Adapting the driving behavior of an autonomous vehicle

Abstract

Method for adapting a driving behavior of an autonomous vehicle (217), wherein - a traffic situation is defined with a driving maneuver to be performed by the autonomous vehicle (217), - a model is trained to perform the driving maneuver in the traffic situation, - whereby the model takes into account and evaluates a traffic flow related to the traffic situation and the execution of the driving maneuver is trained in such a way that the traffic flow is optimized according to its evaluation, - and in the traffic situation, the driving maneuver is carried out by the autonomous vehicle (217) according to the model depending on the traffic flow, - where the traffic situation is reaching an intersection (201), the driving maneuver is driving through the intersection (201) and the following variables are defined and used as measures of traffic flow: - an individual waiting time that the autonomous vehicle (219) waits before the intersection (201) while being the first vehicle waiting before the intersection (201) in the lane (207) it is traveling in, - an integral waiting time indicating a total waiting time of vehicles (215, 216) waiting behind the autonomous vehicle (217) in front of the intersection (201) during the individual waiting time in the lane (207) traveled by the autonomous vehicle (217), and - a dominant traffic flow defined as the product of the individual waiting time with the number of all vehicles (208 to 214) that pass the intersection (201) during the individual waiting time using at least one lane (204 to 206) that the autonomous vehicle (217) must cross when entering the intersection (201) or into which the autonomous vehicle (217) must position itself.

Description

[0001] The invention relates to a method for adapting the driving behavior of an autonomous vehicle.

[0002] An autonomous vehicle is defined here as a self-driving vehicle, i.e., a vehicle that performs driving maneuvers independently without driver intervention. The independent execution of driving maneuvers places diverse demands on an autonomous vehicle. In particular, traffic situations, including other road users and their movements, must be detected and evaluated. For this purpose, an autonomous vehicle has a multitude of sensors to detect its surroundings, such as cameras, lidar, radar, and / or ultrasonic sensors. The sensor signals from these sensors are evaluated and assessed by the autonomous vehicle in order to execute driving maneuvers adapted to the traffic situation. Of particular importance is the requirement that driving maneuvers performed by the autonomous vehicle do not endanger other road users.An autonomous vehicle therefore typically exhibits a so-called defensive driving style, which prioritizes safety aspects. However, excessively defensive driving can lead to traffic disruptions. Furthermore, excessively defensive driving can annoy other road users and encourage them to perform risky driving maneuvers, such as dangerous overtaking maneuvers.

[0003] EP 3 598 414 A1 discloses a method for predicting a trajectory of at least one road user in order to avoid a collision between a host vehicle, in particular an autonomous vehicle, and the road user. A cluster of possible trajectories of the road user is determined, and at least one trajectory for the road user is predicted from this cluster. Based on the at least one predicted trajectory, the host vehicle executes at least one action.

[0004] WO 2020 / 040 975 A1 discloses a method in which data acquired by vehicle sensors is stored, describing the driving behavior of manually controlled vehicles in traffic situations. This data is used, for example, to train a machine-learning model for controlling an autonomous vehicle or to assist a driver of a vehicle. To limit the amount of data to be stored, high-resolution data of a driving behavior is only stored when the driving behavior is of particular interest, for example, when the driving behavior deviates significantly from a predicted driving behavior.

[0005] DE 10 2020 201 931 A1 describes a method for training at least one algorithm for a control unit of a motor vehicle using a self-learning neural network and, in this regard, discloses using an autonomously driving motor vehicle as a leading vehicle for vehicles driving behind it to optimize traffic flow. Regarding the definition of traffic flow, the document discloses that a traffic flow metric used in this context is or includes, for example, an average speed.

[0006] The invention is based on the object of reducing traffic disruptions caused by the driving behavior of an autonomous vehicle.

