WO2025162580A1

WO2025162580A1 - Method and apparatus for assessing risk of deploying at least one traffic participant at a specific location

Info

Publication number: WO2025162580A1
Application number: PCT/EP2024/052412
Authority: WO
Inventors: Iris FUHRMANN; Stefan GADATSCH; Shehzaman Salim KHATIB; Rhena KIEFFER; Sagar PATHRUDKAR; Kaushlendra Pratap Singh; Pruthvi TURPUSEEMA; Kevin Luis VOOGD
Original assignee: Siemens AG; Siemens Corp
Current assignee: Siemens AG; Siemens Corp
Priority date: 2024-01-31
Filing date: 2024-01-31
Publication date: 2025-08-07
Anticipated expiration: 2026-07-31

Abstract

The present disclosure relates to a method and an apparatus (100) for assessing risk of deploying an at least one traffic participant (12, 14) at a specific location. The apparatus comprises interface circuitry (110) configured to receive trajectory data (112) indicating trajectory information of at least one traffic participant (12, 14) of a traffic scenario at a specific location. Further, the apparatus (100) comprises processing circuitry (120) configured to analyze the trajectory data (112) for occurrence of near miss events involving the at least one traffic participant (12, 14). The processing circuitry (120) is further configured to determine a safety criticality of the near miss events based on the analysis, the safety criticality considering a closeness to becoming a collision and an estimated severity in case of such collision. In addition, the processing circuitry (120) is configured to generate output data (122) indicating the safety criticality for the specific location.

Description

Method and apparatus for assessing risk of deploying at least one traffic participant at a specific location

The present disclosure relates to a method and an apparatus for assessing risk of deploying at least one traffic participant at a specific location. Further, the present disclosure relates to a non-transitory machine- readable medium and a computer program.

Assessing safety and/or risk of traffic participants, such as autonomous vehicles, is challenging since there is little historical data available to derive the risk given that there is limited deployment of autonomous vehicles, and the autonomous vehicle operates in complex environments characterized by an explosion of evolutionary possibilities given the sequential and closed loop nature due to the independent decisions taken by multiple actors having their own “free-will”.

Accordingly, the safety and/or risk for deploying the traffic participant, e.g. the autonomous vehicle, may be assessed in loop with a realistically simulated environment and assessment of their response using different safety or risk metrics. The metrics used to evaluate safety response of the autonomous vehicle may include collision frequency and severity, time-to-collision (TTC), responsibility-sensitive-safety (RSS), etc. Thereby, collision incidents are rare and do not provide enough information regarding safety response of the autonomous vehicle and risk corresponding to a location or scenario. As a result, the safety of the autonomous vehicle may only be assessed roughly or with difficulty.

Hence, there may be a need for improved safety and/or risk assessment of deploying a traffic participant.

This need is met by a method for assessing risk of deploying at least one traffic participant at a specific location, and an apparatus for assessing risk of deploying at least one traffic participant at a specific location.

According to a first aspect, there is provided a method for assessing risk of deploying at least one traffic participant at a specific location. The method comprises receiving trajectory data indicating trajectory information of at least one traffic participant of a traffic scenario at a specific location. The method further comprises analyzing the trajectory data for occurrence of near miss events involving the at least one traffic participant. In addition, the method comprises determining a safety criticality of the traffic scenario based on the analysis of the trajectory data, the safety criticality considering a closeness of a near miss event to becoming a collision and an estimated severity in case of such collision. Further, the method comprises generating output data indicating the safety criticality for the specific location.

The method uses information rich non-collision interactions of traffic participants, i.e. a near miss event, which are more frequently than actual collisions between traffic participants. In this way, an accurate, systematic and/or comparable safety assessment and/or risk assessment may be carried out even without actually recording collisions or data about them. The method may be used to assess and/or estimate how safe or unsafe or how risky the deployment of a traffic participant is at a specific location. This may be expressed by the safety criticality determined for this specific location. The safety criticality may then be considered and/or used for the planning and/or actual deployment of any traffic participant, which may also be an autonomous vehicle, at the specific or a comparable location. For example, based on the safety criticality, it may be decided whether the traffic participant in general should be deployed at the specific location at all. Also, it may be decided based on the safety criticality whether and which measures should be taken to make the deployment of the traffic participant as safe as possible.

It is noted that in this disclosure, independent of the grammatical term usage, individuals with male, female or other gender identities are included within the term.

The method may be computer-implemented and may be carried out by any suitable data processor, computation device, or the like. The method may be carried out by a single entity or by multiple entities, a distributed computer system, etc.

As used herein, the at least one traffic participant may be any traffic participant involved in the traffic situation under consideration. In the traffic situation and/or trajectory data, the least one traffic participant may be at least present and/or may interact with its environment, with another traffic participant, or the like. For example, the traffic participant may be any one of a vehicle, such as a car, truck or bus, a bicycle, a motorbike, and/or a pedestrian. Pedestrians, cyclists and/or motorcyclists may be referred to as vulnerable road users, as these are exposed to a higher risk, in particular the risk of injury. Further, the at least one traffic participant may be an autonomous vehicle, i.e. an at least partially automated driving vehicle. The term “vehicle” may include motor vehicles in general such as passenger automobiles including sports utility vehicles (SUV), buses, trucks, various commercial vehicles, and the like, and includes hybrid vehicles, electric vehicles, plug-in hybrid electric vehicles, hydrogen-powered vehicles, and other alternative fuel vehicles (e.g., fuels derived from resources other than petroleum). The terms “autonomous” and/or “at least partially automated driving” may be understood that such vehicle utilizes an at least semi-automated driving system, e.g. according to an SAE level 1 to 5 or the like.

