GB2595856A

GB2595856A - A method for determining a trajectory of an autonomous motor vehicle, as well as an assistance system

Info

Publication number: GB2595856A
Application number: GB2008597.3A
Authority: GB
Inventors: Ulmer Benjamin; Trick Sebastian
Original assignee: Daimler AG
Current assignee: Mercedes Benz Group AG
Priority date: 2020-06-08
Filing date: 2020-06-08
Publication date: 2021-12-15
Also published as: GB202008597D0

Abstract

An assistance system 10 and method for determining a trajectory 20 of an autonomous motor vehicle 12. The method includes capturing the surroundings 16 of the vehicle 12 (e.g. using a sensor 14), and using an electronic computing device 18 to: -detect a critical object 22 (e.g. another vehicle) in the captured surroundings; -determine at least one critical corner 24, 26 of the object, -generate a motion model for a potential traffic participant 28 (e.g. a pedestrian) which is potentially occluded (i.e. hidden) by the critical corner; -determine a safe state map for the autonomous vehicle 12 depending on the motion model; and -determine the trajectory 20 of the autonomous vehicle 12 depending upon the safe state map. The trajectory may be a provably safe trajectory or a minimal risk trajectory. The critical object may be within a predetermined distance from the autonomous vehicle. The motion model may be determined considering characterising parameters of the traffic participant - e.g. legal requirements and/or maximum velocity of pedestrian or other vulnerable road user (VRU).

Description

A method for determining a trajectory of an autonomous motor vehicle, as well as an assistance system

FIELD OF THE INVENTION

[0001] The present disclosure relates to the field of automobiles. More specifically, the present disclosure relates to a method for determining a trajectory of an autonomous motor vehicle, as well as to a corresponding assistance system.

BACKGROUND INFORMATION

[0002] From the state of the art, it is known that automated vehicles are expected to provide human-like behavior and trajectories. However, because humans take risks while driving, a self driving vehicle needs to guarantee safety under a given set of rules and assumptions on the behavior of other traffic participants.

[0003] DE 10 2017 240 531 Al relates to a method for operating a motor vehicle in an automated driving mode, the method comprises receiving a status signal of a respective sensor assigned to the motor vehicle, wherein the status signal comprises information about a performance capability and/or about a functionality of the respective sensor. The method further comprises determining a characteristic value of a parameter which is representative of a performance capability and/or a functionality of a respective sensor as a function of a received status signal, and determining a safe driving range for the automated driving operation of the motor vehicle as a function of the determined characteristic value of a respective sensor. The method also comprises controlling a function of the automated travel operation of the motor vehicle as a function of the determined safe travel range.

[0004] This disclosure classifies safe driving areas for different automated levels, dependent on how many sensors have detected objects. Trajectories are then supposed to be planned within these safe driving areas. However, this disclosure considers only one visible space and makes assumptions on potentially occluded agents outside this visible space and their future impact on an ego vehicle.

[0005] It is an object of the invention to provide a method and an assistance system, by which a potentially occluded traffic participant may be taken into consideration by determining a trajectory of the autonomous motor vehicle.

[0006] This object is solved by a method and an assistance system according to the independent claims. Advantageous embodiments are indicated in the dependent claims.

SUMMARY OF THE INVENTION

[0007] One aspect of the invention relates to a method for determining a trajectory of an autonomous motor vehicle by an assistance system of the motor vehicle. Surroundings of the motor vehicle are captured by at least one capturing device of the assistance system. At least one critical object, wherein the critical object may be a occluding object, is detected in the captured surroundings by computing the captured surroundings by an electronic computing device of the assistance system. At least one critical corner of the at least one critical object is determined in the captured surroundings by the electronic computing device, wherein the critical corner is potentially occluding a potential traffic participant. A motion model for the potentially occluded potential traffic participant is generated by the electronic computing device depending on the at least one critical corner. A safe state map for the autonomous motor vehicle is determined depending on the generated motion model by the electronic computing device. The trajectory of the autonomous motor vehicle is determined depending on the determined safe state map by the electronic computing device.

