Method for selecting an automated driving process using a driver assistance system

The decision tree-based method in driving assistance systems predicts driver and external actor actions to select automated driving processes that align with human behavior, enhancing comfort and safety by optimizing lane changes and speed adjustments.

DE102020215780B4Active Publication Date: 2024-11-07AUMOVIO AUTONOMOUS MOBILITY GERMANY GMBH

Patent Information

Application Number
DE102020215780
Authority / Receiving Office
DE · DE
Patent Type
Patents
Current Assignee / Owner
Filing Date
2020-12-14
Publication Date
2024-11-07
Estimated Expiration
2040-12-14

AI Technical Summary

Technical Problem

Existing driving assistance systems, particularly at autonomy level 2 (SAE level 2), struggle to mimic human driving behavior by accurately predicting and executing lane changes and speed adjustments in response to both driver and external actor actions, leading to suboptimal driving experiences.

Method used

A method using a decision tree-based approach that integrates driver and external actor probabilities, evaluation indicators, and trajectory calculations to select an automated driving process that aligns with human driving behavior, considering comfort and safety.

Benefits of technology

Enables the vehicle to select driving processes that closely resemble human behavior by minimizing abrupt maneuvers and optimizing comfort and safety through probabilistic decision-making and trajectory planning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

The invention relates to a method for selecting an at least partially automated driving process by means of an autonomous driving assistance system of a vehicle.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] The invention relates to a method for selecting an at least partially automated driving process by means of a driver assistance system of a vehicle.

[0002] Driver assistance systems are known that can calculate a driving trajectory based on the current driving scenario, along which the vehicle moves automatically, i.e., for example, without the driver's steering intervention and / or without the driver pressing the accelerator or brake pedal. In particular, driver assistance systems according to Autonomy Level 2 (SAE Level 2) are known that have the capability of keeping in lane but cannot change lanes without the driver's initiative. If a driver initiates a lane change, it can then be performed automatically.

[0003] Depending on the respective driving scenario, it is possible to initiate different driving actions. For example, on a multi-lane road, when approaching a slower vehicle ahead, it is possible to reduce speed at different times or distances from this vehicle. This depends in particular on whether the driver of the ego vehicle with a driver assistance system according to autonomy level 2 or lower is expected to initiate a lane change or not.

[0004] In another driving scenario, the potential traffic action of a third-party actor determines whether it is advantageous to initiate a particular driving maneuver or not. This means that the trigger, which cannot be influenced by the driver assistance system, is not triggered by the driver of the ego vehicle, but by another road user (e.g., another vehicle, pedestrian, cyclist, etc.) or a traffic control device (traffic light, barrier, etc.).

[0005] If the driver is expected to initiate a lane change, it is advantageous not to reduce speed early, as this would result in an unnecessary deceleration from the perspective of a human driver. Conversely, if the driver is expected not to initiate a lane change, it is advantageous to reduce speed early to prevent following too closely to the other vehicle in front and the associated sudden braking. The same applies to driving scenarios in which a third-party actor's traffic action determines which driving procedure is advantageous in terms of safety and / or comfort.

[0006] DE 10 2013 212 359 A1 describes a method for predicting a vehicle's driving paths. Based on the vehicle's current driving position and the position and movement of other road users, possible combinations of several consecutive driving maneuvers are determined using a hypothesis tree. For each driving path in a group of possible driving paths, the probability of the driver executing a driving maneuver, as well as a collision metric or probability of a collision with other road users, are determined and provided.

[0007] DE 10 2014 209 520 A1 describes a method for an automatic distance control system of a vehicle, wherein an interceptor vehicle is detected that is in a lane different from the vehicle's own lane, and the interceptor vehicle's intention to change lanes from the other lane to the vehicle's own lane is detected based, for example, on a turn signal of a turn signal of the interceptor vehicle. Finally, it is determined whether a predetermined minimum distance from the interceptor vehicle can be maintained by decelerating the vehicle with a predetermined deceleration.

[0008] DE 10 2009 020 649 A1 describes a method for collision avoidance for a vehicle by evading an obstacle. In the event of an impending collision, several possible evasive trajectories are determined, and time interval limits are assigned to the evasive trajectories. These time intervals represent the time interval to the obstacle at which an evasive maneuver must be initiated according to the respective trajectory to avoid a collision.

[0009] It is an object of the invention to provide a method for selecting an at least partially automated driving process by means of a driver assistance system, which enables control of the vehicle that approximates human driving behavior as closely as possible.

[0010] This object is achieved by a method having the features of independent patent claim 1. Preferred embodiments are the subject of the dependent claims. A driver assistance system is the subject of independent patent claim 13, and a vehicle with such a driver assistance system is the subject of independent patent claim 14.

[0011] According to a first aspect, the invention relates to a method for selecting an at least partially automated driving process using a driver assistance system of a vehicle. The method comprises the following steps: First, the current driving scenario is determined, i.e., the driving situation in which the vehicle currently finds itself. This could be, for example, "Vehicle is driving in the right lane of a two-lane road, left lane clear, slower vehicle 150 meters ahead." Based on the current driving scenario, the driver assistance system determines several different, future possible driving actions. In the driving scenario described above, these could be, for example, lane changing at a constant speed, lane keeping at a constant speed, or lane keeping with deceleration. It is understood that these driving actions can be initiated at different times.

[0012] A decision structure, also referred to as a decision tree, is then automatically created by a computing unit of the driver assistance system. In this process, several decision levels of the decision tree, also referred to as nodes, are determined, with each node being linked to specific points in time in the future. Depending on the driving scenario, at least one decision to be made by the driver of the vehicle or by an external actor regarding a lane change or traffic action is assigned to the nodes, at least in part. An external actor can be, for example, an external vehicle, a pedestrian, a cyclist or any other object that can perform a traffic action, such as changing lanes, changing direction, entering the roadway, etc.