Further, the trajectory data and/or trajectory information may be derived from real world data and/or simulation data. The trajectory data may refer to at least one traffic scenario. The traffic scenario may also refer to at least one maneuver captured in or derivable from the trajectory data. It is noted that in case that no near miss is detected in the trajectory data, i.e. during the analysis of the trajectory data, this indicates no criticality, so that this may indicate that there is no risk for this location. Such non-criticality may, however, still be considered when determining the safety criticality.

Further, as used herein, a near miss, may be understood as any event that came close to becoming a collision but did not necessarily become one, e.g. due to an evasive maneuver by at least one traffic participant involved. Based on both real-world data and simulation data, near miss events occur more frequently than actual collisions.

As used herein, the safety criticality may be understood, for example, as an indicator, measure, score, or the like, being indicative for the safety and/or risk of deploying the at least one traffic participant at the specific location. The closeness to becoming a collision and/or the estimated severity in case of such collision may be understood as a quantification, score, or the like. The closeness to becoming a collision may indicate how close a near miss came to becoming a collision. The severity in case of collision may indicate how bad it would have been had it become a collision. The less likely a near-miss is to become an actual collision, e.g. an accident, the more are the risk and/or costs reduced. The potential cost-based severity of a near miss event may be regarded as the combination of how bad and how close the near miss event has been, which may also be referred to as the safety criticality.

The output data indicating the safety criticality for the specific location may be output in electronic and/or machine-readable form for further processing, visualization, or the like.

In an embodiment, the output data may be provided and/or used for deploying the at least one traffic participant. The knowledge about the safety criticality for the specific location may be used in a variety of ways in connection with the deployment of such traffic participant. For example, the safety criticality may be used in urban development, site planning, in the planning of a vehicle deployment or a vehicle fleet deployment, in route planning, in driving strategy planning, in driving trajectory planning, in planning an area of deployment of an autonomous vehicle, in the determination of a cost distribution for the vehicle deployment, in the calculation of insurance fees, or the like, wherein this is not limited herein.

According to an embodiment, the specific location may be at least one of a geographical location, fictional location, a geographical area, a road section, a road area, a traffic area, a road route, and a travel route. The specific location may be received, pred-determined, etc.

In an embodiment, multiple near miss events and corresponding safety criticalities may be determined. The method may further comprise aggregating the safety criticalities of the multiple near miss events. Further, the method may comprise determining an overall safety criticality for the specific location based on the aggregated safety criticalities. In addition, the method may comprise determining a safety criticality distribution for the specific location. The overall safety criticality may be used to assess the deployment of the at least one traffic participant at the specific location.

According to an embodiment, the method may further comprise determining a potential cost severity based on the corresponding safety criticality. In other words, the safety criticality may be translated into costs. This may also be used to determine the potential cost severity of a single near miss event. In an embodiment, multiple near miss events and/or corresponding safety criticalities may be determined. Further, the potential cost severities of the multiple near miss events may be aggregated. In addition, a cost severity distribution for the specific location may be determined. Once the cost severity is determined for multiple near misses, it may be aggregated, and a cost severity distribution may be provided for the analyzed specific location.

Optionally, the type of maneuver may be classified. With a more detailed context description of the near miss, its collision conversion rate may be refined, adding to the weighting of the near miss event.

According to an embodiment, the trajectory data may comprise at least kinematic data and metadata of the at least one traffic participant. The determinations and calculations used in the method may be derived by using only the kinematic data and the metadata of the at least one traffic participant involved without the need for any heavy FEA/physics or data-driven models. This may reduce the computational effort. The kinematic data and the metadata may be derived from either a simulation scenario or real- world data, or from both. The kinematic data and the metadata may be contextualized to the traffic situation and/or maneuver.

In an embodiment, the kinematic data may comprise at least one of a position, e.g. coordinates, an orientation, a velocity, an acceleration, and a yaw of the at least one traffic participant.

According to an embodiment, the metadata may indicate the traffic scenario. It is noted that, as used herein, the traffic scenario does not necessarily refer only to public road traffic, but may also refer to private property, construction sites, etc.

In an embodiment, the meta data may comprise at least one of a road user type, a vehicle type, a mass, a length, a shape, a dimension, or the like, of the at least one traffic participant, weather data, and traffic information.

In an embodiment, the determining of the safety criticality may weight whether the at least one traffic participant is a vulnerable road user (VRU) exposed to an increased risk. For example, the vulnerable road user may include at least one of a pedestrian, cyclist, and motorcyclist. For example, the method may account for this factor and may define a respective bounding box, safety envelope, or the like for these vulnerable road users. A different set of metrics may be defined about for different vulnerable road users, such as pedestrian, bicyclist, etc. A corresponding near miss threshold distance may depend on the different characteristics of vulnerable road users and that of the vehicle which is about to be in proximity of the vulnerable road user. The weighting may also be indicated by the trajectory data, the kinematic data and/or the metadata.

According to an embodiment, the determining of the at least one near miss event may comprise determining a safety distance metric involving the at least one traffic participant based on the trajectory data. The at least one near miss event may be determined based on a threshold of the safety distance derived using a combination of a number of factors, such as detection time, judgement and action time which together form the response time along with the time required to evade/maneuverthe at least one traffic participant away from the dangerous situation. This maneuver may comprise stopping the at least one traffic participant, accelerating it away from the dangerous interaction or changing the course of the at least one traffic participant to another direction. The safety distance may also be used as a metric that is influenced by the type of road user, their size, their mass, tire gauge (in case of vehicles) and model of the vehicle. The safety distance combined with the direction of the movement may determine the safety buffer or the safety envelope for each individual road user, i.e. traffic participant. The safety distance used to define the bounding box, safety envelope, etc. may be directional. Depending on the direction in which the at least one traffic participant is moving and proportional to the speed at which they are moving, the safety distance changes. As mentioned above, some traffic participants are more at risk than others and are commonly referred to as vulnerable road users (VRU). The dynamics and kinematics of vulnerable road users are very different from a vehicle perspective. Therefore, there may be defined a different set of metrics for different vulnerable road users. The threshold distance of vulnerable road users may depend on the different characteristics of vulnerable road users and that of the vehicle which is about to be in proximity of the vulnerable road users.