[0007] The object is achieved according to the invention with a method for adapting the driving behavior of an autonomous vehicle, wherein - a traffic situation is defined with a driving maneuver to be performed by the autonomous vehicle, - a model is trained to perform the driving maneuver in the traffic situation, - whereby the model takes into account and evaluates a traffic flow related to the traffic situation and the execution of the driving maneuver is trained in such a way that the traffic flow is optimized according to its evaluation, - and in the traffic situation, the driving maneuver is carried out by the autonomous vehicle according to the model depending on the traffic flow - where the traffic situation is reaching an intersection, the driving maneuver is entering the intersection and the following quantities are defined and used as measures of traffic flow: - an individual waiting time that the autonomous vehicle waits before the intersection while being the first vehicle waiting before the intersection in the lane it is traveling in, - an integral waiting time, which indicates a total waiting time of vehicles waiting in front of the intersection during the individual waiting time in the lane occupied by the autonomous vehicle behind the autonomous vehicle, and - a dominant traffic flow, which is defined as the product of the individual waiting time and the number of all vehicles that pass the intersection during the individual waiting time using at least one lane that the autonomous vehicle must cross when entering the intersection or into which the autonomous vehicle must merge.

[0008] The method according to the invention aligns the driving behavior of an autonomous vehicle in a traffic situation with its effects on a traffic flow related to the traffic situation. A model of the driving behavior of the autonomous vehicle in the traffic situation is trained with the goal of optimizing the traffic flow. This distinguishes the method according to the invention from, for example, the prior art known from EP 3 598 414 A1, which is directed at avoiding collisions between a host vehicle, in particular an autonomous vehicle, and other road users.

[0009] During model training, the impact of the autonomous vehicle's driving behavior on traffic flow is evaluated and optimized using suitable optimization parameters. Thus, the driving behavior of the autonomous vehicle and its effects are retrospectively examined and evaluated during model training from the perspective of optimizing traffic flow. Such a retrospective examination and evaluation of driving behavior is a further difference from the prior art known, for example, from EP 3 598 414 A1 and WO 2020 / 040 975 A1.

[0010] In one embodiment of the invention, the model is based on an artificial neural network. Artificial neural networks are tools of artificial intelligence used, for example, for pattern recognition and machine learning. They are therefore particularly suitable for machine learning of driving behavior based on training data.

[0011] In another embodiment of the invention, the model is trained using data recorded in reality. In particular, the model can use reinforced learning to specifically reward executions of the driving maneuver that were actually performed by non-autonomous vehicles.

[0012] The aforementioned embodiment of the invention makes it possible to adapt the driving behavior of the autonomous vehicle to the driving behavior of non-autonomously driving vehicles. This makes it possible, in particular, to avoid excessively defensive driving behavior of the autonomous vehicle, which could lead to unnecessary delays of the autonomous vehicle and obstructions to the driving of other vehicles.

[0013] In a further embodiment of the invention, the traffic flow in the traffic situation is detected using at least one sensor of the autonomous vehicle. Sensors that can be used to detect traffic flow include, for example, cameras, lidar, radar, and ultrasonic sensors. An autonomous vehicle typically has several such sensors.

[0014] In a further embodiment of the invention, when evaluating the traffic flow, driving delays of both the autonomous vehicle and the driving delays of other vehicles caused by the driving behavior of the autonomous vehicle are taken into account. For example, when optimizing the traffic flow, a Pareto optimization is performed with the goals of minimizing driving delays of the autonomous vehicle and driving delays of other vehicles caused by the driving behavior of the autonomous vehicle.

[0015] Pareto optimization, also known as multi-objective optimization, is the solution of an optimization problem with multiple, often opposing, objectives. In this case, these objectives are minimizing the driving delays of various vehicles, including the autonomous vehicle itself. According to the aforementioned embodiment of the invention, when evaluating and optimizing traffic flow, driving delays of the autonomous vehicle itself are also included in the evaluation. This also helps prevent overly defensive driving behavior by the autonomous vehicle. For example, minor driving delays of other vehicles caused by the driving maneuver of the autonomous vehicle can be accepted in order to avoid excessive driving deceleration of the autonomous vehicle.