[0013] In addition, at least two decisions or action options to be made by the driver assistance system itself are assigned to each node, at least in part. Thus, according to the decision tree, multiple decisions can be made at the points in time linked to the nodes of the decision tree, either initiated by the driver of the ego vehicle (own lane change) or by an external actor (e.g., lane change of another vehicle), or by a decision of the driver assistance system itself (e.g., lane keeping with deceleration, lane keeping without deceleration, heavy braking to avoid a collision).

[0014] The decisions to be made at the respective nodes define sub-drive processes, which are components of a longer-duration drive process. In other words, a drive process consists of several sub-drive processes assigned to the branches of the decision tree. A path through the decision tree, for example, which extends over several nodes, is thus composed of several decisions and associated sub-drive processes.

[0015] Driving trajectories are then calculated, with each driving trajectory being assigned to a defined path through the decision tree, and each path comprising one or more decisions to be made sequentially. The respective driving trajectory defines, for example, the vehicle's trajectory in a two-dimensional coordinate system and preferably also specifies the vehicle's speed along this trajectory. When calculating the driving trajectory, the longitudinal and lateral acceleration, or the longitudinal and lateral jerk, can also be determined. Thus, in this step, several driving trajectories are calculated that can be followed in the current driving scenario, depending on the respective decisions that can be made according to the decision tree.

[0016] In addition, evaluation indicators are calculated for the driving trajectories. The evaluation indicator is a measure of the comfort and / or safety of the respective driving trajectory. The evaluation indicator is preferably a number, so that by comparing the evaluation indicators, the individual driving trajectories can be compared with each other in terms of comfort and / or safety.

[0017] In addition, probability values ​​are calculated that are assigned to the nodes of the decision tree. These probability values ​​indicate the probability that the driver of the vehicle or an external actor will make a decision to change lanes or perform another traffic action at the time assigned to the node. In other words, the probability value indicates how likely it is that a decision will be made by the driver of the vehicle or an external actor at the time assigned to the node, to which the driver assistance system of the ego vehicle must respond by providing a driving trajectory.

[0018] Subsequently, several decision indicators are calculated, each of which is assigned to a node. The calculation of the decision indicator is based at least in part on the probability value of the respective node, which indicates that the driver of the vehicle or the third-party actor will make a decision to change lanes or perform a traffic action at the time assigned to the node. Furthermore, the calculation of the decision indicator is based at least on an evaluation indicator of a driving trajectory determined by a decision to be made at the time of the respective node.

[0019] Finally, an at least partially automated driving process is selected by the autonomous driving assistance system based on the decision tree, the decision indicators, and whether the driver of the vehicle or the external actor initiates a lane change or a traffic action at the time assigned to the node. In particular, if a decision by the driver of the ego vehicle or a decision by an external actor does not already force the initiation of another driving process, the decision tree is traversed to the node with the lowest decision indicator, and there, the driving assistance system automatically initiates a lane-keeping decision with a simultaneous low deceleration (for example, at least 80% or 90% less deceleration than during emergency braking).

[0020] The technical advantage of the method according to the invention is that, by considering the probability of initiating a lane change or a traffic action and by considering the trajectory characteristics with regard to comfort and / or safety, a driving procedure is automatically selected that enables the most natural driving experience possible, oriented towards human driving behavior. In other words, a driving procedure is selected that closely corresponds to the driving behavior of a human driver.

[0021] To select an at least partially automated driving process, the values ​​of the decision indicators of the nodes of the decision tree are compared. The selected driving process is based on a driving trajectory that initiates a lane change or deceleration process at the time associated with the node with the lowest decision indicator value. In other words, a comparison of the decision indicators of the nodes is performed, and it is determined which node is assigned the lowest decision indicator value.The lowest decision indicator value is an indication that, taking into account the probability that the driver initiates a lane change at the time of the node and the evaluation indicators of the driving trajectories of this node, it is advantageous to either initiate the lane change at this time (which must be initiated by the driver of the ego vehicle or the other vehicle) or to stay in the lane and decelerate the vehicle (initiated by the driver assistance system alone).

[0022] It should be noted that, preferably, only the driver should initiate a lane change. The lane change itself can be performed automatically by the driver assistance system.

[0023] According to one embodiment, the decision indicators of the nodes are calculated at least partially by adding the results of a first multiplication and a second multiplication. During the first multiplication, the probability value that the driver of the vehicle or the external actor initiates a lane change or a traffic action at the time assigned to the node is multiplied by the evaluation indicator of the driving trajectory by means of which the lane change is carried out. During the second multiplication, the probability value that the driver of the vehicle or the external actor does not initiate a lane change or a traffic action at the time assigned to the node is multiplied by the evaluation indicator of the driving trajectory by means of which the lane is maintained and the vehicle is decelerated.In order to calculate the decision indicators that are assigned to the nodes and thus to the respective decision times, both the evaluation indicators of the driving trajectories that are selected at this time and the probabilities that these driving trajectories are selected due to the behavior of the driver of the own vehicle or the external actor are taken into account.

[0024] According to one embodiment, a decision to be made by the vehicle driver by activating the turn signal and / or a steering intervention that causes a lane change is detected. This allows the driver assistance system to recognize when the driver intends to change lanes.

[0025] According to one embodiment, a decision to be made by the external actor is detected by detecting a turn signal and / or a lane change of a third-party vehicle, by a third-party vehicle turning, by a third-party vehicle approaching and merging into the driver's lane, or by a person entering the lane. This can be detected, for example, using a suitable sensor system, which may include a camera, a radar sensor, and / or a LIDAR sensor. This allows the driver assistance system to detect whether the third-party vehicle is changing lanes or whether the third-party actor is performing a traffic action that may require a driving maneuver to be performed by the driver assistance system.

[0026] According to one embodiment, the calculation of a driving trajectory includes the calculation of multidimensional coordinates along which the vehicle moves while traveling along the driving trajectory, the calculation of longitudinal and lateral acceleration values, and / or the calculation of longitudinal and lateral jerk values. This makes it possible to determine whether the trajectory is collision-free and what forces act on the vehicle's occupants when the vehicle is moving along the driving trajectory. These are essential parameters for the comfort and safety perception of the driving trajectory.