In an embodiment, detecting of the near miss event may comprise determining a respective bounding box of the at least one traffic participant based on kinematic data and metadata of the at least one traffic participant and determining the near miss event based on a distance and/or overlap with the bound box. The bounding box may be a polygon or the like enclosing and/or enveloping the at least one traffic participant.

According to an embodiment, the bounding box of the at least one traffic participant and the corresponding distance and/or overlap may be determined for a number of timesteps to determine the near miss event. For example, the kinematic data and the metadata may be contextualized to the traffic scenario and/or maneuver, wherein a directional safety distance and/or envelope and corresponding expanded bounding boxes for the at least one traffic participant may be calculated at every timestep.

In an embodiment, for near miss events occurred, a worst point where the near miss is closest to become a collision may be determined. The worst point may also be referred to as a critical point in time, where the near miss is closest to become a collision - or is a collision. The worst point may be used for determining the kinematic parameters as referred to herein.

In an embodiment, the worst point may be determined from a minimum enhanced time-to-collision, minETTC, indicating the time where a trajectory prediction of the vehicle is closest in time to collide. The minETTC is defined in e.g. ISO 23376, 15623. The minETTC may depend on the relative positions, velocities and accelerations of the traffic participants involved. The minETTC may define, which kinematic data may used at the worst point. According to an embodiment, the closeness to becoming a collision of the near miss event may be determined by calculating a closeness weighting factor based on a distance with or to a bounding box assigned to the at least one traffic participant and/or a corresponding minimum enhanced time-to- collision, minETTC. The closeness weighting factor c may provide a static function between 0, at the threshold of the near miss determination, when bounding boxes only slightly touch, to 1 , equal to a collision, considering the minETTC.

In an embodiment, the estimated severity in case of collision may comprise at least one of a bodily injury severity and a property damage severity.

According to an embodiment, the bodily injury severity may be estimated based on maximum abbreviated injury scale, MAIS. The MAIS is an anatomical-based coding system to classify and describe the severity of injuries. For example, based on empirical formulas the injury level of the passengers inside the vehicle and another vehicle, or any other traffic participant or vulnerable road user, e.g. pedestrian, bicyclist, motorcyclist, etc, may be predicted.

In an embodiment, the estimating of the severity in case of collision may comprise determining pre-crash parameters, e.g. at the worst point, based on a respective mass and respective velocity of the at least one traffic participant. It may further comprise determining a collision configuration being indicative for a location of impact comprising at least one of a front impact, a rear impact, a driver-side impact, and a passenger-side impact. Further, it may comprise determining a speed difference, which is also referred to herein as DeltaV, between a speed of the at least one traffic participant before and after the collision based on the pre-crash parameters and the collision configuration. Alternatively, it may comprise determining a closing speed corresponding to a relative velocity of one of the at least one traffic participant moving. It may further comprise estimating the severity in case of collision based on the estimated speed difference or the closing speed and the collision configuration.

For example, the pre-crash parameters may be determined at minETTC as an example of the worst point. In other words, the pre-crash parameters may be taken at the moment where the near miss becomes closest to a collision. The collision configuration may be determined from a projected collision angle, offset, length and width of the above-mentioned bounding box(es). The speed difference, which may be referred to as DeltaV, is the speed difference between a speed of the at least one traffic participant before and after the collision. In general, a higher DeltaV results in more severe injuries to the passengers of the vehicles. DeltaV may be estimated based on the pre-crash parameters of the at least one traffic participant and the collision configuration. For example, DeltaV may be determined by determining the closing speed, i.e., the relative velocity of one moving object with respect to the other at the worst point, e.g. minETTC. Then, a bumper's restitution coefficient is calculated based on a suitable metric, such as, for example, Antonetti, V., "Estimating the Coefficient of Restitution of Vehicle-to- Vehicle Bumper Impacts," SAE Technical Paper 980552, 1998, https://doi.org/10.4271/980552. Further, the system's mass may be determined. In addition, the projected contact plane between the two objects may be determined. The closing speed may be decomposed into the normal and tangential directions to the contact plane. Further, the impulse and velocity ratios (tangential over normal) may be computer. Further, the deformation energy of the vehicle subject to the speeds in the normal and tangential directions of the contact plane, system mass, the restitution coefficient, and the impulse ratio may be determined. DeltaV may be obtained by an empirical formula which depends on the subject object’s mass, the system’s mass, the restitution coefficient, and the deformation energy.

According to an embodiment, the method may further comprise determining an injury level and/or a property damage level based on maximum abbreviated injury scale, MAIS, based on the speed difference or the closing speed. For example, based on empirical formulas the injury level of the passengers inside the vehicle and another vehicle, or any other vulnerable road user may be predicted. The main predictor for injuries caused by vehicle-to-vehicle collision is the DeltaV followed by the location of the impact, i.e. the collision configuration, e.g. rear-end, side, or front. The probability for a certain MAIS injury level to happen based on these parameters as well as the associated property damage levels may be determined. In the case of vulnerable road users, the principal predictor for injury level is the closing speed, so that the MAIS injury as well as property damage level probabilities may be determined based on the closing speed.