[0016] In a further embodiment of the invention, the traffic situation is reaching an intersection, and the driving maneuver to be performed by the autonomous vehicle is entering the intersection, for example turning at the intersection or crossing the intersection on a lane already occupied by the autonomous vehicle before reaching the intersection. An intersection is understood here to be any meeting of traffic routes. For example, a junction (“T-junction”) where one traffic route ends and meets another traffic route is also referred to as an intersection. Reaching an intersection is a potentially challenging traffic situation for an autonomous vehicle, as it can be entered by vehicles coming from different directions, and relatively complex right-of-way rules often have to be observed, especially if the intersection is not controlled by traffic lights, for example.The method according to the invention therefore aims in particular at the driving behavior of an autonomous vehicle when reaching and entering an intersection.

[0017] In the aforementioned embodiment of the invention, the following quantities are defined and used as measures of traffic flow: - an individual waiting time that the autonomous vehicle waits before the intersection while being the first vehicle waiting before the intersection in the lane it is traveling in, - an integral waiting time, which indicates a total waiting time of vehicles waiting in front of the intersection during the individual waiting time in the lane occupied by the autonomous vehicle behind the autonomous vehicle, and - a dominant traffic flow, which is defined as the product of the individual waiting time and the number of all vehicles that pass the intersection during the individual waiting time using at least one lane that the autonomous vehicle must cross when entering the intersection or into which the autonomous vehicle must merge.

[0018] The individual waiting time is a measure of the autonomous vehicle's travel delay. The integral waiting time is the sum of the waiting times of other vehicles waiting behind the autonomous vehicle at the intersection during the individual waiting time. The integral waiting time is therefore a measure of the travel delays of all vehicles that are prevented from continuing in the lane occupied by the autonomous vehicle due to the autonomous vehicle waiting at the intersection during the individual waiting time. The dominant traffic flow is a measure of the traffic flow in all other lanes where vehicles could potentially be impeded by the autonomous vehicle's maneuver.

[0019] In the case of reaching an intersection, for example, an ego-traffic flow ratio is minimized to optimize traffic flow. This ratio is defined as the sum of the individual waiting time and the integral waiting time divided by the dominant traffic flow. According to the definition of the ego-traffic flow ratio, minimizing the ego-traffic flow ratio aims to achieve an optimized balance between the driving delays of the autonomous vehicle and the driving delays of other vehicles caused by the driving behavior of the autonomous vehicle.

[0020] Alternatively, the ego-traffic flow ratio is minimized, for example, under the constraint that maneuver deceleration is minimal. The maneuver deceleration is defined as a function of maneuver-induced speed differences for the speeds of vehicles decelerated as a result of executing the maneuver, with each maneuver-induced speed difference being defined as the difference between a reference speed for the decelerated vehicle and a speed to which this vehicle was decelerated.

[0021] For example, the maneuver deceleration is defined as a sum of speed differences caused by driving maneuvers during an evaluation period.

[0022] The evaluation period extends, for example, from the start of the driving maneuver to a latest point in time at which speed differences caused by the driving maneuver can be detected by at least one sensor of the autonomous vehicle.

[0023] The reference speed of a decelerated vehicle is, for example, the minimum speed of the decelerated vehicle that is compatible with the traffic flow immediately before the maneuver is executed and the maximum permitted speed in a lane occupied by the decelerated vehicle. Alternatively, the reference speed of a decelerated vehicle is, for example, the speed of the decelerated vehicle immediately before the maneuver is executed.

[0024] Embodiments of the invention are explained in more detail below with reference to the drawings. Fig. 1 a flowchart of an embodiment of the method according to the invention, Fig. 2 a street scene at an intersection.

[0025] Fig. 1 ( Fig. 1) shows a flowchart 100 of an embodiment of the method according to the invention with method steps 101 to 103 for adapting a driving behavior of an autonomous vehicle.

[0026] The method steps 101 to 103 are also described below with reference to Fig. 2 described.