[0008] For example, a guaranteed safe trajectory is selected out of a plurality of possible trajectories by discarding any trajectory options that violate the safe state map at any point in time. The original trajectory pool may be computed by any state of the art trajectory generation method. The existence of a guaranteed safe trajectory may be recursively ensured by including the trajectories used for the safe state map generation into the trajectory pool.

[0009] Thereby, it is facilitated that depending on the detected critical corner a potentially occluded potential traffic participant may be taken into consideration by determining the trajectory of the autonomous motor vehicle.

[0010] Therefore, an algorithm that may incorporate arbitrary assumptions on the behavior of the other traffic participants is provided in order to guarantee comfortable behavior of the autonomous motor vehicle without endangering other agents. The presented method also works for agents that have not been detected due to occlusions or sensor insufficiencies. While guaranteeing passive motion safety, the behavior of the autonomous motor vehicle is adapted in order to maximize driving comfort.

[0011] According to an embodiment, the trajectory is determined as a provably safe trajectory of the autonomous motor vehicle.

[0012] In another embodiment, the trajectory is determined as a minimal risk trajectory of the autonomous motor vehicle.

[0013] In another embodiment, the motion model of the potentially occluded traffic participant is generated in consideration of at least one characterizing parameter of the traffic participant.

[0014] In another embodiment, the critical object is detected in a predetermined distance to the autonomous motor vehicle. In particular, this is performed as the first step of the algorithm.

[0015] A further aspect of the invention relates to an assistance system for determining a trajectory of an autonomous motor vehicle, wherein the assistance system comprises at least one capturing device and one electronic computing device and is configured to perform a method according to the preceding aspect. In particular, the method is performed by the assistance system.

[0016] Another aspect of the invention relates to a motor vehicle comprising an assistance system according to the preceding aspect. The motor vehicle is in particular an autonomous motor vehicle.

[0017] Further advantageous embodiments of the method are to be regarded as advantageous embodiments of the assistance system as well as of the motor vehicle. The assistance system as well as the motor vehicle for this purpose comprise substantive features, which facilitate a performance of the method or advantageous embodiments thereof.

[0018] Further advantages, features, and details of the invention derive from the following description of preferred embodiments as well as from the drawings. The features and feature combinations previously mentioned in the description as well as the features and feature combinations mentioned in the following description of the figure and/or shown in the figure alone may be employed not only in the respectively indicated combination but also in any other combination or taken alone without leaving the scope of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The novel features and characteristics of the disclosure are set forth in the appended claims. The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate exemplary embodiments and together with the description, serve to explain the disclosed principles. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The same numbers are used throughout the figures to refer to identical features and components. Some embodiments of system and/or methods in accordance with embodiments of the present subject matter are now described below, by way of example only, and with reference to the accompanying figures.

[0020] The drawings show in: [0021] Fig. 1 is a schematic view of an embodiment of the assistance system; and [0022] Fig. 2 is another schematic top view of an embodiment of an autonomous motor vehicle comprising the assistance system.

[0023] In the figures same elements or elements having the same function are indicated by the same reference signs.

DETAILED DESCRIPTION

[0024] In the present document, the word "exemplary" is used herein to mean "serving as an example, instance, or illustration." Any embodiment or implementation of the present subject matter described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments.

[0025] While the disclosure is susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and will be described in detail below. It should be understood, however, that it is not intended to limit the disclosure to the particular forms disclosed, but on the contrary, the disclosure is to cover all modifications, equivalents, and alternatives falling within the scope of the disclosure.

[0026] The terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a setup, device or method that comprises a list of components or steps does not include only those components or steps but may include other components or steps not expressly listed or inherent to such setup or device or method. In other words, one or more elements in a system or apparatus preceded by "comprises" or "comprise" does not or do not, without more constraints, preclude the existence of other elements or additional elements in the system or method.