In an embodiment, estimating the severity in case of collision may comprise determining a respective bounding box of the at least one traffic participant based on kinematic data and metadata of the at least one traffic participant, e.g. at the near miss event. It may further comprise determining at least one severity metric comprising at least one of an impact velocity indicating relative velocity between the centers of mass of the traffic participants, e.g. at the worst point, a near miss angle indicating an angle between longitudinal axes of the traffic participants involved at the time of near miss, an offset distance between longitudinal axes of the traffic participants involved at the time of near miss, an offset distance between two closest central axes of the traffic participants involved, an acceleration of the traffic participants involved, e.g. at the worst point, and an angular velocity of the traffic participants involved. In addition, it may comprise determining the severity in case of collision based on the at least one severity metric.

For example, the kinematic data and metadata may be derived from either a simulation scenario or real- world data. The data may comprise traffic participant coordinates, velocities, accelerations, and yaw along with metadata such as masses, shapes and dimensions of the actors involved. This data may be contextualized to the traffic scenario and/or maneuver, the directional safety distance envelope, and corresponding bounding boxes for the at least one traffic participant may be calculated at every timestep. The near miss event may be determined based on overlap with the bounding box and/or the bounding boxes of the interacting traffic participants involved at each timestep. The dynamic bounding box of the at least one traffic participant may then be analyzed at determined near miss to determine the at least one severity metric. Factors may be defined to determine the values of each individual factor, e.g. numerical values between 0 and 1 , based on empirical correlations observed. The near miss severity may be determined as normalized weighted averages of different near miss factors described above, expressed as Severity _score = duration_f actor * weight_duration + type_f actor * weight_type + of fset_f actor * weight_off set + angle_f actor * weight_off set )/sum of all weights. The following factors may be computed on the basis for severity score: time interval, impact velocity, near miss angle, traffic participant offset, type of configuration of Near miss with respect to each traffic participant, e.g. vehicle, (rear, head on, side impact).

According to a further aspect, there is provided an apparatus for assessing risk of deploying an at least one traffic participant at a specific location. The apparatus comprises interface circuitry and processing circuitry coupled to each other. The interface circuitry is configured to receive trajectory data indicating trajectory information of at least one traffic participant of a traffic scenario at a specific location. The processing circuitry is configured to analyze the trajectory data for occurrence of near miss events involving the at least one traffic participant. The processing circuitry is further configured to determine a safety criticality of the near miss events based on the analysis, the safety criticality considering a closeness to becoming a collision and an estimated severity in case of such collision. In addition, the processing circuitry is configured to generate output data indicating the safety criticality for the specific location.

The apparatus is configured to carry out the method according to the first aspect. Therefore, it may be modified in accordance with any one of the examples described herein. For the technical effects of the apparatus, reference is made to the above. The apparatus may be implemented as a single entity or may be distributed over multiple entities, such as a distributed computer system.

In at least some examples, the apparatus may be operationally connected to control circuitry for at least one vehicle, the control circuitry being configured to operate the at least one vehicle based on the output data. The apparatus may also be used during development of the vehicle, e.g. to generate driving software, etc. However, it may also be used in traffic planning, traffic control, etc.

In a further aspect, there is provided a non-transitory machine-readable medium having stored thereon a (computer) program having a program code for performing the method according to the first aspect and/or the method according to the second aspect, when the program is executed on a processor or a programmable hardware. Examples may also cover program storage devices, such as digital data storage media, which are machine-, processor- or computer-readable and encode and/or contain machine-executable, processor-executable or computer-executable programs and instructions. Program storage devices may include or be digital storage devices, magnetic storage media such as magnetic disks and magnetic tapes, hard disk drives, or optically readable digital data storage media, for example. Other examples may also include computers, processors, control units, (field) programmable logic arrays ((F)PLAs), (F)PGA), graphics processor units (GPU), ASICs, integrated circuits (ICs) or system-on-a-chip (SoCs) systems programmed to execute the steps of the methods described above. In a further aspect, there is provided a (computer) program having a program code for performing the method according to the first aspect, when the program is executed on a processor or a programmable hardware. Thus, steps, operations or processes of different ones of the methods described above may also be executed by programmed computers, processors or other programmable hardware components.

Some embodiments of apparatuses and/or methods will be described in the following by way of example only, and with reference to the accompanying figures, in which

Fig. 1 illustrates in a schematic block diagram an exemplary apparatus for assessing risk of deploying at least one traffic participant at a specific location according to an embodiment;

Fig. 2 illustrates in a schematic block diagram an example of assessing risk of deploying at least one traffic participant at a specific location according to an embodiment;

Fig. 3 illustrates in a schematic block diagram an example of determining a near miss event according to an embodiment;

Fig. 4 illustrates an example of determining a closeness weighting factor c according to an embodiment;

Fig. 5 illustrates in a block diagram an example of determining costs based on pre-crash parameters according to an embodiment;

Fig. 6 illustrates in a block diagram multiple examples of determining costs based on pre-crash parameters according to an embodiment;

Fig. 7 illustrates in an example several methods to translate severity scores or MAIS level distributions into costs that would arise assuming a near miss event would be a collision, according to an embodiment; and

Fig. 8 illustrates in a flow chart an exemplary method for assessing risk of deploying at least one traffic participant at a specific location according to an embodiment.

Embodiments of the present disclosure are described more fully hereinafter with reference to the accompanying drawings. Elements that are identified using the same or similar reference signs refer to the same or similar elements. The various embodiments of the present disclosure may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are pro-vided so that this disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art. Fig. 1 illustrates an exemplary apparatus 100 for assessing risk of deploying an at least one traffic participant at a specific location. The specific location may be at least one of a geographical location, fictional location, a geographical area, a road section, a road area, a traffic area, a road route, and a travel route.

