Methods for planning the behavior of a vehicle

Abstract

Method for planning the behavior of a vehicle (1) with respect to at least one concealed area (4), the method comprising: a phantom object generation step (S1) in which at least one phantom object (3) is generated for the concealed area (4), wherein the phantom object generation step (S1) comprises a calculation of an appearance probability for the phantom object (3), wherein the appearance probability describes a probability that the phantom object (3) will emerge from the concealed area (4) into a field of view of the vehicle (1), wherein the method comprises an appearance probability modification step (S2) in which a modified appearance probability is determined based on a message received from a V2X device (6) communicating with the vehicle (1) and on the basis of the appearance probability.

Description

Technical area

[0001] The invention relates to a method for planning the behavior of a vehicle according to the preamble of claim 1. The invention further relates to a system, a vehicle, a computer program and a computer-readable medium according to the dependent claims. Technical background

[0002] The field of autonomous driving, which typically comprises several levels of automated driving, such as assisted driving, semi-autonomous driving, or even fully autonomous driving, is a field subject to constant technological development. One reason for this is that the automated movement of a vehicle is typically a very complex task and entails a large number of technical challenges.

[0003] For example, adequate road safety must be ensured when automated vehicles participate in road traffic. On the other hand, a certain level of traffic flow must also be ensured: To avoid traffic jams, for example, blockades and other forms of standstill must be avoided.

[0004] In order to achieve an appropriate compromise between traffic safety on the one hand and traffic flow on the other, methods for automatic behavior planning of autonomous vehicles typically calculate probability values ​​that describe the probabilities for the occurrence of certain traffic scenarios and then base their planning at least partially on these probabilities.

[0005] However, the known methods and systems have several shortcomings. One problem is that it is still difficult for vehicles in the field of autonomous driving to deal with occlusion scenarios. An occlusion scenario is a traffic situation in which a certain part of a vehicle's surroundings is not visible because this part of the surroundings is obscured by an object. In other words, in an occlusion scenario, the field of view of a vehicle for which automated behavior is planned is obscured by an object. Such objects can be, for example, other vehicles such as cars, buses, or trucks, or stationary infrastructure elements such as bridges, walls, buildings, or the like. Disclosure of the invention

[0006] It is an object of the invention to overcome or at least reduce the above disadvantages.

[0007] The object is achieved by a method for planning the behavior of a vehicle with respect to at least one hidden area, wherein the method comprises a phantom object generation step in which at least one phantom object is generated for the hidden area, wherein the phantom object generation step comprises a calculation of an appearance probability for the phantom object, wherein the appearance probability describes a probability that the phantom object appears from the hidden area in a field of view of the vehicle, wherein the method comprises an appearance probability modification step in which a modified appearance probability is determined on the basis of a message received from a V2X device that communicates with the vehicle and on the basis of the appearance probability.

[0008] The method is typically a computer-implemented method. The vehicle is typically autonomous, meaning it is designed for at least one type of autonomous driving, such as assisted driving, semi-autonomous driving, or ultimately fully autonomous driving. The term "behavior" describes a reaction of the vehicle to an obscured area, e.g., deceleration when approaching and / or passing an object responsible for the obscured area, or a gradual forward creep when passing an object causing the obscured area.

[0009] A "phantom object" is a virtual object located within a hidden area. Examples of phantom objects include other vehicles or pedestrians. In the term "V2X device," the term "V2X" stands for "vehicle-to-everything." V2X devices are specifically designed for communication between vehicles (V2V), vehicles and infrastructure (V2I), vehicles and pedestrians (V2P), and vehicles and other entities such as cloud services (V2C). V2X technology enables vehicles to exchange information and data with their surroundings. V2X devices typically include on-board units (OBUs) installed in vehicles, roadside units (RSUs) positioned along roads, and various sensors and communication modules.These devices utilize wireless communication technologies such as Dedicated Short-Range Communication (DSRC) or Cellular Vehicle-to-Everything (C-V2X) to enable the exchange of information between vehicles and other entities in the V2X ecosystem. In typical embodiments, the V2X device includes a camera and / or is configured to interact with a camera, e.g., by receiving and / or analyzing and / or processing an image and / or video stream provided by a camera. These V2X devices are typically all part of a network of devices, with the vehicle whose behavior is to be planned (this vehicle is also referred to as the ego vehicle) also being part of this network. In typical embodiments, the method handles a plurality of occluded areas and a plurality of phantom objects.When the terms "a hidden" area or "a phantom object" are used in this specification and the following claims, this in principle also includes cases where there are multiple such objects. In specific embodiments, a modified appearance probability is determined for each phantom object. There may be cases where the modified appearance probability for a phantom object exactly corresponds to the appearance probability (in other words, where no modification of the original appearance probability occurs).