[0027] According to one embodiment, the evaluation indicator of a driving trajectory is calculated taking into account the vehicle's distance from other vehicles and / or objects, the difference from a set speed, and / or the lane change frequency. These are also essential parameters for the comfort and safety perception of the driving trajectory.

[0028] According to one embodiment, a driving trajectory corresponds to a path through the decision tree, whereby the driving trajectory is based on decisions associated with the path of the decision tree. Thus, the driving trajectory can be composed of several sub-trajectories, for example, an initial lane-keeping period at a constant speed followed by a later lane change.

[0029] According to one embodiment, the probability value that the driver will initiate a lane change at a future time associated with a node is calculated based on a neural network trained using data obtained from the driver's past driving maneuvers. This allows the probability of a lane change to be estimated or calculated from statistical correlations obtained in the past.

[0030] According to one embodiment, the probability value that the driver will initiate a lane change at a future time associated with a node is calculated based on the evaluation indicators of the driving trajectories initiated by decisions made at the times of the respective nodes. A human driver often initiates a lane change when a driving trajectory is available that allows for a sufficiently good lane change in terms of comfort and / or safety. Thus, the evaluation indicators of the driving trajectories can be used to determine or estimate the probability of a driver-initiated lane change.

[0031] According to one embodiment, the probability that the driver will initiate a lane change at a future time associated with a node is calculated based on driver behavior detected by at least one sensor. For example, the sensor can detect whether the driver has already looked in the side mirror, checked their shoulder, looked forward toward their own lane for an extended period, or similar factors. This allows a lane change probability to be calculated or estimated based on the sensor.

[0032] According to one embodiment, at least for some of the nodes, a plurality of decision indicators are calculated, each based on different probability values ​​that the driver of the vehicle or an external actor will make a decision regarding a lane change or a traffic action at the time assigned to the node. For each node, the highest decision indicator is used to select an at least partially automated driving process. The probability values ​​relating to a lane change or a traffic action are subject to a degree of uncertainty, which increases the further in the future the time to which the probability value refers lies.By calculating the decision indicators based on different probability values, a driving process can be selected that enables a driving process optimized in terms of comfort and / or safety, even if the probability values ​​are inaccurate.

[0033] According to one embodiment, the decision to be made by the driver of the vehicle includes a lane change from a right lane to a left lane or a lane change from a center lane to a left or right lane.

[0034] According to a further aspect, the invention relates to a driver assistance system for a vehicle that is designed to select an at least partially automated driving process from a plurality of possible driving processes. The driver assistance system is designed to perform the following steps: a) Determining the current driving scenario and several different possible future driving processes based on the current driving scenario by the driver assistance system; b) Automated creation of a decision tree by means of a computing unit of the driver assistance system by determining a plurality of nodes of the decision tree, wherein the nodes are linked to points in time in the future, wherein the nodes are at least partially assigned at least one decision to be made by the driver of the vehicle or by an external actor regarding a lane change or a traffic action depending on the driving scenario, and wherein the nodes are at least partially assigned at least two decisions or options for action to be made by the driver assistance system itself; c) calculating travel trajectories using an automated trajectory planner, wherein each travel trajectory is associated with a defined path through the decision tree and each path comprises one or more decisions to be made successively; d) Automated calculation of an evaluation indicator for the travel trajectories, wherein the evaluation indicator is a measure of the comfort and / or safety of the respective travel trajectory; (e) calculating probability values, wherein the probability values ​​are assigned to nodes of the decision tree and the probability values ​​each indicate the probability with which the driver of the vehicle or an external actor makes a decision to change lanes or perform a traffic action at the time assigned to the node; f) calculating a plurality of decision indicators, each decision indicator being assigned to a node, the calculation of the decision indicator being based at least in part on - the probability value of the respective node, which indicates that the driver of the vehicle or a third party will make a decision to change lanes or perform a traffic action at the time associated with the node; and - at least one evaluation indicator of a travel trajectory determined by a decision to be taken at the time of each node; g) selecting an at least partially automated driving process by the autonomous driving assistance system based on the decision tree, the decision indicators and depending on whether the driver of the vehicle or the external actor initiates a lane change or a traffic action at the time assigned to the node or not, wherein the values ​​of the decision indicators of the decision levels of the decision structure are compared to select an at least partially automated driving process and wherein the selected driving process is based on a driving trajectory that initiates a lane change or a deceleration process at the time assigned to the decision level with the smallest decision indicator value.

[0035] According to yet another aspect, the invention relates to a vehicle with a driver assistance system according to one of the previously described embodiments.

[0036] A "decision tree" in the sense of the present invention is a decision structure with multiple decision levels (also called nodes). The decision levels are assigned to different points in the future. Decisions must be made at these decision levels, either by the driver assistance system itself or by the driver of the vehicle itself, or by the driver or driver assistance system of the other vehicle.

[0037] For the purposes of the present invention, "external actor" refers to any actor that can perform an action that can influence the driving processes of the ego vehicle. This includes, for example, an external vehicle, a bus approaching from a stop, a pedestrian likely to enter the roadway, a cyclist possibly transitioning from the bike path to the roadway, a car possibly turning from a side street at an intersection, but also a traffic light or a barrier.

[0038] A "traffic action" within the meaning of the present invention refers to any action by the driver of the ego vehicle or an external actor whose occurrence limits the options for action or the decision-making freedom of the driver assistance system. For example, a person stepping onto the roadway limits the options for action of the driver assistance system to the extent that a "lane-staying at a constant speed" driving trajectory is no longer possible, and instead the system attempts to either change lanes or brake.

[0039] The terms “approximately”, “essentially” or “about” mean, in the sense of the invention, deviations from the exact value by + / - 10%, preferably by + / - 5% and / or deviations in the form of changes that are insignificant for the function.