The apparatus 100 comprises at least interface circuitry 110 and processing circuitry 120. The processing circuitry 120 is operatively coupled to the interface circuitry 110.

The interface circuitry 110 is configured to receive trajectory data 112 indicating trajectory information 10 of at least one traffic participant in a traffic scenario, wherein the at least one traffic participant may be represented by vehicle 12 and vehicle 14. The trajectory data 112 may be at least partially derived from real-life data capturing the traffic scenario, such as sensor data, video data, or the like, and/or may be at least partially derived from simulated data comprising a number or plurality of traffic scenario and/or situation simulations involving the at least one traffic participant 12, 14. It may also be possible to first determine the traffic scenario and/or situation from real-life data and then run, e.g. different, simulations for that traffic situation to obtain the trajectory data 112. It is noted that although the at least one traffic participant 12, 14 is illustrated as a vehicle, the at least one traffic participant 14 may be any one of a pedestrian, cyclist, motorcyclist, or the like. Also the type of vehicle is not limited herein, and includes, for example, a bus, truck, or the like.

The processing circuitry 120 is configured to receive and process the trajectory data 112. Further, the processing circuitry 120 is configured to analyze, e.g. compute or the like, the trajectory data 112 for occurrence of near miss events involving the at least one traffic participant 12, 14. The processing circuitry 120 is further configured to determine a safety criticality of the traffic scenario based on the analysis of the trajectory data 112. The safety criticality considers a closeness of a near miss event to becoming a collision and an estimated severity in case of such collision. Further, the processing circuitry 120 is configured to generate output data 122 indicating the safety criticality for the specific location.

For instance, the processing circuitry 120 may be a single dedicated processor, a single shared processor, or a plurality of individual processors, some of which or all of which may be shared, a digital signal processor (DSP) hardware, an application specific integrated circuit (ASIC), a neuromorphic processor or a field programmable gate array (FPGA). The processing circuitry 120 may optionally be operatively connected to, e.g., read only memory (ROM) for storing software, random access memory (RAM) and/or non-volatile memory. Optionally, the processing circuitry 120 may be operatively connected to a network controller to communicate via a network in order to re-motely control a self-driving car, e.g. the traffic participant 12, perform traffic control, or the like.

For example the trajectory data 112 may comprise and/or may indicate kinematic data and metadata of the the at least one traffic participant 12, 14. By way of example, the kinematic data may comprise at least one of a position, an orientation, a velocity, an acceleration, and yaw of the at least one traffic participant. The metadata may comprise and/or indicate at least one of a road user type, a vehicle type, a mass, a length, a shape, a dimension of the at least one traffic participant, weather data, and traffic information.

In at least some embodiments, the apparatus 100, e.g. the processing circuitry 120, may be configured to determine multiple near miss events and corresponding safety criticalities. The apparatus 100 may be further configured to aggregate the safety criticalities of the multiple near miss events and to determine an overall safety criticality for the specific location based on the aggregated safety criticalities.

Further, in at least some embodiments, the apparatus 100, e.g. the processing circuitry 120, may be configured to determine a potential cost severity for the at least one near miss event based on the corresponding safety criticality. Thereby, multiple near miss events and corresponding safety criticalities may be determined. Further, the potential cost severities of the multiple near miss events are aggregated and a cost severity distribution for the specific location may be determined therefrom.

In at least some embodiments, the apparatus 100 may be configured to weight whether the at least one traffic participant 12, 14 is a vulnerable road user exposed to an increased risk.

For example, the apparatus 100, e.g. the processing circuitry 120, may be configured to determine a safety distance metric involving the at least one traffic participant 12, 14 based on the trajectory data. Further, by way of example, the apparatus 100 may be configured to determine a respective bounding box 12a, 14a of the at least one traffic participant 12, 14 based on kinematic data and metadata of the at least one traffic participant 12, 14, and to analyze it for near miss events based on a distance and/or overlap with at least one of or between the bounding boxes 12A, 14A. The bounding box 12A, 14A of the at least one traffic participant 12, 14 and the corresponding distance and/or overlap may be determined for a number of timesteps to determine the near miss event.

In at least some embodiments, the apparatus 100, e.g. the processing circuitry 120, may be configured to determine, for near miss events occurred, a worst point where the near miss is closest to become a collision. The worst point may also be referred to as a critical point in time, where the near miss is closest to become a collision - or is a collision. The worst point may be used for determining the kinematic data and/or parameters. For example, the worst point may be determined based on or from a minimum enhanced time-to-collision, minETTC, indicating the time where a trajectory prediction of the at least one traffic participant 12, 14 is closest in time to collide.

Fig. 2 illustrates in a schematic block diagram an example 200 of assessing risk of deploying at least one traffic participant at a specific location .

At block 202, the trajectory data 112 is received. At block 204, the trajectory data 112 is analyzed for occurrence of near miss events, e.g. by performing a detection or the like. At block 206, optionally, the worst point where the near miss is closest to become a collision is determined for the at least one near miss event.

At block 208, the closeness to becoming a collision of the at least one near miss event is determined. For example, the closeness may be determined by calculating a closeness weighting factor based on a distance between bounding polygons assigned to the at least one traffic participant 12, 14 and a corresponding minimum enhanced time-to-collision (minETTC). At block 210, the closeness is weighted by using the closeness weighting factor, thereby quantifying how close the near miss becomes to a collision.

In Fig. 2, blocks 212, 214, 216 form a first branch and blocks 218, 220, 222 and 224 form a second branch. The first branch and the second branch form alternative operations to estimate severity in case of such collision for the at least one near miss event.