[0010] Using such a vehicle behavior planning method, it becomes possible to achieve a better compromise between traffic safety and traffic flow, since the fact that a message received from a V2X device is taken into account when estimating the probability of a phantom object's appearance makes the calculation of this probability more realistic. This was surprisingly discovered by the inventors.

[0011] In preferred embodiments, the method comprises a confidence modifier calculation process in which a confidence modifier to be used in the appearance probability modification step is preferably calculated for each phantom object, wherein the confidence modifier calculation process comprises: an assignment sub-process in which the phantom object, preferably all phantom objects, is / are assigned to a list of V2X devices that are / are capable of observing the respective phantom object, and / or a detection sub-process in which it is checked whether any real objects are present in the hidden area(s) of the respective phantom object(s), and / or a determination sub-process in which a list of confidence modifier candidates is determined, wherein the list preferably comprises a confidence modifier candidate for each V2X device present in the list of V2X devices,and / or a selection sub-process in which the confidence modifier is selected, preferably from the list of confidence modifier candidates.

[0012] The term “capable of observation” is typically understood to mean that the respective V2X device, which may be a camera, for example, can actually observe (in the case of a camera: see) the respective phantom object and / or the obscured area in which the phantom object is located at a specific time. In some embodiments, the list of V2X devices comprises only a single V2X device. In some embodiments, the list of confidence modifier candidates comprises only a single confidence modifier candidate. In typical embodiments, the method goes through the list of V2X devices during the detection sub-process and checks which of the V2X devices in the list of V2X devices are observing or seeing a real road user at a specific time.

[0013] In typical embodiments, the determination sub-process comprises calculating a maximum number of confidence steps for each V2X device present in the list of V2X devices, wherein a confidence step is typically a planning step of the method in which it is certain that no undetected real objects will enter the observation range of the respective V2X device.In typical embodiments, the determination sub-process comprises calculating a delay factor for each V2X device present in the list of V2X devices, wherein the delay factor typically describes the extent to which a message received by the respective V2X device should be trusted and / or wherein the delay factor typically has a value of 0 if a message is older than one second and / or wherein the delay factor typically has a value of less than 1 but more than 0 if a message is newer than one second. In typical embodiments, the determination sub-process comprises determining a trust factor for each V2X device present in the list of V2X devices, wherein the trust factor typically describes the extent to which the respective V2X device can be trusted, wherein each trust factor preferably has a value between 0.7 and 1.

[0014] In typical embodiments, each confidence modifier candidate is preferably calculated according to the following equation: pcm,kc:=pd,kc⋅ft,kc⋅fq,kc, if th <tt,kc.

[0015] In typical embodiments, if more than one confidence modifier candidate has been calculated, the confidence modifier is calculated according to the equation: pcm,k*:=f(Pcm,k):={max(|Pcm,k|),if ∀pcm,kc∈Pcm,k:pcm,kc<0max(Pcm,k),otherwise.

[0016] In typical embodiments, if only one confidence modifier candidate has been computed, the confidence modifier corresponds to the confidence modifier candidate.

[0017] Where c is the number of a specific V2X device, k is the number of a specific phantom object and pcm,kc a confidence modifier candidate for the V2X device with the number c and the phantom object with the number k. pd,kc a detection result indicating whether a real object is detected by the V2X device with the number c in the occluded area of ​​the phantom object with the number k. ft,kc a delay factor for the V2X device with the number c and the phantom object with the number k. fq,kc a confidence factor for the V2X device with the number c and the phantom object with the number k. Where t h a time step within a planning horizon of the process and tt,kc a maximum number of confidence steps for a V2X device with the number c for the particular phantom object with the number k. All confidence modifier candidates for a particular phantom object with the number k together form the confidence modifier candidate list P cm,kfor the corresponding phantom object with the number k. From this list, the confidence modifier is typically pcm,k* for each phantom object with the number k.