[0040] Further developments, advantages, and possible applications of the invention will become apparent from the following description of exemplary embodiments and from the figures. All described and / or illustrated features, individually or in any combination, are fundamentally part of the invention, regardless of their summary in the claims or their reference back to them. The content of the claims is also incorporated into the description.

[0041] The invention is explained in more detail below with reference to exemplary embodiments and the figures. They show: Fig. 1 shows, by way of example and schematically, a vehicle in a driving scenario in which a decision has to be made regarding a lane change or lane keeping with speed reduction; Fig. 2 shows an example of a decision tree that allows a selection of several possible driving processes for the Fig. 1 enables the driving scenario shown; Fig. 3 shows an example of a schematic representation of the driving processes according to the decision tree of the Fig. 2; Fig. 4 example of the decision tree according to Fig. 2 with probability values ​​assigned to the decisions; Fig. 5 example of the decision tree according to Fig. 4 with evaluation indicators for the travel trajectories and decision indicators assigned to the nodes; Fig. 6 shows, by way of example and schematically, a block diagram of a driver assistance system for carrying out the selection procedure; and Fig. 7 shows, by way of example and schematically, a flow diagram illustrating the steps of the method for selecting an at least partially automated driving process.

[0042] Fig. 1 shows an example of a driving scenario in which a vehicle 1, also referred to as ego vehicle, is driving on a multi-lane road on which one or more other vehicles 2, 2' are also driving.

[0043] The driving scenario is characterized in particular by the fact that in front of vehicle 1 in the same lane there is another vehicle 2' which, however, has a lower speed, so that the distance d between the vehicles 1, 2' is increasingly reduced.

[0044] The driving situation may also be characterized by another vehicle 2 approaching from behind in a left lane of the road.

[0045] In such a driving scenario, several driving actions are conceivable. First, vehicle 1 can initiate a lane change, as indicated by the arrow, or reduce its speed without changing lanes and continue to follow the other vehicle 2 traveling in front of it.

[0046] These two basic driving processes can be initiated at different times. For example, at a large distance d from the other vehicle 2', the speed of vehicle 1 can be reduced slightly in order to slowly approach the other vehicle 2' in front. Alternatively, the deceleration process can be initiated later, which requires a greater deceleration. Similarly, the lane change of vehicle 1 can be initiated at different times and thus at different distances d from the other vehicle 2'.

[0047] In vehicles with a driver assistance system that complies with Autonomy Level 2 (SAE Level 2), a lane change must be initiated by the driver. This means that the autonomous driver assistance system cannot initiate a lane change independently. It can only do so after the driver has initiated it, for example, by activating the turn signal or by intervening in the steering wheel. After this initiation, the lane change can then be performed automatically.

[0048] The method described below enables, depending on the probability that the driver will initiate a lane change at a given time, a driving maneuver that is as comfortable as possible and adapted to the natural driving behavior of a human driver. This means, in particular, that driving maneuvers that require abrupt braking or an abrupt lane change can be effectively avoided.

[0049] Fig. 2 shows a decision tree 3, based on which a selection of an at least partially automated driving process can be made.

[0050] The decision tree 3 comprises a trunk 3a, along which several nodes 3.1 are provided. Branches 3b branch off from the nodes 3.1. Each branch 3b illustrates a decision 3.2. The branches 3b of the tree each end in a so-called leaf, which corresponds to a travel trajectory 3.3. In other words, making one or more decisions 3.2 leads to a certain travel trajectory 3.3. The travel trajectory is automatically calculated by a trajectory planner and specifies the path along which the vehicle 1 will automatically move if the decisions 3.2 leading to this travel trajectory 3.3 are made consecutively.

[0051] Nodes 3.1 are each linked to times t0 to t3, with the times t0 to t3 following chronologically from the current time ("now"). The root 3a of decision tree 3, with its nodes 3.1, thus forms a time axis. The times t0 to t3 linked to node 3.1 indicate which decisions linked to node 3.1 can be made at which time.

[0052] Decision tree 3 thus maps the decisions 3.2 to be made at different times, along with the resulting driving processes and driving trajectories 3.3. The entire decision tree can, for example, cover a future period of 5 to 15 seconds; the time intervals between two nodes can, for example, be between 1 and 5 seconds, preferably 1, 2, or 3 seconds.

[0053] In the decision tree 3 according to Fig. 2 For example, at time t0 the decisions "change lane and maintain speed", "keep lane and maintain speed" and "keep lane and reduce speed" are possible.

[0054] For example, some of the decisions 3.2 to be made at a node must be initiated by the driver themselves. Other decisions 3.2 can be initiated by the autonomous driving assistance system itself. For example, the decision 3.2 "lane change" must be initiated by the driver of vehicle 1, whereas the decisions 3.2 that involve maintaining the lane can be initiated by the autonomous driving assistance system itself, meaning no special driver initiative, such as activating the turn signal, is required.

[0055] Fig. 3 illustrates the decision tree according to Fig. 2 graphically illustrates the possible driving processes. If vehicle 1 moves toward another vehicle 2', the lane change can be initiated at different times and thus at different distances between the vehicles. For example, the driving trajectories Trj1, Trj3, Trj5 correspond to the lane change trajectories if the lane change is initiated at times t0, t1, or t2, respectively.

[0056] It should be noted that a driving trajectory does not refer to a single decision, but rather corresponds to a complete path through the decision tree 3 from "Now" to the respective leaf, thus encompassing a chain of several consecutive decisions 3.2. For example, the use of the driving trajectory Trj5 requires that the decision "Keep lane and maintain speed" was made at times t0 and t1, and that the driver initiated the lane change at time t2.

[0057] The driving trajectories Trj2, Trj4, Trj6 refer to the fact that no lane change is initiated by the driver and the deceleration is initiated at different times t0 to t2.