Regarding the first branch, at block 212, a number of factors, which may also be referred to as near-miss severity factor, is determined. The number of factors comprise at least one of a duration factor corresponding to the duration of the near miss event, a type factor corresponding to the type of traffic participant, and an angle factor corresponding to an impact and/or collision angle, a relative velocity of impact, an offset between the traffic participants involved, and a near miss configuration. For example, the relative velocity of impact may be directly proportional to the severity, wherein the higher the relative velocity of impact, the more severe the near miss will be in case of a collision. Regarding the near miss angle, for a full-frontal near miss, the chances of it turning into a severe collision is high. For a head on Near miss occurring at an angle, the severity may depend on the direction of the impact - if it is from the driver’s side, then the probability of collision is increased. For a side impact collision, the severity may depend on whether the line of impact was severe enough to turn into collision. The amount of severity may depend on the angle from which it was impacted. The offset may be proportional to the difference between 1 and modulus (absolute value) of offset. For example, if it is a zero offset near miss then its most critical with severity decreasing linearly with increasing offset of near miss. The offset between traffic participants, e.g. vehicles, determines the line of feree and momentum transfer - for example, in a head-on near miss, the incident is most severe when the offset is zero, i.e., it can turn into a full-frontal impact. For zero offset near miss, the kinetic energies of both the vehicles do not create rotation due to aligned momentums and hence most of the energy transfer from vehicles would go towards each other’s vehicle bodies. The more the offset in the collision - the farther are the centers of mass of both vehicles and more linear kinetic energy is converted into rotational kinetic energy and less dissipated into deformations of vehicle bodies. The near miss configuration may differ between head on impact, side impact and rear end impact. It may be determined from e.g. existing datasets that signify how severe what kind of near miss configuration was and assign a correlative empirical score to each collision configuration in terms of its severity. The direct empirical correlations data on which near miss configuration results in more severity of a near miss. The values of the individual factors may be numerical values between 0 and 1 . For example, the angle of near miss factor may be determined, e.g. calculated, as: Contact_angle_factor = function (line of impact of moment of near miss and relative positioning of both vehicles and their occupant compartments). The offset of near miss factor may be determined, e.g. calculated, as: offsetFactor = max(l - \of f setOfNearCollision\, 0.1), meaning linearly decrease from a peak at offset of zero (axes of both vehicles aligning and intersecting during Near miss) implying maximum conversion of kinetic energies leading to a critical Near miss or even collision. Thereby, the more the offset of collision the more the kinetic energies would get converted to angular kinetic energies and less would be dissipated and absorbed in vehicle bodies. The near miss configuration factor may be determined, e.g. calculated, considering that headOnCollisionF actor > sidelmpactF actor > rearEnd Factor), wherein the head on near miss risk factor is higher than side impact and rear-end impact. The duration factor may dur at ion be determined, e.g. calculated, as: durationFactor = min ( - , 1), wherein duration is a proportionate metric which tells the amount of time the vehicle was in a near miss configuration. Thereby, the higher the duration the vehicles are in close proximity and in near miss configuration, the higher the chances of its severity and thus higher the chances to turn it into a collision. The importance or weightage of the factor for near miss severity may be high, medium, or low. For example, high importance may be expressed by a value of 0.9, medium importance may be expressed by a value of 0.6, and low importance may be expressed by a value of 0.3, wherein other values are conceivable for each factor. The importance of relative velocity of impact may be high, the importance of near miss angle may be high, the importance of the offset may be high, and the importance of the near miss configuration may be high.

The severity of the near miss event may be determined, e.g. calculated, as normalized weighted averages of different near miss factors described above, and may be expressed as: Severity_score = duration_f actor * weight_duration + type_f actor * weight_type + of fset_f actor * weight_offset + angle_f actor * weight_offset )/sum of all weights . The following factors may be computed on the basis for severity score: time interval, impact velocity, Near miss angle, vehicle offset, type of configuration of Near miss with respect to each vehicle (rear, head on, side impact).

Regarding the second branch, at block 218, pre-crash parameters at the worst point are determined. The pre-crash parameters at the worst point may be determined based on a respective mass and respective velocity of the at least one traffic participant 12, 14. It may further comprise determining a collision configuration being indicative for a location of impact comprising at least one of a front impact, a rear impact, a driver-side impact, and a passenger-side impact. Further, it may comprise, at block 220, determining a speed difference between a speed of the vehicle 12 before and after the collision based on the pre-crash parameters and the collision configuration. Alternatively, at block 222, it may comprise determining a closing speed corresponding to a relative velocity of the at least one traffic participant moving. It may further comprise estimating the severity in case of collision based on the estimated speed difference or the closing speed and the collision configuration. For example, the pre-crash parameters may be determined at minETTC as an example of the worst point. In other words, the pre-crash parameters may be taken at the moment where the near miss becomes closest to a collision. The collision configuration may be determined from a projected collision angle, offset, length and width of the above-mentioned bounding boxes. The speed difference, which may be referred to as DeltaV, is the speed difference between a speed of the vehicle before and after the collision. In general, a higher DeltaV results in more severe injuries to the passengers of the vehicles. DeltaV may be estimated based on the pre-crash parameters of the the at least one traffic participant 12, 14 and the collision configuration. For example, DeltaV may be determined by determining the closing speed, i.e., the relative velocity of one moving object with respect to the other at the worst point, e.g. minETTC. Then, a bumper’s restitution coefficient is calculated based on a suitable metric, such as, for example, Antonetti, V., “Estimating the Coefficient of Restitution of Vehicle-to- Vehicle Bumper Impacts,” SAE Technical Paper 980552, 1998, https://doi.org/10.4271/980552. Further, the system’s mass may be determined. In addition, the projected contact plane between the two objects may be determined. The closing speed may be decomposed into the normal and tangential directions to the contact plane. Further, the impulse and velocity ratios (tangential over normal) may be computer. Further, the deformation energy of the vehicle subject to the speeds in the normal and tangential directions of the contact plane, system mass, the restitution coefficient, and the impulse ratio may be determined. DeltaV may be obtained by an empirical formula which depends on the subject object’s mass, the system’s mass, the restitution coefficient, and the deformation energy. At block 224, an injury level and/or a property damage level is determined based on maximum abbreviated injury scale, MAIS, based on the speed difference or the closing speed. For example, based on empirical formulas the injury level of the passengers inside the vehicle 12 and another vehicle, e.g. the at least one traffic participant 14, or any other vulnerable road user may be predicted. The main predictor for injuries caused by vehicle-to-vehicle collision is the DeltaV followed by the location of the impact, i.e. the collision configuration, e.g. rear-end, side, or front. The probability for a certain MAIS injury level to happen based on these parameters as well as the associated property damage levels may be determined. In the case of vulnerable road users, the principal predictor for injury level is the closing speed, so that the MAIS injury as well as property damage level probabilities may be determined based on the closing speed.