[0018] In typical embodiments, the modified appearance probability is calculated according to the following equation papp,k*={max(0, min(1, papp,k+pcm,k*)).if pcm,k* available,papp,k,otherwise.

[0019] This is papp,k* is the modified probability of appearance for the phantom object with the number k and is p app,k the appearance probability for the phantom object with number k before modification, which can also be called the initial appearance probability.

[0020] In typical embodiments, the method comprises a hidden region selection process in which the hidden region is selected from a list of types of hidden region, wherein the list of types of hidden region preferably comprises five types of hidden region. In typical embodiments, the types of hidden region comprise a high priority lane, an oncoming lane, a low priority lane, an ego navigation lane, and / or an occluded region, which may also be referred to as another hidden region. In typical embodiments, the another hidden region comprises or is a crosswalk or a bus stop.

[0021] In typical embodiments, the initial appearance probability has a static component, wherein the static component preferably takes into account map and / or road topology information, and / or wherein the static component preferably depends on an initial environment probability and / or a phantom object distance and / or a distance threshold. The initial environment probability is typically a value between 0 and 1. The "phantom object distance" typically describes a distance between the beginning of a high-risk area located within a hidden area and the phantom object. A high-risk area is typically the area of ​​particular interest in a hidden area of ​​a specific type.For example, the high-risk area within a occluded area of ​​the "crosswalk" type is typically the crosswalk itself, although the occluded area of ​​the "crosswalk" type may also include areas near the crosswalk that are not part of the crosswalk itself. Similar definitions apply to other high-risk areas such as bus stops, taxi stands, intersections, or the like. The "distance threshold" is a special value of the phantom object distance after which the static component has the value 0. The distance threshold is preferably located outside the high-risk area but within the occluded area. In typical embodiments, the static component has the value of the initial ambient probability over the entire high-risk area.In preferred embodiments, the static component preferably decreases linearly from one end of the high-risk range to the distance threshold, where it takes the value 0.

[0022] In preferred embodiments, the appearance probability has a dynamic component, wherein the dynamic component preferably at least indirectly takes into account a geometric modification of the concealed area between two points in time, and / or wherein the dynamic component preferably depends on a one-phantom object length, wherein the one-phantom object length is defined as a length within a concealed area in which exactly one phantom object is expected, wherein the one-phantom object length is preferably measured along a driving path (if the phantom object is a phantom vehicle) and / or along a walking path (if the phantom object is a phantom pedestrian), and / or wherein the dynamic component preferably depends on a field of view change, wherein the field of view change is preferably a length,which is measured along a driving path (if the phantom object is a phantom vehicle) and / or along a walking path (if the phantom object is a phantom pedestrian), wherein the single-phantom object length and the field of view magnification are typically parallel. In typical embodiments, the driving path is a lane extracted from a high-resolution map. In typical embodiments, the walking path is a pseudo-priority walking path. In typical embodiments, the walking path consists of a plurality of waypoints that start at one edge of the field of view (FoV) and point to another side of the road.

[0023] In typical embodiments, the static component is calculated according to the following equation: penv(d)=max((KenvDs−dDs),0) where p env (d) the static component is K envis the initial environment probability, Ds is the distance threshold and d is the phantom object distance, and / or the dynamic component is preferably calculated according to the following equation: pFoV(Δs)={0 ,for Δs≤0ΔsL,for>0∧Δs <L1 ,fu¨r Δs≥L where p FoV (Δs) is the dynamic component, Δs is the visual field change and L is the one-phantom object length, and / or the probability of occurrence is preferably calculated according to the following equation: pa(d, Δs)=min((penv(d)+pFoV(Δs)), 1), where p a (d, Δs) of the probability of occurrence p app corresponds, but with “p a (d, Δs)” to indicate that the probability of occurrence p app in this particular case depends on the phantom object distance and the field of view magnification.

[0024] The problem is further solved by a system for planning the behavior of a vehicle with respect to a hidden area, the system being designed to generate a phantom object for the hidden area and to calculate an appearance probability for the phantom object, the appearance probability describing the probability for the phantom object to emerge from the hidden area in a field of view of the vehicle, characterized in that the system is designed to determine a modified appearance probability based on a message received from a V2X device communicating with the vehicle and on the basis of the appearance probability.