[0058] If the driver does not initiate a lane change and does not reduce the driving speed beforehand, an emergency braking maneuver is finally initiated automatically by the emergency braking assistant of vehicle 1.

[0059] The following describes how, depending on the probability that the driver initiates a lane change in the respective driving situation, a selection of an at least partially automated driving process from a multitude of possible driving processes is made possible, which improves comfort and safety.

[0060] Fig. 4 shows the decision tree 3 according to Fig. 2, with additional probability values ​​attached to branches 3b of decision tree 3. In particular, branches 3b, which refer to decisions 3.2 concerning a driver-initiated lane change, are assigned the probability values ​​p0, p1, p2. The probability values ​​are each linked to a node and thus indicate the probability p0 to p2 with which a driver makes a lane change decision at a time t0 to t2.

[0061] Thus, in the embodiment shown, the driver assistance system can make one of the two other decisions at the respective times t0 to t2 with the probability 1-p0, 1-p1, 1-p2, ie keep the lane and either decelerate or keep the speed constant.

[0062] The probabilities p0 to p2 that the driver initiates a lane change are calculated or estimated in advance for all nodes 3.1 or times t0 to t2.

[0063] The probabilities p0 to p2 for a lane change at times t0 to t2 can be calculated in various ways. Some examples of calculating the probabilities are given below. It is understood that other methods for calculating the probability also exist.

[0064] For example, information from the respective driver's driving history, i.e., in particular, from past lane-changing events, can be used to determine the probability of a lane change at a specific time. In particular, this information from the respective driver's driving history can be used as training data for a machine-learning system, in particular a neural network, so that it can recognize statistical relationships regarding the time at which a driver initiates a lane change in a given driving scenario. After training, the machine-learning system can be used in the current driving scenario to calculate the probability of a lane change at the respective future times.

[0065] According to a further alternative, evaluation indicators assigned to the respective driving trajectories are used to estimate the probabilities p0 to p2 for a lane change at times t0 to t2. The evaluation indicator assigned to a driving trajectory indicates the comfort (e.g., longitudinal and lateral acceleration below a predefined threshold, longitudinal and lateral jerk below a predefined threshold, lane change frequency) and the safety (e.g., distance to other vehicles and / or objects in the surrounding area) of the driving trajectory. The calculation of the probabilities p0 to p2 for a lane change at times t0 to t2 based on the evaluation indicators is carried out against the background that a human driver typically performs a lane change when this can be done with a high level of comfort and a high level of safety.Thus, the probability of changing lanes will most likely be highest at the time when a driving trajectory with the highest comfort or safety is available.

[0066] According to another alternative, other available information can be used to calculate lane change probabilities. This information could include, for example, whether: - the indicator is on; - a driver camera can detect that the driver is looking in the side mirror which is directed towards the target lane; - it can be determined by means of environment detection means (ie one or more sensors) that the target lane is clear; - it can be determined by means of a driver camera that the driver is looking over his shoulder in the direction of the target lane; - the navigation system suggests a route for which a lane change is advantageous in order to reach the navigation destination.

[0067] According to another alternative, other available information can be used to calculate lane change probabilities. This information could include, for example, whether: - a driver monitoring camera shows that the driver is only looking at the current lane for a long period of time; - a driver monitoring camera shows that the driver is looking only at an obstacle in the current lane for an extended period of time; - the navigation system suggests a route for which staying in the lane is advantageous in order to reach the navigation destination.

[0068] A combination of these examples to calculate the probabilities p0 to p2 for a lane change at times t0 to t2 is also possible.

[0069] As previously explained, a driving trajectory is calculated, at least in part, for decisions to be made according to decision tree 3. The driving trajectory specifies, on the one hand, the vehicle's spatial trajectory, for example, based on two-dimensional coordinates. On the other hand, information such as speed in the x- and y-directions, longitudinal and lateral acceleration, and longitudinal and lateral jerk are also assigned to the driving trajectory. This makes it possible to determine how comfortable the trajectory is for the driver.

[0070] In addition, additional information can be determined when calculating the driving trajectory, such as the distance the vehicle maintains from other vehicles or objects in the surrounding area.

[0071] It should be noted that the calculated driving trajectories must be collision-free in order to be considered in decision tree 3.

[0072] The planning of the driving trajectory can be carried out by a trajectory planner. This can calculate trajectories for various driving maneuvers. The driving maneuvers can include, in particular: - Stay in lane; - changing lanes; - keeping the lane with delay; - Maintain lane without delay for x seconds followed by deceleration, where x is an integer.

[0073] In addition, a unit is provided that enables an assessment of the planned driving trajectories in terms of comfort and safety. Preferably, an evaluation indicator assigned to the respective driving trajectory is calculated, indicating the comfort and safety of the driving trajectory in the form of a number. This evaluation indicator can be used to determine the driver's impression of comfort and safety when the vehicle travels the respective driving trajectory.

[0074] For example, a first driving trajectory in which the driving trajectory is driven with low longitudinal and lateral acceleration has a lower evaluation indicator than a second driving trajectory in which high longitudinal and lateral acceleration occurs.

[0075] Likewise, a driving trajectory in which the vehicle is decelerated slowly will have a lower evaluation indicator than a driving trajectory in which the vehicle is decelerated abruptly.

[0076] In other words, the evaluation indicator of a driving trajectory is a measure of how a human driver perceives a driving trajectory in terms of comfort and safety.

[0077] Fig. 5 shows the decision tree according to Fig. 4, where the individual trajectories Trj1 to Trj6 are each assigned an evaluation indicator C sw0 , C sw1 , C sw2 , C sh0 , C sh1 , C sh2 The subscript suffix "sw" stands for lane change and the subscript suffix "sh" for lane keeping. The assessment indicator C sw0 thus forms, for example, the evaluation indicator for the trajectory Trj1, the evaluation indicator C sh0for example, the evaluation indicator for the trajectory Trj2, etc.