At block 226, injury cost or injury cost distribution may be determined based on the severity determined in the first branch, i.e. blocks 212, 214, and 216, or in the second branch, i.e. blocks 218, 220, 222, and 224. At block 228, property damage cost or property damage cost distribution may be determined based on the severity determined in the first branch, i.e. blocks 212, 214, and 216, or in the second branch, i.e. blocks 218, 220, 222, and 224. The estimated severity in case of collision comprises at least one of a bodily injury severity and a property damage severity.

At block 230, the safety criticality of the at least one near miss event based on the worst point is determined. It may be determined per near miss event. This may be done for multiple near miss events. At block 232, the safety criticalities of the multiple near miss events may be aggregated. At block 234, an overall safety criticality and/or cost severity for the specific location may be determined based on the aggregated safety criticalities. Optionally, a cost severity distribution may be determined.

At optional block 236, the type of maneuver may be classified. At block 238, with a more detailed context description of the near miss, its collision conversion rate may be refined, adding to the weighting of the near miss event.

Fig. 3 illustrates in a schematic block diagram an example 300 of determining a near miss event.

As mentioned above, the trajectory data 112 may comprise kinematic data 302 and metadata 304 of the at least one traffic participant 12, 14. The kinematic data 302 may comprise at least one of a position, an orientation, a velocity, an acceleration, and yaw of the at least one traffic participant 12, 14. The metadata 304 may comprise at least one of a road user type, a vehicle type, a mass, a length, a shape, a dimension of the at least one traffic participant 12, 14, weather data, and traffic information.

At block 306, a vehicle safety distance metric involving the at least one traffic participant 12, 14 may be determined based on the trajectory data 112 and/or the kinematic data 302 and the metadata 304. Reference sign 320 denotes considering parameters comprising at least one of a detection time, judgement time, response time, and breaking time configurable with respect to at least one of a vehicle type, location, weather, traffic density, or the like. At block 308, the detection time may be used to determine, e.g. calculate, a distance traveled at same speed and/or an acceleration, denoted by reference sign 310. At block 312, the judgement and/or response time may be used to determine, e.g. calculate, a distance travelled at same speed, denoted by reference sign 314. At block 316, the braking time may be used to determine, e.g. calculate a braking distance at best breaking maneuver. Block 306 may output the directional vehicle safety distance and/or a corresponding envelope, for the at least one traffic participant 12, 14.

At block 322, the near miss event is determined, e.g. calculated. It provides information about at least one of a vehicle location, derived contact point, vehicle ID, and vehicle movement direction.

At block 324, the near miss configuration is determined, e.g. calculated. It provides information about at least one of a duration of the near miss event, yaw, vehicle velocity, collision configuration, e.g. front, side, rear, a shortest distance, an offset distance, and a (near-)collision angle.

At block 326, characteristics and/or configurations of the near miss event are output.

Fig. 4 schematically illustrates an example 400 of determining the closeness weighting factor c. Fig. 4 shows some exemplary bounding boxes, wherein each bounding box is indicative for the at least one traffic participant 12, 14. In Fig. 4, AE is the actual shortest distance between the two bounding boxes, expanded or assigned with a safety envelope, and AB + CD is the maximum possible shortest distance between the two vehicles and/or traffic participants for them to still be in a near miss in the given scenario. As not all near misses are equal, the near miss closeness weighting c may be determined. This provides a static function between 0 (at the threshold of the near miss detection, when bounding boxes only slightly touch) to 1 , equal to a collision, considering the worst point, e.g. minETTC. For example, the near miss closeness weighting c may be determined by:

Fig. 5 illustrates in a block diagram an example 500 for determining costs based on the above-mentioned pre-crash parameters.

At block 502, the pre-crash parameters at the worst point, e.g. at minETTC, are determined and/or received. The pre-crash parameters may comprise at least one of masses of the traffic participants involved, their velocities, a projected collision angle, collision offsets, bounding polygons and/or boxes, and the closing speed.

At block 504, a projected contact plane between two objects, i.e. traffic participants may be determined. It may provide velocity in tangential and normal direction, indicated by an arrow directed to block 502.

At block 506, a coefficient of restitution may be determined using an empirical computer model.

At block 508, a deformation energy may be determined using a physics-based computer model.

At block 510, the above-mentioned DeltaV may be determined using a physics-based computer model.

At block 512, a property damage level may be determined based on at least one regression function based on historical and/or test data.

At block 514, an injury severity level may be determined based on at least one regression function based on historical and/or test data.