[0025] The system is preferably designed to carry out a method according to one of the previously described embodiments. In typical embodiments, the system comprises a module for identifying occluded areas and / or a module for generating phantom objects and / or a module for calculating appearance probabilities, which is designed to calculate the aforementioned appearance probability, typically including the static component and / or the dynamic component. In typical embodiments, the system comprises a module for modifying appearance probabilities to determine the modified appearance probability, in particular based on the message received from the V2X device and the appearance probability.In preferred embodiments, the system comprises a module for determining confidence modifiers and / or a module for determining modified occurrence probabilities. In typical embodiments, the system comprises a module for calculating observable regions and / or a module for identifying occlusion regions and / or an assignment module and / or a detection module and / or a selection module.

[0026] In typical embodiments, at least one of these modules, preferably all of these modules, is / are implemented using software code. The system typically comprises means for performing at least one method according to one of the aforementioned embodiments, in particular computer hardware means such as processing units, storage devices, or the like for participating in the various methods and / or steps and / or processes and / or subprocesses and / or routines described above.

[0027] The object is further achieved by a vehicle, wherein the vehicle has a system according to one of the above-mentioned embodiments and / or wherein the vehicle is designed to carry out one of the above-mentioned methods.

[0028] In a typical embodiment of the invention, a computer program comprises instructions which, when executed by a computer, cause the computer to perform a method according to one of the aforementioned embodiments. The term "computer" is to be understood as referring to any device or structure capable of executing the instructions. The computer program may also be referred to as a computer program product.

[0029] In one embodiment of the invention, a computer-readable medium comprises a computer program code for carrying out a method according to one of the aforementioned embodiments and / or comprises a computer program according to the aforementioned embodiment. The term "computer-readable medium" can be understood to mean, in particular but not exclusively, hard drives and / or servers and / or memory sticks and / or flash drives and / or DVDs and / or Blu-ray discs and / or CDs. Furthermore, the term "computer-readable medium" can also refer to a data stream that arises, for example, when downloading a computer program and / or a computer program product from the Internet. Short description of the drawings

[0030] The invention is explained below with reference to the figures, in which: Fig. 1: a schematic visualization of a problem underlying the invention, Fig. 2: a schematic visualization of an aspect of the invention, Fig. 3: a flowchart of a method according to an embodiment of the invention, Fig. 4: an overview of five different types of occlusion, Fig. 5: a schematic visualization of a first algorithm executed in an embodiment of the invention, Fig. 6: a schematic visualization of a second algorithm executed in an embodiment of the invention, and Fig. 7: Trust factors for selected types of V2X devices. Description of preferred embodiments

[0031] Fig. Figure 1 shows a schematic visualization of a problem underlying the invention. In Fig. 1, in particular, a road intersection 5 is shown. A vehicle 1 is approaching this road intersection 5. In the vicinity of the road intersection 5, there is an obstacle 2, for example a truck or a bus. This obstacle 2 forms a hidden area 4 for the vehicle 1, for which the objects located in the hidden area 4 are not visible. The hidden area 4 is in Fig. 1 is represented by dashed lines. In the concealed area 4 there is a concealed vehicle 7, which is therefore for the vehicle 1 in the Fig. 1 is not visible. One aspect of the present invention is the creation of so-called phantom vehicles for hidden areas, which are used in the planning of autonomous driving scenarios for the vehicle 1. For each of these phantom objects, an appearance probability is calculated, which describes a probability that the phantom object appears from the hidden area 4 in a field of view of the vehicle 1. For reasons of clarity, the phantom object is in Fig. 1 not shown.

[0032] Fig. Figure 2 shows a schematic visualization of an aspect of the invention. In particular, Fig. 2 the same intersection 5 as in Fig. 1 with the vehicle 1 and the obstacle 2. Here too, the obstacle 2 forms the hidden area 4 for the vehicle 1. The Fig. However, the hidden vehicle 7 shown in Figure 1 is in Fig. 2 is not shown. Instead, Fig. 2 a phantom object 3 in the hidden area 4, which is used during the planning process for the autonomous driving of the vehicle 1. Furthermore, in Fig. 2 shows a V2X device 6 installed next to the road intersection 5. The V2X device 6 may, for example, be a camera. The V2X device 6 may observe details of the hidden area 4 and send messages about the hidden area 4 to the vehicle 1, for example, about real objects located in this hidden area 4, for example the hidden vehicle 7 (in Fig. 1, but not in Fig. 3 because of the presence of the phantom object in Fig. 2). With the message received by the V2X device 6, the vehicle 1 could also receive information about the Fig. 1 shown hidden vehicle 7. However, as already mentioned, this vehicle 7 is in Fig. 2 not shown, since Fig. 2 instead shows the phantom object 3. Since the V2X device 6 exists, an appearance probability calculated for the phantom object 3 can be replaced by a modified appearance probability determined during the method based on a message received from the V2X device 6 and the initial appearance probability.