[0078] Based on the probabilities (p0 to p2) for a lane change and the evaluation indicators of the driving trajectories, decision indicators can be calculated. Each decision indicator is preferably assigned to a node 3.1. Fig. 5, the decision indicator of the first node is labeled Cost_0, the decision indicator of the second node is labeled Cost_1 and the decision indicator of the third node is labeled Cost_2.

[0079] Based on a comparison of the decision indicators of the different nodes 3.1, the driver assistance system can decide whether, in the event that the driver does not initiate a lane change at a particular node, it is more advantageous, in terms of the driver's perception of comfort and safety, to stay in the lane and initiate a deceleration or to stay in the lane without deceleration.

[0080] The calculation of the decision indicators Cost_x of the different nodes 3.1 (x stands for the node or time index) can be done, for example, as follows: Cost_x=px*Cswx+(1−px)*Cshx

[0081] This includes: Cost_x decision indicator of the respective node; p x Probability of a lane change initiated by the driver at time tx C swx Evaluation indicator of the lane change trajectory at node x C shxEvaluation indicator of the lane keeping trajectory at node x

[0082] In addition, the emergency braking option is assigned the highest possible evaluation indicator or decision indicator value C NB (NB: emergency braking) assigned.

[0083] After a decision indicator Cost_x has been calculated for all nodes 3.1, the node with the lowest decision indicator value is first determined. For example, if node 0 has a decision indicator Cost_0 of 40, node 1 has a decision indicator Cost_1 of 30, node 2 has a decision indicator Cost_2 of 50, and node 3 has a decision indicator C NB of 100, node 1 is selected because it has the lowest decision indicator value.

[0084] The lowest decision indicator value means that, taking into account the probability that the driver initiates a lane change at the node to which this smallest decision indicator value is assigned, the costs of the decisions or the costs of the driving trajectories associated with these decisions are the lowest.

[0085] Alternatively, the lowest decision indicator value may mean that the expected cost at this node is lowest. This may be because all leaves of this node's branches have low-cost trajectories, or the lower-cost trajectory has a high probability compared to the high-cost trajectory.

[0086] Depending on the driver's driving behavior and the decision indicators of nodes 3.1, decision tree 3 is then traversed. For example, at node 0 (time t0), which has a higher decision indicator value than node 1 (time t1), if the driver does not initiate a lane change, the decision tree is traversed further to node 1, i.e., the trajectory Trj2 is not traversed, since optimized driving maneuvers with regard to comfort and safety are available at node 1.

[0087] If the driver does not initiate a lane change at node 1 either, the decision is made based on the minimum decision indicator value that the automated driver assistance system initiates the trajectory Trj4 at time t1, ie the lane is maintained and the speed is reduced.

[0088] In other words, if the driver does not initiate a lane change at time t0 and thus the trajectory Trj1 is traversed, at time t1 either the trajectory Trj3 is initiated, namely when the driver initiates a lane change, or the trajectory Trj4 is initiated, ie the lane is maintained and the speed is reduced.

[0089] The method described above for selecting an at least partially automated driving process can be modified if necessary, as described below: As previously described, the decision indicators (Cost_0 to Cost_2) assigned to node 3.1 are influenced by the probability values ​​(p0 to p2) that a driver will initiate a lane change at a specific time associated with node 3.1. These probability values ​​are subject to inaccuracy. The inaccuracy of the probability values ​​increases the further in the future the node or the time to which the probability value is assigned lies.

[0090] It is therefore proposed to determine the decision indicators for different probability deviations from the originally calculated probability values, for example, for a first scenario with no deviation, a second scenario with a deviation of +20%, a third scenario with a deviation of -20%, a fourth scenario with a deviation of +30%, and a fourth scenario with a deviation of -30%. The decision indicators of the nodes can then be calculated for all of these scenarios, and the decision indicator with the largest value can be selected for each node. The selection of the at least partially automated driving processes is then made based on these decision indicators with the largest value for the nodes.

[0091] According to a further embodiment of the invention, when traversing the decision tree 3, a decision can be made at the nodes as to whether trajectory planning should be performed again based on the current environmental information. Therefore, the decision tree 3 should be regenerated and the possible trajectories, probabilities, evaluation indicators of the driving trajectories, and decision indicators should be calculated. As a condition, collision-free trajectories should be available for the respective decisions at at least two future points in time / nodes.

[0092] The invention has been described above using an exemplary embodiment in which the vehicle changes lanes to a left lane to initiate an overtaking maneuver. It is understood that the invention is also applicable to other driving scenarios in which the vehicle changes lanes to a right lane. It is also conceivable to apply the invention to driving scenarios in which changing lanes to the left and right is possible.

[0093] Furthermore, the invention can also be applied to driving scenarios in which an automated driving process is carried out depending on whether or not an external vehicle is expected to merge into the driver's lane. In other words, the probability of initiating a lane change by an external vehicle is not predicted by the driver of the driver's own vehicle (ego vehicle).

[0094] Thus, the probability values ​​associated with the nodes of the decision tree indicate how likely it is that the other vehicle will initiate a lane change into the lane of the own vehicle. The decision tree can be used to decide whether to initiate a driving maneuver when the other vehicle performs a lane change or, if the other vehicle does not perform a lane change, to decide whether it is better to prepare for the other vehicle's lane change by reducing speed moderately or to maintain speed and initiate a harder braking maneuver later when the lane change occurs.

[0095] Fig. 6 shows a schematic representation of a driver assistance system for carrying out at least partially automated driving operations.

[0096] The driver assistance system preferably has a central control unit 11, which provides computing and storage means for carrying out the processes described below. Alternatively, a decentralized computer architecture can be provided, or the processes can be carried out or calculated on a computer unit of a sensor, provided the sensor provides sufficient computing power.