[0027] Fig. 2 ( Fig. 2) schematically shows a street scene 200 at an intersection 201. At intersection 201, a first street 202 and a second street 203 intersect, with the first street 202 being the priority road at intersection 201. The first street 202 has two lanes 204, 205 with different prescribed directions of travel. The second street 203 also has two lanes 206, 207 with different prescribed directions of travel. The respective prescribed directions of travel of lanes 204 to 207 are shown in Fig. 2 represented by arrows.

[0028] Also shown are vehicles 208, 209, 210 traveling in a first lane 204 of the first road 202, vehicles 211, 212, 213 traveling in the second lane 205 of the first road 202, a vehicle 214 traveling in a first lane 206 of the second road 203, and vehicles 215, 216, 217 traveling in the second lane 207 of the second road 203. The vehicle 217 is an autonomous vehicle.

[0029] In a first method step 101, a traffic situation with a driving maneuver to be performed by the autonomous vehicle 217 is defined.

[0030] The definition of the traffic situation includes, for example, a description of the static structure of a traffic environment. This includes, for example, traffic routes in the traffic environment, such as roads and their lanes, as well as their connections, but also pedestrian crossings or traffic lights. Furthermore, the definition of the traffic situation includes, for example, traffic regulations applicable in the traffic environment, such as right-of-way rules or speed limits. Furthermore, the definition of the traffic situation includes a description of a driving maneuver to be performed by the autonomous vehicle 217, i.e., a driving task for the autonomous vehicle 217.

[0031] In the Fig. 2, the traffic situation is reaching the intersection 201 with a driving task for the autonomous vehicle 217. The corresponding definition of a traffic situation includes an abstraction of the Fig. 2 with two intersecting streets 202, 203 and their lanes 204 to 207. Furthermore, the definition of the traffic situation includes the traffic regulations applicable at the intersection 201, in particular the characterization of the first street 202 as a priority road and the directions of travel prescribed for lanes 204 to 207. Furthermore, the definition of the traffic situation includes a description of the driving task for the autonomous vehicle 217, i.e., the driving maneuver to be performed by the autonomous vehicle 217 at the intersection 201. The driving task is, for example, turning at intersection 201 from the second lane 207 of the second road 203 onto the first lane 204 of the first road 202, turning at intersection 201 from the second lane 207 of the second road 203 onto the second lane 205 of the first road 202, or crossing intersection 201 on the second lane 207 of the second road 203.

[0032] In a second method step 102, a model is trained to execute the driving maneuver in the traffic situation. The model is based, for example, on an artificial neural network.

[0033] The model considers and evaluates a traffic flow related to the traffic situation. The execution of the driving maneuver is trained in such a way that the traffic flow is optimized according to its evaluation. When evaluating the traffic flow, driving delays of both the autonomous vehicle 217 and the driving delays of other vehicles 208 to 216 caused by the driving behavior of the autonomous vehicle 217 are taken into account. When optimizing the traffic flow, for example, a Pareto optimization is performed with the goals of minimizing driving delays of the autonomous vehicle 217 and driving delays of other vehicles 208 to 216 caused by the driving behavior of the autonomous vehicle 217.

[0034] In the following, the training of such a model is described using the example of a traffic situation that requires the achievement of a Fig. 2, wherein the driving maneuver to be performed by the autonomous vehicle 217 at the intersection 201 is a turn at the intersection 201 from the lane 207 of the second road 203 onto one of the lanes 204, 205 of the first road 202 or a crossing of the intersection 201 on the lane 207 of the second road 203.

[0035] For example, a stopping point is defined at which the autonomous vehicle 217 must stop on lane 207 before the intersection 201 at the latest in order not to encroach into lane 205 of the first road 202. At this stopping point, the autonomous vehicle 217 is in Fig. 2. At the stopping point, the autonomous vehicle 217 must generally stop and wait until it can enter intersection 201. Due to the right-of-way nature of the first road 202, it must first allow vehicles 208 to 213 to pass through intersection 201. These vehicles are entering intersection 201 from one of the lanes 204, 205 of the first road 202. At the stopping point, the autonomous vehicle 217 is the first vehicle waiting in front of intersection 201 in the lane 207 it is traveling in. Vehicles 215, 216 traveling behind the autonomous vehicle 217 in lane 207 must also stop and wait until the autonomous vehicle 217 departs from the stopping point.