[0033] Fig. Figure 3 shows a flowchart of a method according to an embodiment of the invention. In particular, Fig. 3 a phantom object generation step S1, in which a phantom object 3 for the hidden area 4, which for example in Fig. 2. The phantom object generation step S1 comprises a calculation of the appearance probability for the phantom object 3. The Fig. 3 further comprises an appearance probability modification step S2, in which a modified appearance probability is calculated on the basis of a probability of appearance transmitted by a V2X device 6 communicating with the vehicle 1 in Fig. 2 received message and based on the probability of appearance calculated during the phantom object generation step S1. The flowchart in Fig. Figure 3 shows that the method is performed in an endless loop. This may be the case in some embodiments, but does not always have to be the case. In some embodiments, the method may also be performed only at specific times or only when needed, for example, when vehicle 1 approaches an intersection or when a message sent by a V2X device is received by vehicle 1 via a hidden area.

[0034] In a specific embodiment, an on-board occlusion analysis module in an ego vehicle is used to calculate observable areas Praise and the identification of occlusion areas Locc along the ego route r ego of the ego vehicle using data from the on-board perception. The observable areas Praise are the areas that can be detected by the on-board sensors of the ego vehicle. First, a complete area centered on the ego vehicle and defined by the maximum range D p of the perception system. Then, this fully observable area is mapped to the observable areas based on perceived static and dynamic obstacles Locc shrunk. The map then shows the route r egofor hidden lanes and areas Locc A catalog of occlusions is introduced to describe the different occlusion types φ occ,k for assignment to Locc to be defined. The occlusion type φ occ,k describes the lane relationships to the ego vehicle, such as oncoming lanes or the ego navigation lane. In addition, the relative priority of Locc relative to the ego vehicle at intersections without traffic signs or traffic lights, i.e., high-priority lanes and low-priority lanes. This is important in countries like Germany, where drivers must follow the right-before-left rule when approaching an intersection without traffic signs or traffic lights. Furthermore, unobservable areas around pedestrian crossings and bus stops are considered as occluded regions of interest.

[0035] In a particular embodiment, an observation model for phantom objects in the ego vehicle is used to generate phantom road users N k responsible, including their attributes, states χ k and estimated (in other words: initial) probabilities of occurrence p app,k Phantom road users are identified depending on the types of masking areas Locc As in Fig. 4, each phantom road user N k Attributes assigned, including the occlusion type φ occ,k , the object type φ obj,k and the priority φ pri,k . These attributes are not in the states χ k the phantom road user N k because they are static values. The object type φ occ,k represents a phantom vehicle or a phantom pedestrian. The priority φ pri,krepresents the relative priority of a phantom object compared to the ego vehicle.

[0036] Fig. 4 also contains information about the location (x k , y k ), the orientation θ k and the speed v k to generate the state χ k the phantom road user. For example, a phantom vehicle N k at the edge of the FoV with the arc length s k along a lane r k with high priority at a speed of v k the maximum speed. The Cartesian coordinates (x k , y k ) and the orientation θ k are removed from the intended path r k according to the path geometry based on the position sk. A static phantom vehicle is positioned at a safe distance D fplaced in front of the leading vehicle. This takes into account situations in which the leading vehicle suddenly performs an evasive maneuver due to a static obstacle, and allows the ego vehicle sufficient reaction time. Pseudo-priority walking paths are generated for a phantom pedestrian. The walking paths consist of waypoints that start at the edge of the FoV and point to the other side of the street. The lengths of all phantom road users are defined as infinite in the occluded areas, which allows the representation of a set of achievable states with only one configuration of phantom road users.