[0097] The control unit 11 has an environmental model unit 12, in which the environmental model of the vehicle's surroundings is provided. This environmental model unit 12 receives information from one or preferably several sensors 13. The sensors 13 can be, for example, cameras, radar sensors, LIDAR sensors, ultrasonic sensors, etc. The environmental model unit 12 generates an environmental model from the information from the sensors 13, which is provided to a driving process planner 14. The driving process planner 14 is a planner that can calculate one or more driving trajectories based on the environmental model, preferably depending on the current driving scenario, several driving trajectories that are applicable in this driving scenario and can be carried out without collisions.

[0098] Preferably, the driving process planner 14 also calculates the evaluation indicator for the respective driving trajectory. The previously described method for selecting a driving process can also be carried out in the driving process planner 14.

[0099] Once a travel trajectory along which the vehicle is to be moved has been determined, corresponding information is then transmitted to a vehicle motion controller 15. This vehicle motion controller 15 can either be provided in the control unit 11 or be provided decentrally, i.e., in a separate control unit. The vehicle motion controller 15 generates output signals for the steering, braking system, and drivetrain.

[0100] Fig. 7 schematically shows a flow diagram of a method for selecting an at least partially automated driving process by means of a driver assistance system of a vehicle.

[0101] First, the current driving scenario and several different possible future driving processes based on the current driving scenario are determined by the driver assistance system (S10).

[0102] A decision tree is then created automatically (S11). The decision tree comprises several nodes, each linked to a point in time in the future. Depending on the driving scenario, at least one decision regarding a lane change to be made by the driver of the vehicle or another vehicle is assigned to each node, at least in part. In addition, at least two decisions to be made by the driver assistance system itself are assigned to each node, at least in part.

[0103] In addition, driving trajectories are calculated, whereby each driving trajectory is assigned to a defined path through the decision tree and each path includes one or more decisions to be made sequentially (S12).

[0104] Subsequently, evaluation indicators for the driving trajectories are calculated, whereby the evaluation indicator is a measure of the comfort and / or safety of the respective driving trajectory (S13).

[0105] In addition, probability values ​​are calculated, whereby the probability values ​​are assigned to nodes of the decision tree and the probability values ​​each indicate the probability with which the driver of the vehicle or of a third-party vehicle makes a decision to change lanes at the time assigned to the node (S14).

[0106] Subsequently, several decision indicators are calculated (S15). Each decision indicator is assigned to a node. The calculation of the decision indicator is based at least in part on the probability value assigned to the respective node, which indicates that the driver of the vehicle or another vehicle will make a lane change decision at the time assigned to the node. Furthermore, the calculation of the decision indicator is based on at least one evaluation indicator of a driving trajectory, which is determined by a decision to be made at the time of the respective node.

[0107] Finally, an at least partially automated driving process is selected by the autonomous driving assistance system (S16) based on the decision tree, the decision indicators and depending on whether the driver of the vehicle or the other vehicle initiates a lane change at the time point assigned to the node or not.

[0108] The invention has been described above using exemplary embodiments. It is understood that numerous changes and modifications are possible without departing from the scope of protection defined by the patent claims. List of reference symbols 1 vehicle 2, 2' third-party vehicle 3 Decision tree 3a tribe 3b Branch 3.1 Nodes 3.2 Decision 3.3 Driving trajectory 10 Driver assistance system 11 Control unit 12 Environmental model unit 13 Sensor 14 driving process planner 15 Vehicle motion control d distance