[0036] Training the model serves, in particular, to determine the time at which the autonomous vehicle 217 departs from the stopping point to execute the driving maneuver. This time is determined such that executing the driving maneuver optimizes the traffic flow according to its evaluation.

[0037] To evaluate and optimize traffic flow, the following variables are defined and used as traffic flow measures: - an individual waiting time that the autonomous vehicle 217 waits at the stopping point, or that the autonomous vehicle 217 waits before the intersection 201 while it is the first vehicle waiting before the intersection 201 in the lane 207 it is traveling in, - an integral waiting time indicating a total waiting time of vehicles 215, 216 waiting behind the autonomous vehicle 217 in front of the intersection 201 during the individual waiting time, i.e. the sum of the waiting times of all vehicles 215, 216 waiting behind the autonomous vehicle 217 in front of the intersection 201 during the individual waiting time, and - a dominant traffic flow, which is defined as the product of the individual waiting time with the number of all vehicles 208 to 214 that pass the intersection during the individual waiting time using at least one lane 204 to 206 that the autonomous vehicle 217 must cross when entering the intersection 201 or into which the autonomous vehicle 217 must enter.

[0038] In the case where the driving maneuver is crossing the intersection 201 on lane 207, the autonomous vehicle 217 must cross the two lanes 204, 205 of the first road 202. In this case, the dominant traffic flow is the product of the individual waiting time and the number of all vehicles 208 to 214 that pass the intersection during the individual waiting time using one of the lanes 204, 205 of the first road 202.

[0039] In the case where the driving maneuver is a turn from lane 207 to lane 205, the autonomous vehicle 217 must enter lane 205. In this case, the dominant traffic flow is the product of the individual waiting time and the number of all vehicles 211 to 213 that pass the intersection during the individual waiting time using lane 205 of the first road 202.

[0040] In the case where the driving maneuver is a turn from lane 207 to lane 204, the autonomous vehicle must cross lanes 205 and 206 and enter lane 204. In this case, the dominant traffic flow is the product of the individual waiting time and the number of all vehicles 208 to 214 that pass the intersection during the individual waiting time using at least one of lanes 204 to 206.

[0041] From the individual waiting time, the integral waiting time, and the dominant traffic flow, an ego-traffic flow ratio related to the autonomous vehicle 217 is formed, which is defined as the sum of the individual waiting time and the integral waiting time divided by the dominant traffic flow. The model for executing the driving maneuver is then trained, for example, in such a way that the ego-traffic flow ratio is minimized.

[0042] Alternatively, the model for executing the driving maneuver is trained, for example, such that the ego-traffic flow ratio is minimized under the constraint that maneuver deceleration is minimal. The maneuver deceleration is defined as a function of maneuver-induced speed differences for speeds of vehicles 208 to 214 decelerated as a result of executing the driving maneuver, wherein each maneuver-induced speed difference is defined as the difference between a reference speed for the decelerated vehicle 208 to 214 and a speed to which this vehicle 208 to 214 was decelerated.

[0043] For example, maneuver deceleration is defined as the sum of maneuver-induced speed differences during an evaluation period. The evaluation period extends, for example, from the start of the maneuver to the latest point in time at which maneuver-induced speed differences can be detected by at least one sensor of the autonomous vehicle.

[0044] The reference speed of a braked vehicle 208 to 214 is, for example, the minimum of a speed of the braked vehicle 208 to 214 that is compatible with the traffic flow immediately before the driving maneuver is carried out and a maximum permissible speed on a lane 204 to 206 traveled by the braked vehicle 208 to 214. Alternatively, the reference speed of a braked vehicle 208 to 214 is the speed of the braked vehicle 208 to 214 immediately before the driving maneuver is carried out.