[0027] In the following detailed description of the embodiments of the disclosure reference is made to the accompanying drawings, which form a part hereof and in which are shown by way of illustration specific embodiments in which the disclosure may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the disclosure, and it is to be understood that other embodiments may be utilized and that changes may be made without departing from the scope of the present disclosure. The following description is, therefore, not to be taken in a limiting sense.

[0028] Fig. 1 shows a schematic view of an embodiment of an assistance system 10 of a motor vehicle 12 (Fig. 2). The assistance system 10 comprises at least one capturing device 14 (Fig. 2) for capturing the surroundings 16 (Fig. 2) of the motor vehicle 12. Furthermore, the assistance system 10 comprises at least one electronic computing device 18 (Fig. 2).

[0029] The method for determining a trajectory 20 (Fig. 2) of the motor vehicle 12 by the assistance system 10, wherein the motor vehicle 12 is an autonomous motor vehicle 12,comprises at least capturing the surroundings 16 of the motor vehicle 12 by the at least one capturing device 14 of the assistance system 10 in a first step Si. In a second step S2 detecting at least one critical object 22 (Fig. 2) in the captured surroundings 16 by computing the captured surroundings 16 by the electronic computing device 18 is performed. At least one critical corner 24, 26 (Fig. 2) of the at least one critical object 22 is determined by the electronic computing device 18 in a third step S3, wherein the critical corner 24, 26 is potentially occluding a potential traffic participant 28 in the captured surroundings 16. In a fourth step S4 generating a motion model for the potentially occluded potential traffic participant 28 depending on the at least one critical corner 24, 26 by the electronic computing device 18 is performed. Specifically, the motion model is a worst-case motion model for the potentially occluded potential traffic participant 28. In a fifth step 55 determining a safe state map 34 (Fig. 2) for the autonomous motor vehicle 12 is performed depending on the generated motion model by the electronic computing device 18. In a sixth step 56 determining the trajectory 20 of the autonomous motor vehicle 12 depending on the determined safe state map by the electronic computing device 18 is performed.

[0030] Fig. 1 further shows that the trajectory 20 is determined as a provably safe trajectory 20a of the autonomous motor vehicle 12, wherein computing and selecting trajectories complying with the safe state maps 34 is performed in a seventh step 57. Furthermore, the trajectory 20 is determined as a minimal risk trajectory 20b of the autonomous motor vehicle 12, wherein any kind of measuring of a distance D (Fig. 2) in order to quantify the controllability of the worst-case scenario may be used. Therefore, information from an Automotive Safety Integrity Level (ASIL) 30 may be used, for example, to describe the severity, the exposure, and the controllability in the given scenario.

[0031] Furthermore, it is shown that the motion model of the potentially occluded potential traffic participant 28 is generated in consideration of at least one characterizing parameter 32 of the traffic participant 28. In particular, the characterizing parameter 32 describes legal requirements and physical boundaries, for example it is assumed that a pedestrian may not move faster than a certain velocity, for the traffic participant 28 relating to situations that need to be handled, such as a maximum velocity of the traffic participant 28 and the route compliance of the traffic participant 28, for example, no pedestrian on express ways.

[0032] Fig. 2 shows in a schematic top view an embodiment of the motor vehicle 12 with the assistance system 10. According to this embodiment the occluded critical object 22 is another motor vehicle. In an occluded area 36, which is occluded from a first critical corner 24 and a second critical corner 26, the potentially occluded potential traffic participant 28 is shown.

[0033] The worst-case assumption of the traffic participant 28 may follow any model, given the legal requirements on what motion model and actions of the traffic participant 28 need to be considered when driving autonomously. The safe state maps 34 and ultimately the safe trajectory 20 are computed on the assumption that the autonomous motor vehicle 12 may perform a full braking maneuver without steering. For example, in a first step the motor vehicle 12 starts in a safe state by definition. In a second step, the safe state map 34 is generated based on physical braking capacities and the worst case scenario, for example given occlusion, and worst case assumptions on occluded agent trajectories. In a third step a trajectory 20 in compliance with the safe state map 34 is executed, which may be one that just stays in the same safe state, but preferable the trajectory 20 may be the fastest and most comfortable one. After the third step, the first step is performed again.