[0037] In typical embodiments, the appearance probabilities of the generated phantom road users are initialized. The assumption that the phantom road users always emerge from the occlusion area during each planning cycle would lead to the ego vehicle driving overcautiously. In some cases, this assumption could even block the ego vehicle and lead to a "frozen state" without further movement. To avoid this, different initial appearance probabilities are used depending on the occlusion type φ. occ,k and the phantom object type φ obj,k introduced. The probability of occurrence p app,k= 0 is assigned to phantom vehicles in low-priority lanes, meaning that with exclusive onboard perception, the risk of vehicles appearing and initiating illegal behavior from these areas is not taken into account. This prevents overly cautious driving by the ego vehicle. However, with V2X communication, the appearance probability of low-priority phantom vehicles can be increased if a vehicle is detected that violates the right-of-way rule and crosses the intersection before the ego vehicle. In addition, the appearance probability p app,k = 0 is also assigned to static phantom objects in the ego lane. For the remaining group of phantom road users, their probability of appearance is calculated with p app,k = p a (d, Δs) initialized: pa(d, Δs):=min((penv(d)+pFoV(Δs)), 1).

[0038] The probability of occurrence p a (d,Δs) consists of a domain-specific component p env (d), which takes map information into account, and a dynamic component p FoV (Δs), which detects the change in the field of view (FoV) of the ego vehicle in future time steps. Its components are presented below. The first component p env (d) represents the environmental context: penv(d):=max((KenvDs−dDs), 0), where K env represents an initial environmental probability. The distance d and a distance threshold Ds denote the location (x k , y k ) of the phantom object to the boundary of the occlusion area. K envreflects the phenomenon that the appearance of road users depends on the location of the occlusion. For example, the probability of pedestrians appearing at pedestrian crossings, near bus stops, and in school zones is higher than on normal road sections. The probability of phantom road users appearing decreases with distance d.

[0039] The second part of the probability p FoV (Δs) describes the probability of the appearance of phantom objects due to a change in the FoV with Δs = s' k - s k : pFoV(Δs):={0,if Δ≤0,ΔsL,if Δs>0∧Δs <L,1,sonst. L is defined as the length within which exactly one phantom object is expected to be observed. The FoV can change at each time step during the forward simulation. With an extended FoV, the chance that the ego vehicle observes a previously hidden object increases. It is assumed that the phantom vehicles and pedestrians are normally distributed in the hidden lane or hypothetical path. When the field of view is extended by Δs meters, the probability that at least one phantom object appears is ΔsL. If the FoV decreases or does not change, no additional probability is assigned for the occurrence of phantom objects.

[0040] In a preferred embodiment of the invention, a V2X communication module of the ego vehicle first finds a subset of V2X devices C from all available V2X devices Casc,k, the phantom road user N k It then calculates and selects the most appropriate confidence modifier pcm,k* to determine the degree of confidence regarding the existence of road users at the location of the phantom road user N k to be evaluated. The confidence modifier of a V2X device c for the phantom object N k takes the detection result into account pd,kc, the maximum number of trust steps tt,kc for the device due to the coverage of the observation area for the phantom road user N k , the delay factor ft,kc to evaluate communication latency and the trust factor fq,kc, which represents the reliability of the sensor.

[0041] The Fig. Algorithm 1, shown in Figure 5, outlines the assignment step in the V2X communication module. The inputs to the algorithm are all generated phantom traffic participants N k and a list of V2X devices C from which the ego vehicle receives messages. Each V2X device c delivers a message with the observable areas Lobs,c, which represent the areas that the device c can observe, an object list χ c , which contains the detected objects, and a timestamp T c , which indicates the time at which the message was sent. For each phantom object N k an empty list of V2X devices Cask,k initialized (lines 2-3). Then it is determined whether the observable areas Lobs,c each V2X device c the location (x k , y k ) of the phantom road user N k(lines 4-6), and it will be added to the corresponding device list Cask,k added (line 7). The output of the algorithm is the list of associated V2X devices Cask, the subsets of V2X devices Cask,k for each phantom road user N k contains.

[0042] The Fig. Algorithm 2 shown in Figure 6 shows the calculation and selection of the confidence modifier pcm,k∗. First, the observable areas Lobs,c, the list χ c of the observed objects and the timestamp T c from the assigned V2X device c for the phantom object N k (line 3). Then the detection result pd,kc, the maximum number of trust steps tt,kc, the delay factor ft,kc, the trust factor fq,kc and a confidence modifier candidate pcm,kc, which determines the confidence of the V2X device c for the phantom object N k (see lines 4-8). pcm,kc List P cm,k in which all confidence modifier candidates for the phantom object N k are stored (line 9). Finally, the most reliable confidence modifier candidate is used as the confidence modifier pcm,k∗ for the phantom object N k selected when it is observed by multiple V2X devices (line 11). The algorithm is explained in detail below.