[0045] In this case, the model serves as a classifier of a situation at the stopping point of autonomous vehicle 217 before intersection 201. The model determines a current recommendation for stopping at the stopping point or driving off from the stopping point as an output parameter. The model's input parameters are the description of the traffic situation with the driving maneuver to be performed and dynamic boundary conditions, such as the current occupancy of the individual lanes 204 to 207, current traffic flows in lanes 204 to 207, or average vehicle spacing in lanes 204 to 207.

[0046] In a third method step 103, the model is applied in reality by the autonomous vehicle 217. The autonomous vehicle 217 typically follows a route from a starting position to a destination position. For navigation and object detection along the route, the autonomous vehicle 217 has a navigation device and a plurality of sensors, for example cameras, lidar, radar and / or ultrasonic sensors. The navigation device comprises a positioning unit that receives and evaluates satellite signals from navigation satellites of one or more navigation satellite systems such as GPS, GLONASS, Galileo and / or Beidou in order to determine and track the current position of the autonomous vehicle 217. Furthermore, the navigation device evaluates map data from one or more digital maps that are stored in a memory unit of the autonomous vehicle 217 and / or from a data cloud service.Cloud). In addition, the autonomous vehicle 217 can be configured to receive and evaluate data from a V2X communication. V2X communication is also referred to as vehicle-to-everything communication, car2X communication, or car-to-X communication and enables the exchange of data between a vehicle and the vehicle's surroundings, in particular also communication between vehicles, which is also referred to as car2car communication, car-to-car communication, V2V communication, or vehicle-to-vehicle communication. V2X communication is usually carried out via radio signals in frequency bands of the WLAN, with DSRC (abbreviation for Dedicated Short Range Communication), or mobile network.

[0047] To apply the trained model, the autonomous vehicle 217 continuously records the dynamic input parameters of the model as it approaches the traffic situation. In the example described above using Fig. 2, the autonomous vehicle 217, for example, uses its sensors when approaching the intersection 201 to detect the current occupancy of the individual lanes 204 to 207, the current speeds of vehicles 208 to 216 in the lanes 204 to 207, and the current vehicle spacing in the lanes 204 to 207. Depending on the dynamic input parameters, the model provides a continuous recommendation for action to the autonomous maneuver planning of the autonomous vehicle 217. In the example described above using Fig. In the traffic situation described in section 2, the recommended course of action at the stop point on lane 207 before intersection 201 is, for example, the recommendation to wait at the stop point or to drive off from the stop point.

[0048] The second method step 102 is first performed in a training phase. The model is preferably trained using data recorded in reality. For example, through reinforcement learning, the model particularly rewards executions of the driving maneuver that were actually performed by non-autonomous vehicles. This allows the execution of the driving maneuver by the autonomous vehicle 217 to be adapted to the behavior of drivers of non-autonomous vehicles. After the training phase, the trained model is transferred to the autonomous vehicle 217.

[0049] In the third method step 103, the autonomous vehicle 217 can record the traffic flows during and after the application of the driving maneuver and determine the resulting optimization parameters, which are optimized during training of the model. In the method described above with reference to Fig.In the traffic situation described in Figure 2, the autonomous vehicle 217 records, for example, its individual waiting time, the integral waiting time, the dominant traffic flow and the speed differences of other vehicles 208 to 213 caused by driving maneuvers, and the ego-traffic flow ratio and the maneuver delay are determined from this as optimization parameters.