[0034] It is proposed to use any kind of a distance D measure, for example, a distance D of the trajectories 20 from the safe state map 34, in order to quantify the controllability of the computed trajectories 20. The provably safe trajectory 20 within an autonomous motor vehicle 12 is guaranteed to find the safe trajectory 20a for the potentially endangered traffic participant 28. The further the trajectory 20 deviates from the safe state, the smaller the controllability becomes. The controllability does not immediately collapse to zero when the safe state map 34 is violated since the safe states are overly conservative and do not consider evasive steering or "escaping" into neighboring lanes.

[0035] The vehicle is considered to be in a safe state if a trajectory 20 exists which ensures passive motion safety. Therefore the trajectory 20 will not be in collision at any time, except potentially at standstill. By always finding a transition from one safe state to another passive motion safety can be guaranteed at all times. How this is possible will be outlined in the following.

[0036] The state of the ego vehicle consists of an x and a y location in the world, as well as the current velocity at this position. Its rotation is aligned with the reference line, which is why it is not considered inside of the state. The current acceleration is not taken into account.

[0037] It should be appreciated that, a safe state always accounts for the worst-case scenario. Therefore it can guarantee safety for all possible scenarios. This is a very conservative assumption, but has the benefit of being provably safe.

[0038] The European Commission for Mobility and Transport defines Vulnerable Road Users (VRUs) as "non-motorised" road users, such as pedestrians and cyclists as well as motorcyclists and persons with disabilities or reduced mobility and orientation".

[0039] In the context of this invention, VRUs or the traffic participants 28, will be used for traffic participants 28 moving into traffic out of an occluded area. The main focus is particularly on pedestrians. Without loss of generality, it can be assumed that the occluded VRU has no physical expansion. Therefore it can be in even the smallest occluded space. As soon as it leaves the occluded space, its shape is over-approximated by a 2D bounding box, to check potential collisions with the vehicle. A VRU may be modeled with a starting acceleration, maximum reachable acceleration, and maximum velocity to define its reachable set. For a worst case scenario, however, only the maximum velocity is of importance since that creates the largest reachable set, and therefore, also the worst-case scenario.

[0040] Only considering driving forward, there may be three types of collisions. A frontal collision, a collision from the right side, or one from the left side. It can be differentiated further between a collision in motion and a collision at standstill. A collision at standstill implies that the ego vehicle is not moving at the time of impact. In order to compute a safe state, it needs to be determined which scenario would be the worst case.

[0041] Visible space is the area that is perceived by the sensor setup arranged on the ego vehicle. Everything else is considered occluded space. The occluded space may either be created by a static obstacle, a dynamic obstacle, or limited sensor range. A simplifying assumption is that a VRU would only step out of an occlusion created by a static object. Therefore, a VRU may not step out from behind a moving car. Visible space is dependent on the sensor setup of the ego vehicle. The assumption is that there is at least one sensor at the very front of the vehicle. As soon as the sensor passes an occluded corner that space becomes visible.

[0042] For example the propagation of an occlusion over three time steps may be considered. Obviously occlusions are time-dependent. However, the currently observed occlusion may be used as a worst-case occlusion, which is time-independent. Assuming the scenario at t = tO: in the next time step t = ti, the occlusion has moved. However, it only covers space that was previously visible or uncovers space that was occluded at t = tO. In particular, there is always a corner that creates the occlusion. That corner will also keep creating an occlusion over all time steps until the first sensor arranged on the ego vehicle has passed it. Those occlusion generating corners will be referred to as critical corners in the following.