[0043] The detection result pd,kc is calculated by evaluating whether a detected object N j from the object list χ c the V2X device c with the phantom road user N kmeets the following conditions: pd,kc:={1,if djk≤Dc∧Δθjk≤Θc,−1,otherwise. where d jk and D c the Euclidean distance or the distance threshold between the phantom object N k and the detected object N j represent. Δθ jk and Theta c denote the relative alignment and the threshold of relative alignment between them.

[0044] The maximum number of trust steps tt,kc is calculated for the V2X device c to determine the number of time steps for which the confidence modifier can be applied to estimate the probability of appearance of the phantom object N k to adjust: tt,kc:=dk,mincvk.

[0045] This is dk,minc the minimum distance of the phantom object N k to the limit of the observable area Lobs,c from the V2X device c and v k the assumed speed of the phantom object N k . Beyond the maximum number of trust steps tt,kc other undetected objects may enter the observable area and reach the position of the phantom object.

[0046] Delays can also be taken into account when the ego vehicle uses communication messages. During delays, the detected objects continue to move, making the perception data unreliable. Therefore, a delay factor ft,kc∈[0,1] introduced to indicate the degree of trust that the ego vehicle should assign to the message from the V2X device c: ft,kc:={1−Δtc,if 0≤Δtc≤1,0,otherwise,

[0047] Where Δt c = T cur - T c the time interval between time T cur, at which the message is processed by the ego vehicle, and T c to which it is sent. If messages from V2X device c are older than 1 second, where Δt c ≥ 1 second, the delay factor ft,kc=0, which means that such a message will not be taken into account.

[0048] It becomes a trust factor fq,kc to prioritize more reliable devices when multiple V2X devices share the location of the phantom object N k observe. Fig. Figure 7 shows the trust factors for selected types of V2X devices. For example, certified devices, such as the ZF ProConnect connectivity platform, are more trusted than non-certified devices, such as webcams. Furthermore, a V2X vehicle of the same brand can provide more accurate calibrated data than vehicles of other brands.

[0049] The confidence modifier candidate pcm,kc, the detection result pd,kc, the maximum number of trust steps tt,kc, the delay factor ft,kc and the trust factor fq,kc combined, is calculated as follows: pcm,kc:=pd,kc⋅ft,kc⋅fq,kc, if th <tt,kc, where t h ∈ {0, ...,H} is a time step within the planning horizon H ∈ R+. If several confidence modifiers pcm,kc are available, the confidence modifier pcm,k* with the highest confidence as follows: pcm,k*:=f(Pcm,k):={max(|Pcm,k|),if ∀pcm,kc∈Pcm,k:pcm,kc<0max(Pcm,k),otherwise.

[0050] If pcm,kc>0 the V2X device detects an object that corresponds to the phantom object N kThe device with the highest confidence is selected. In preferred embodiments, in the case of conflicting detection results from multiple V2X devices, a conservative preference is implied. In such cases, the device that detects the object with pcm,kc>0 detected, preferred to the device that does not detect the object (pcm,kc<0). If all devices at the location of the phantom object do not detect an object (pcm,kc<0), the device with the highest absolute confidence (max (|P cm,k |)) selected.

[0051] Finally, the selected confidence modifier pcm,k∗ applied to determine the initial probability of occurrence p app,k for the phantom object N k to modify, if available, the modified probability of occurrence papp,k∗ for each phantom object as follows: papp,k∗={max(0, min(1, papp,k+pcm,k∗)),if pcm,k∗ available,papp,k,otherwise.

[0052] The invention is not limited to the preferred embodiments described herein. The scope is defined by the claims.