[0050] The optimization parameters determined during and after applying the driving maneuver can then be used to further train the model, which can lead to an improvement of the model. This further training of the model can be carried out, for example, by an in-vehicle unit of the autonomous vehicle 217. Alternatively or additionally, the determined optimization parameters are sent, for example, to a data cloud that manages an instance of the model and continuously trains and improves the model based on data sent to it, including from other vehicles. The instance of the model trained in the data cloud can be used, for example, to implement it in an autonomous vehicle 217 or to replace or update an instance of the model already implemented in an autonomous vehicle 217. List of reference symbols 100 Flowchart 101 to 103 process step 200 Street Scene 201 intersection 202, 203 Street 204 to 207 lanes 208 to 216 vehicles 217 autonomous vehicle

Claims

[1] Method for adapting a driving behavior of an autonomous vehicle (217), wherein - a traffic situation is defined with a driving maneuver to be performed by the autonomous vehicle (217), - a model is trained to perform the driving maneuver in the traffic situation, - whereby the model takes into account and evaluates a traffic flow related to the traffic situation and the execution of the driving manoeuvre is trained in such a way that the traffic flow is optimised according to its evaluation, - and in the traffic situation, the driving maneuver is carried out by the autonomous vehicle (217) according to the model depending on the traffic flow, - where the traffic situation is reaching an intersection (201), the driving maneuver is driving through the intersection (201) and the following variables are defined and used as measures of traffic flow: - an individual waiting time that the autonomous vehicle (219) waits before the intersection (201) while being the first vehicle waiting before the intersection (201) in the lane (207) it is traveling in, - an integral waiting time indicating a total waiting time of vehicles (215, 216) waiting behind the autonomous vehicle (217) in front of the intersection (201) during the individual waiting time in the lane (207) traveled by the autonomous vehicle (217), and - a dominant traffic flow defined as the product of the individual waiting time with the number of all vehicles (208 to 214) passing the intersection (201) during the individual waiting time using at least one lane (204 to 206) that the autonomous vehicle (217) must cross when entering the intersection (201) or into which the autonomous vehicle (217) must position itself. [2] The method according to claim 1, wherein the model is based on an artificial neural network. [3] Method according to claim 1 or 2, wherein the model is trained with data recorded in reality. [4] Method according to claim 3, wherein the model, through reinforcement learning, particularly rewards executions of the driving maneuver that were carried out in reality by non-autonomous vehicles. [5] Method according to one of the preceding claims, wherein the traffic flow in the traffic situation is detected with at least one sensor of the autonomous vehicle (217). [6] Method according to one of the preceding claims, wherein, when evaluating the traffic flow, driving delays of both the autonomous vehicle (217) and driving delays of other vehicles (208 to 216) caused by the driving behavior of the autonomous vehicle (217) are taken into account. [7] Method according to claim 6, wherein, when optimizing the traffic flow, a Pareto optimization is carried out with the objectives of minimizing driving delays of the autonomous vehicle (217) and driving delays of other vehicles (208 to 216) caused by the driving behavior of the autonomous vehicle (217). [8] Method according to one of the preceding claims, wherein to optimize the traffic flow an ego-traffic flow ratio is minimized, which is defined as the sum of the individual waiting time and the integral waiting time divided by the dominant traffic flow. [9] Method according to claim 8, wherein the ego traffic flow ratio is minimized under the constraint that a maneuver delay is minimal, wherein the maneuver delay is defined as a function of maneuver-induced speed differences for speeds of vehicles (208 to 214) braked as a result of performing the maneuver, wherein each maneuver-induced speed difference is defined as the difference between a reference speed for the braked vehicle (208 to 214) and a speed to which this vehicle (208 to 214) was braked. [10] Method according to claim 9, wherein the maneuver deceleration is defined as a sum of maneuver-induced speed differences during an evaluation period. [11] Method according to claim 10, wherein the evaluation period extends from a start of the driving maneuver to a latest time at which speed differences caused by the driving maneuver can be detected with at least one sensor of the autonomous vehicle (217). [12] Method according to one of claims 9 to 11, wherein the reference speed of a braked vehicle (208 to 214) is the minimum of a speed of the braked vehicle (208 to 214) compatible with the traffic flow immediately before the driving maneuver is carried out and of a maximum permissible speed on a lane (204 to 206) traveled by the braked vehicle (208 to 214). [13] Method according to one of claims 9 to 11, wherein the reference speed of a braked vehicle (208 to 214) is the speed of the braked vehicle (208 to 214) immediately before the driving maneuver is carried out.

Images