[0043] Critical corners are also the point where the worst-case scenario originates, for example a pedestrian coming out of any point in the occlusion. In the same amount of given time the VRU can travel the same distance, assuming a worst-case maximum velocity. It is apparent that the worst case must originate at the point where the pedestrian has the ability to walk closest towards the approaching vehicle. That point is a critical corner since the origin of an occlusion is the closest point of the occlusion to the ego vehicle.

[0044] A reachable set for the VRU describes the area that a VRU can reach in a certain time span. In particular it is important to know the area that the VRU can travel in the time it takes the ego vehicle to come to a complete stop. Rather than modeling the whole reachable set, it is sufficient to only model a reachable set around the critical corner.

[0045] In order to compute the safe state map 34 simplifying but still worst case compliant assumptions on the braking maneuver are made.

[0046] The car applies full braking power (maximum deceleration) along the reference line until it comes to a full stop. The assumption is that maximum braking power is constant and known. While driving on a non-straight stretch of road acceleration is dependent on curvature and velocity. Here the assumption is that an over-approximation of the available deceleration that can be applied is always known.

[0047] Computing safe states refers to actually computing an upper bound velocity where a state s is still safe.

[0048] To check if a state is safe for an arbitrary path is rather simple. It takes tstop for ago to come to a full stop, applying maximum braking force along the reference line. Each reachable set assigned to a critical corner needs to be expanded based on tstop. As long as the stopped ego vehicle does not intersect with any of the reachable sets, the state is considered to be safe. So checking if a state is safe or not is straight forward. However, it is not trivial to find the maximum of the velocity dependent on x and y that is still considered part of the safe state. In the case of arbitrary paths there is no closed form solution and a solution with an error margin may be found by conducting a search in the velocity space.

[0049] However, this approach is not computationally effective. Therefore it is better to distinguish between two cases. The case where the upcoming reference line can be approximated by a straight line and the case where a curved stretch can be approximated as a curve with a fixed radius. Furthermore, it can be shown that in the straight line case a closed form solution exists and in the curved stretch case a very close over-approximation exists.

[0050] In tstop a potential VRU can travel a distance. With tstop the distance traveled by the ego vehicle may also be calculated. A collision will occur if the distance between the stopping point of the ego vehicle and the distance traveled by the VRU are equal. Solving for v will give four solutions, however, only one of them is the desired value. If substituting v square with u, there would be two solutions. One for the first intersection of the straight driving line with the reachable set, and one of the line leaving the reachable set again. Obviously only the first intersection is of relevance. It will be the solution with the lower v. The additional two solutions by going from u back to v are for potential negative velocities, which also are not a solution. Therefore there is only one valid solution.

[0051] To obtain a non-imaginary solution the discriminant has to always be positive. It takes VRU tvru = y to travel the distance y. So the ego vehicle needs to start braking at x to come to a full stop. Rearranged for y, this gives the distance the ego vehicle needs to drive away from the critical corner depending on x, where v is irrelevant. This is always smaller than or equal to the analytically given boundary condition.

[0052] The solution is safe as long as the query points are inside the boundary conditions. It needs to be distinguished between frontal and side collisions. A frontal collision should only be considered when the critical corner is directly in front of the car. In those cases the safe velocity is to be computed with v(x, 0). Side collisions may only happen if the query point at this point in time is outside the boundaries. Since the boundaries are known, it is easy to compute the missing values. If a query point is outside the boundaries, instead of computing for one of the front corners for the car, the rear end corner needs to be considered. This essentially just means that when "exiting" the well-defined states the current velocity has to be maintained for the entire car length.

[0053] So far system delay has not been considered in the computing of safe states. A human driver has a certain reaction time, and so does an automated system. It takes time to process all the sensor data, compute an appropriate behavior and generate trajectories 20. To account for this time a time delay tdelay is introduced. The reasoning to arrive at the solution stays essentially the same. In computing tdelay is added, because the VRU can travel for that additional time. The term v*tdelay is added, since the vehicle travels for the time until braking occurs at the velocity that it currently has. Again, potential acceleration or deceleration of the ego vehicle in that time span is not considered.

