the two best trajectories T1 to Tn are determined from both categories A and F, and a target trajectory is selected from the two trajectories T1 to Tn.

[0052] FIG. 5 shows a possible embodiment of a method procedure for such a determination and selection.

[0053] Now, in method step S6, from all the unfiltered trajectories T1 to Tn of category A, the trajectory T1 to Tn having the lowest cost is determined.

[0054] In another method step S7, from all trajectories T1 to Tn of the filtered category F, the trajectory T1 to Tn with the lowest cost is determined.

[0055] Next, in branch V, it is checked whether the minimum acceleration min(ax(t)) in the unfiltered trajectories T1 to Tn of category A is greater than or equal to the minimum acceleration min(ax(t)) in the filtered trajectories T1 to Tn of category F.

[0056] In the positive case, represented by the Yes branch J, the corresponding trajectory T1 to Tn of the unfiltered category A is selected as the target trajectory in method step S8. In the negative case, represented by the No branch N, the corresponding trajectory T1 to Tn of the unfiltered category A is selected as the target trajectory in method step S9.

[0057] FIG. 6 shows a plan view of a traffic scenario in which there is a vehicle 1 and several objects 2, 3, where object 2 has a low quality value Q and object 3 has a high quality value Q.

[0058] For example, if trajectory T5 is the most cost-effective trajectory candidate from the filtered category F, and the trajectory T5 reduces the acceleration ax(t) to, for example, -5 m / s 2 If you specify that the velocity should be reduced to a value of min(ax5(t))=-5m / s 2 applies.

[0059] In the unfiltered category A, for example, trajectory T1 is the trajectory candidate with the best cost. Trajectory T1 has an acceleration ax(t) of, for example, -3 m / s 2 This specifies that the velocity should be reduced to a value of min(ax1(t))=-3m / s. 2 applies.

[0060] -3 m / s 2 ≧-5m / s 2 In this case, the condition min(ax1(t))≧min(ax5(t)) is satisfied, so the trajectory T1 is selected as the target trajectory. On the other hand, if this condition is not satisfied, the trajectory T5 is selected as the target trajectory.

[0061] For example, min(ax1(t))=-5m / s 2 In this case, both trajectories T1 and T5 are braked with the same acceleration ax(t) towards the object 3 in front. Trajectory T1 also takes into account the object 2 with a low quality value Q, whereas trajectory T5 does not. Since the condition min(ax1(t)) ≥ min(ax5(t)) is also met here, trajectory T1 is selected as the target trajectory and therefore object 2 with a low quality value Q is avoided. [Prior art documents] [Patent documents]

[0062] [Patent Document 1] German Patent Application Publication No. 102020108857 [Patent Document 2] German Patent Application Publication No. 102020200183 [Patent Document 3] German Patent Application Publication No. 102015016544 [Patent Document 4] German Patent Application Publication No. 102016218121

Claims

1. A method for planning a target trajectory of an autonomous vehicle (1), comprising: a set of trajectory candidates for a given planning period (P) is set, each trajectory candidate setting a path that the vehicle (1) should follow if that trajectory candidate is selected as the target trajectory, and setting an acceleration (ax(t)) with which the vehicle (1) should follow that path; - an object (2, 3) is detected within said planning period (P), - assigning a trajectory cost to each of the trajectory candidates using a predetermined cost function, the cost function including an object cost that depends on the detected object (2, 3), the object cost of the object (2, 3) for the trajectory candidate increasing as the distance between the object (2, 3) and the trajectory candidate decreases; and The target trajectory is selected from the set of trajectory candidates depending on the trajectory cost, - for each detected object (2, 3), a quality value (Q) is determined which represents a measure for the reliability of the detection of said object (2, 3); - for each object (2, 3), a minimum allowable acceleration (aQ) is determined that allows the vehicle (1) to brake towards the object (2, 3) depending on the quality value (Q) of the object (2, 3); the trajectory candidates and the object costs determined for them are evaluated depending on the acceleration (ax(t)) set by each trajectory candidate and depending on the minimum allowable acceleration (aQ) when braking towards the object (2, 3); The selection of the target trajectory is additionally performed taking into account the evaluation of the trajectory candidates and the object costs determined for the trajectory candidates. A method characterized by:

2. The trajectory candidate having the lowest trajectory cost is selected from the set of trajectory candidates as the target trajectory.

2. The method of claim 1.

3. - during said evaluation, categorization is carried out, - the categorization distinguishes between unfiltered categories (A) and filtered categories (F), - the unfiltered category (A) is assigned all object costs and the corresponding trajectory candidates; To the filtered category (F) are assigned only the object costs and the corresponding trajectory candidates whose minimum value (min(ax(t))) of the acceleration (ax(t)) set by each of the trajectory candidates is greater than the minimum allowable acceleration (aQ) of each of the objects (2, 3).

3. The method according to claim 1 or 2.

4. When selecting the target trajectory, - the trajectory candidate with the lowest trajectory cost is selected from the unfiltered category (A) and the filtered category (F), respectively; - for each of the selected trajectory candidates, the minimum value (min(ax(t))) of the acceleration (ax(t)) set by the respective trajectory candidate is determined; The trajectory candidate having the minimum value (min(ax(t))) greater than - is selected as the target trajectory.

3. The method according to claim 3, which is dependent on claim 1.

Images