Claims

[1] Method for selecting an at least partially automated driving process by means of a driver assistance system of a vehicle (1) comprising the following steps: a) determining the current driving scenario and several different possible future driving processes based on the current driving scenario by the driver assistance system (S10); b) Automated creation of a decision structure (3) by means of a computing unit of the driver assistance system by determining a plurality of decision levels (3.1) of the decision structure (3), wherein the decision levels (3.1) are linked to points in time in the future, wherein at least one decision (3.2) to be made by the driver of the vehicle (1) or an external actor (2) regarding a lane change or a traffic action is at least partially assigned to the decision levels (3.1) depending on the driving scenario, and wherein at least two decisions (3.2) to be made by the driver assistance system itself are at least partially assigned to the decision levels (3.1) (S11); c) calculating travel trajectories (3.3), each travel trajectory being associated with a defined path through the decision structure (3) and each path comprising one or more decisions (3.2) to be made successively (S12); d) calculating evaluation indicators for the travel trajectories (3.3), wherein the evaluation indicator is a measure of the comfort and / or safety of the respective travel trajectory (3.3) (S13); e) calculating probability values, wherein the probability values are assigned to decision levels (3.1) of the decision structure (3) and the probability values each indicate the probability with which the driver of the vehicle (1) or an external actor (2) makes a decision (3.2) to change lanes or perform a traffic action at the time point assigned to the decision level (3.1) (S14); f) Calculating a plurality of decision indicators (S15), each decision indicator being assigned to a decision level (3.1), the calculation of the decision indicator being based at least in part on - the probability value of the respective decision stage (3.1), which indicates that the driver of the vehicle (1) or an external actor (2) makes a decision (3.2) to change lanes or perform a traffic action at the time associated with the decision stage (3.1); and - at least one evaluation indicator of a travel trajectory (3.3) which is determined by a decision to be taken at the time of the respective decision stage (3.1); g) selecting an at least partially automated driving process by the autonomous driving assistance system based on the decision structure (3), the decision indicators and depending on whether the driver of the vehicle (1) or the external actor (2) initiates a lane change or a traffic action at the time assigned to the decision stage (3.1) or not (S16), wherein in order to select an at least partially automated driving process, the values of the decision indicators of the decision stages (3.1) of the decision structure (3) are compared and wherein the selected driving process is based on a driving trajectory (3.3) which initiates a lane change or a deceleration process at the time assigned to the decision stage (3.1) with the smallest decision indicator value. [2] Method according to claim 1, characterized bythat the decision indicators of the decision stages (3.1) are calculated at least partially by adding the results of a first multiplication and a second multiplication, wherein in the first multiplication the probability value that the driver of the vehicle (1) or the external actor (2) initiates a lane change at the time assigned to the decision stage (3.1) is multiplied by the evaluation indicator of the driving trajectory by means of which the lane change is carried out, and in the second multiplication the probability value that the driver of the vehicle (1) or the external actor (2) does not initiate a lane change or traffic action at the time assigned to the decision stage (3.1) is multiplied by the evaluation indicator of the driving trajectory by means of which the lane is kept and the vehicle (1) is decelerated. [3] Method according to one of the preceding claims, characterized bythat a decision (3.2) to be made by the driver of the vehicle (1) is detected by setting the indicator and / or a steering intervention that causes a lane change. [4] Method according to one of the preceding claims, characterized by that a decision to be made by the external actor (2) is detected by detecting an indicator, by turning and / or changing lanes of an external vehicle or by a person (2) entering the lane. [5] Method according to one of the preceding claims, characterized by that the calculation of a travel trajectory (3.3) comprises the calculation of multi-dimensional coordinates along which the vehicle (1) moves when traveling along the travel trajectory, the calculation of longitudinal and lateral acceleration values and / or the calculation of longitudinal and lateral jerk values. [6] Method according to one of the preceding claims, characterized bythat the calculation of the evaluation indicator of a driving trajectory (3.1) takes into account the distance of the vehicle (1) to other vehicles (2, 2') and / or objects, the difference to a set speed and / or the lane change frequency. [7] Method according to one of the preceding claims, characterized by that a driving trajectory (3.3) corresponds to a path through the decision structure (3), wherein the driving trajectory (3.3) is based on decisions (3.2) that are assigned to the path of the decision structure (3). [8] Method according to one of the preceding claims, characterized by that the probability value that the driver initiates a lane change at a future time associated with a decision stage (3.1) is calculated based on a neural network trained using data obtained from the driver's past driving actions. [9] Method according to one of the preceding claims, characterized by that the probability value that the driver initiates a lane change at a future time associated with a decision stage (3.1) is calculated based on the evaluation indicators of the driving trajectories (3.3) that are initiated by decisions at the times of the respective nodes (3.1). [10] Method according to one of the preceding claims, characterized by that the probability value that the driver initiates a lane change at a future time associated with a decision stage (3.1) is calculated based on driver behavior detected by at least one sensor. [11] Method according to one of the preceding claims, characterized bythat at least for some of the decision stages (3.1) a plurality of decision indicators are calculated which are based on different probability values that the driver of the vehicle (1) or the external actor (2) makes a decision to change lanes or perform a traffic action at the time associated with the decision stage (3.1), and that for each decision stage (3.1) the largest decision indicator is used for selecting an at least partially automated driving process. [12] Method according to one of the preceding claims, characterized by that the decision to be taken by the driver of the vehicle (1) involves a lane change from a right-hand lane to a left-hand lane or a lane change from a centre lane to a left or right-hand lane. [13] Driver assistance system for a vehicle (1) which is designed to select an at least partially automated driving process from a plurality of possible driving processes, wherein the driver assistance system is designed to carry out the following steps: a) Determining the current driving scenario and several different possible future driving processes based on the current driving scenario by the driver assistance system; b) Automated creation of a decision structure (3) by means of a computing unit of the driver assistance system by determining a plurality of decision levels (3.1) of the decision structure (3), wherein the decision levels (3.1) are linked to points in time in the future, wherein at least one decision (3.2) to be made by the driver of the vehicle (1) or an external actor (2) regarding a lane change or a traffic action is at least partially assigned to the decision levels (3.1) depending on the driving scenario, and wherein at least two decisions (3.2) to be made by the driver assistance system itself are at least partially assigned to the decision levels (3.1); c) calculating travel trajectories (3.3) by means of an automated trajectory planner, wherein each travel trajectory is assigned to a defined path through the decision structure (3) and each path comprises one or more decisions (3.2) to be made one after the other; d) Automated calculation of an evaluation indicator for the travel trajectories (3.3), wherein the evaluation indicator is a measure of the comfort and / or safety of the respective travel trajectory (3.3); e) calculating probability values, wherein the probability values are assigned to decision levels (3.1) of the decision structure (3) and the probability values each indicate the probability with which the driver of the vehicle (1) or an external actor (2) makes a decision to change lanes or perform a traffic action at the time assigned to the decision level (3.1); f) Calculating a plurality of decision indicators, each decision indicator being assigned to a decision level (3.1), the calculation of the decision indicator being based at least partly on - the probability value of the respective decision stage (3.1), which indicates that the driver of the vehicle (1) or an external actor (2) makes a decision (3.2) to change lanes or perform a traffic action at the time associated with the decision stage (3.1); and - at least one evaluation indicator of a driving trajectory (3.3) which is determined by a decision (3.2) to be taken at the time of the respective decision stage (3.1); g) selecting an at least partially automated driving process by the autonomous driving assistance system based on the decision structure (3), the decision indicators and depending on whether the driver of the vehicle (1) or the external actor (2) initiates a lane change or a traffic action at the time assigned to the decision stage (3.1) or not, wherein the values of the decision indicators of the decision stages (3.1) of the decision structure (3) are compared to select an at least partially automated driving process and wherein the selected driving process is based on a driving trajectory (3.3) which initiates a lane change or a deceleration process at the time assigned to the decision stage (3.1) with the smallest decision indicator value. [14] Vehicle comprising a driver assistance system according to claim 13.

Citation Information

Patent Citations

  • Method for operating a longitudinal guidance system in a motor vehicle

    DE102006001710A1

  • Method for avoiding collision of motor vehicle towards obstacle, involves producing warning for driver of vehicle optically, acoustically and / or haptically according to different escalation stages when time interval limits are reached

    DE102009020649A1

  • Predicting a vehicle's trajectory

    DE102013212359A1

  • Automatic distance control system for a vehicle

    DE102014209520A1

  • STEERING WHEEL ACTION

    DE102018107340A1

Cited By

  • Separate detection of lane change intention and lane change probability

    DE102024126108A1