[0054] So far only the isolated case of one single critical corner has been computed. This may easily be generalized for multiple critical corners. For this purpose the maximum velocity for each corner is to be computed and the minimum results in the maximum velocity for that query state are to be used. To minimize computational cost, this raises the question of which critical corners are worth being considered in the first place.

[0055] In a real scenario there may be a plurality of occlusions, which increases computation time. Rather than always computing the full velocity value for each critical corner, it is enough to check if the corner is relevant to the current query point. It is not necessarily the closest corner which is the most relevant, but it is a good heuristic to start searching for the generating corner of the safe velocity. Therefore it is a good idea to start with a list of critical corners sorted according to their distance to the ego vehicle. Defining a bounding box for a critical corner is intuitive. The length has to be the stopping distance of the ego vehicle at a certain speed plus the distance a VRU can travel in that time. The width of the bounding box is limited by the distance the VRU travels.

[0056] There may be for example four critical corners and the query point. The four corners are just exemplarily. There may be for example more than for critical corners. As a first step it is assumed that the ego vehicle is not driving faster than the allowed speed limit. So the speed limit is used to calculate the bounds for a critical corner. The query point is inside the bounds, but the velocity for that point remains undefined. Again the point is inside the bounds, but computing an actual maximal velocity for that point is possible. So in the next query step the bounds for the critical corners are considerably smaller. The irrelevant corners may be left unconsidered. At first glance, this whole procedure may appear to be an unnecessary extra step, but the actual bound check is faster to compute.

[0057] Figs. 1 and 2 show a method for ensuring passive motion safety of an autonomous motor vehicle 12 while passing potentially occluded vulnerable road users.

Reference Signs assistance system 12 motor vehicle 14 capturing device 16 surroundings 18 electronic computing device trajectory 20a provable safe trajectory 20b minimal risk trajectory 22 critical object 24 first critical corner 26 second critical corner 28 traffic participant

ASIL

32 characterizing parameter 34 safe state map 36 occluded area Si first step 52 second step 53 third step S4 fourth step fifth step 56 sixth step 37 seventh step 38 eighth step

Claims

CLAIMSA method for determining a trajectory (20) of an autonomous motor vehicle (12) by an assistance system (10) of the motor vehicle (12), the method including the following steps: -Capturing a surroundings (16) of the motor vehicle (12) by at least one capturing device (14) of the assistance system (10); -detecting at least one critical object (22) in the captured surroundings (16) by computing the captured surroundings (16) by an electronic computing device (18) of the assistance system (10); - determining at least one critical corner (24, 26) of the at least one critical object (22) by the electronic computing device (18), wherein the critical corner (24, 26) is potentially occluding a potential traffic participant (28) in the captured surroundings (16); - generating a motion model for the potentially occluded potential traffic participant (28) depending on the at least one critical corner (24, 26) by the electronic computing device (18); -determining a safe state map (34) for the autonomous motor vehicle (12) depending on the generated motion model by the electronic computing device (18); and -determining the trajectory (20) of the autonomous motor vehicle (12) depending on the determined safe state map (34) by the electronic computing device (18).
The method according to claim 1, characterized in that the trajectory (20) is determined as a provably safe trajectory (20a) of the autonomous motor vehicle (12).
The method according to claim 1 or 2, characterized in that the trajectory (20) is determined as a minimal risk trajectory (20b) of the autonomous motor vehicle (12).
The method according to any one of claims 1 to 3, characterized in that the motion model of the potentially occluded potential traffic participant (28) is generated in consideration of at least one characterizing parameter (32) of the traffic participant (28).
The method according to any one of claims 1 to 4, characterized in that the critical object (22) is detected in a predetermined distance (D) to the autonomous motor vehicle (12) .
6. An assistance system (10) for determining a trajectory (20) of an autonomous motor vehicle (12), wherein the assistance system (10) comprises at least one capturing device (14) and one electronic computing device (18) and is configured to perform a method according to any of claims 1 to 5.