[0053] It should further be noted that the methods, processes, sub-processes and / or routines disclosed in the description or in the claims may be implemented by an apparatus having means for performing the respective acts of these methods, processes, sub-processes and / or routines. List of reference symbols 1 vehicle 2 Obstacle 3 Phantom object 4 hidden area 5 intersection 6 V2X device 7 concealed vehicle S1 Phantom object creation step S2 Appearance probability modification step

Claims

[1] Method for planning the behavior of a vehicle (1) with respect to at least one hidden area (4), the method comprising: - a phantom object generation step (S1) in which at least one phantom object (3) is generated for the hidden area (4), wherein the phantom object generation step (S1) comprises a calculation of an appearance probability for the phantom object (3), wherein the appearance probability describes a probability that the phantom object (3) appears from the hidden area (4) in a field of view of the vehicle (1), characterized by , that - the method comprises an appearance probability modification step (S2) in which a modified appearance probability is determined on the basis of a message received from a V2X device (6) communicating with the vehicle (1) and on the basis of the appearance probability. [2] Method according to claim 1, characterized by that the method comprises a confidence modifier calculation process in which a confidence modifier to be used in the appearance probability modification step (S2) is preferably calculated for each phantom object (3), the confidence modifier calculation process comprising: - an assignment sub-process in which the phantom object (3), preferably all phantom objects (3), is / are assigned to a list of V2X devices (6) that are / are capable of observing the respective phantom object (3), and / or - a detection sub-process in which it is checked whether real objects are present in the hidden area(s) of the respective phantom object(s) (3), and / or - a determination sub-process in which a list of confidence modifier candidates is determined, wherein the list preferably comprises a confidence modifier candidate for each V2X device (6) present in the list of V2X devices, and / or - a selection sub-process in which the confidence modifier is selected, preferably from the list of confidence modifier candidates. [3] Method according to claim 2, characterized by that the determination sub-process includes: - a calculation of a maximum number of confidence steps for each V2X device (6) present in the list of V2X devices (6), wherein a confidence step is typically a planning step of the method which ensures that no undetected real objects enter the observable area of ​​the respective V2X device (6), and / or - a calculation of a delay factor for each V2X device (6) present in the list of V2X devices (6), wherein the delay factor typically describes the extent to which a message received by the respective V2X device (6) should be trusted, and / or wherein the delay factor typically has a value of 0 if a message is older than one second, and / or wherein the delay factor typically has a value of less than 1 but more than 0 if a message is newer than one second, and / or - a determination of a trust factor for each V2X device present in the list of V2X devices (6), wherein the trust factor typically describes the extent to which the respective V2X device (6) can be trusted, wherein each trust factor preferably has a value between 0.7 and 1. [4] Method according to one of claims 2 to 3, characterized by , that: - each confidence modifier candidate is preferably calculated according to the following equation pcm,kc:=pd,kc⋅ft,kc⋅fq,kc, if th <tt,kc. and / or - where, if more than one confidence modifier candidate has been calculated, the confidence modifier is calculated according to the following equation pcm,k∗:=f(Pcm,k):={max(|Pcm,k|),if ∀pcm,kc∈Pcm,k:pcm,kc<0max (Pcm,k),otherwise. and / or - where in the case that only one confidence modifier candidate was calculated, the confidence modifier corresponds to the confidence modifier candidate. [5] Method according to one of the preceding claims, characterized by that the modified probability of occurrence is calculated according to the following equation papp,k∗={max(0, min(1, papp,k+pcm,k∗)),if pcm,k∗ available,papp,k,otherwise. [6] Method according to one of the preceding claims, characterized bythat the method comprises a process for selecting hidden regions, in which the hidden region is selected from a list of types of hidden regions, wherein the list of types of hidden regions preferably comprises five types of hidden regions. [7] System for planning the behavior of a vehicle (1) with respect to a hidden area (4), wherein the system is designed to generate a phantom object (3) for the hidden area (4) and to calculate an appearance probability for the phantom object (3), wherein the appearance probability describes the probability that the phantom object (3) appears from the hidden area (4) in a field of view of the vehicle (1), characterized bythat the system is designed to determine a modified appearance probability based on a message received from a V2X device (6) communicating with the vehicle (1) and on the basis of the appearance probability. [8] Vehicle (1), characterized by , that - the vehicle (1) comprises a system according to claim 7, and / or - the vehicle (1) is designed to carry out a method according to one of claims 1 to 6. [9] A computer program comprising instructions which, when executed by a computer, cause the computer to carry out a method according to any one of claims 1 to 6. [10] A computer-readable medium comprising a computer program code for carrying out a method according to any one of claims 1 to 6 and / or comprising a computer program according to claim 9.

Images