At block 516, the costs may be determined using e.g. at least one lookup table.

Fig. 6 illustrates in an example 600 that multiple other options may be used to determine, e.g. calculate, injury and damage levels based on the above-mentioned pre-crash parameters and turn them into costs. These options may be combined as indicated by the diagram and/or arrows. In Fig. 6, block 602 denotes the pre-crash parameters. Block 604 denotes empirical and/or physics-based formulae. Block 606 denotes a physics-based computer model. Block 608 denotes machine learning methods, e.g. based on historical and/or simulated crash databases. Block 610 denotes post-crash parameters that may be determined, comprising at least one of DeltaV, deformation energy, momentum transfer, and coefficient of restitution. Block 612 denotes regression functions based on historical and/or test data. Block 614 denotes machine learning models. Block 616 denotes crash test reports. Block 618 denotes the property damage level. Block 620 denotes the injury severity level. Block 622 denotes a cost calculation module.

Fig. 7 illustrates in an example 700 several methods to translate severity scores or MAIS level distributions into costs that would arise assuming the near miss event would be a collision.

Block 602 denotes the property damage level and block 604 denotes the injury severity level. Block 606 denotes a cost calculation module, wherein block 608 denotes a lockup table based on historical and/or test data, block 610 denotes a machine learning model, and block 612 denotes insurance company data. Block 614 denotes the cost of a collision. Block 616 denotes the near miss closeness weighting, e.g. the weighting factor c. Block 616 denotes, as an output, a potential cost-based severity of the near mis event.

For further highlighting the risk assessment, Fig. 8 illustrates in a flowchart a method 800 for assessing risk of deploying an at least one traffic participant at a specific location. The method comprises receiving 810 trajectory data indicating trajectory information of at least one traffic participant of a traffic scenario at a specific location. Further, the method comprises analyzing 820 the trajectory data for occurrence of near miss events involving the at least one traffic participant. The method further comprises determining 830 a safety criticality of the traffic scenario based on the analysis of the trajectory data. The safety criticality considers a closeness of a near miss event to becoming a collision and an estimated severity in case of such collision. In addition, the method comprises generating 840 output data indicating the safety criticality for the specific location.

Claims

Patent claims

1 . A method (800) for assessing risk of deploying an at least one traffic participant (12, 14) at a specific location, the method comprising: receiving (810) trajectory data indicating trajectory information of at least one traffic participant of a traffic scenario at a specific location; analyzing (820) the trajectory data for occurrence of near miss events involving the at least one traffic participant; determining (830) a safety criticality of the traffic scenario based on the analysis of the trajectory data, the safety criticality considering a closeness of a near miss event to becoming a collision and an estimated severity in case of such collision; and generating (840) output data indicating the safety criticality for the specific location.

2. The method of claim 1 , wherein the output data is provided and/or used for deploying the at least one traffic participant.

3. The method of claim 1 or 2, wherein the specific location is at least one of a geographical location, fictional location, a geographical area, a road section, a road area, a traffic area, a road route, and a travel route.

4. The method of any one of the preceding claims, wherein multiple analyses for near miss events are performed and corresponding safety criticalities are determined, and the method further comprises: aggregating the safety criticalities of the multiple analyses for near miss events; and determining an overall safety criticality for the specific location based on the aggregated safety criticalities.

5. The method of any one of the preceding claims, further comprising: determining a potential cost severity for the based on the corresponding safety criticality.

6. The method of claim 5, wherein multiple near miss events and corresponding safety criticalities are determined, the potential cost severities of the multiple near miss events are aggregated and a cost severity distribution for the specific location is determined.

7. The method of any one of the preceding claims, wherein the trajectory data comprises kinematic data and metadata of the vehicle and the at least one traffic participant.

8. The method of claim 7, wherein the kinematic data comprises at least one of a position, an orientation, a velocity, an acceleration, and yaw of the at least one traffic participant.

9. The method of claim 7 or 8, wherein the metadata comprises information about the traffic scenario.

10. The method of any one of claims 7 to 9, wherein the metadata comprises at least one of a road user type, a vehicle type, a mass, a length, a shape, a dimension of the at least one traffic participant, weather data, and traffic information.

11. The method of any one of the preceding claims, wherein the determining of the safety criticality weights whether the at least one traffic participant is a vulnerable road user exposed to an increased risk.

12. The method of any one of the preceding claims, wherein a near miss event is detected by determining a vehicle safety distance metric involving the at least one traffic participant based on the trajectory data.

13. The method of any one of the preceding claims, wherein a near miss event is detected by determining a respective bounding box of the at least one traffic participant based on kinematic data and metadata of the at least one traffic participant, and determining the near miss event based on a distance and/or overlap with the bounding box.

14. The method of claim 12, wherein the bounding box of the at least one traffic participant and the corresponding distance and/or overlap is determined for a number of timesteps to detect the near miss event.

15. The method of any one of the preceding claims, wherein, for near miss events occurred, a worst point where the near miss is closest to become a collision is determined.

16. The method of claim 15, wherein the worst point is determined from a minimum enhanced time-to- collision, minETTC, indicating the time where a trajectory prediction of the vehicle is closest in time to collide.

17. The method of any one of the preceding claims, wherein the closeness to becoming a collision of the at least one near miss event is determined by calculating a closeness weighting factor based on a distance between bounding boxes assigned to the at least one traffic participant and/or a corresponding minimum enhanced time-to-collision, minETTC.

18. The method of any one of the preceding claims, wherein the estimated severity in case of collision comprises at least one of a bodily injury severity and a property damage severity.

19. The method of claim 18, wherein the bodily injury severity is estimated based on maximum abbreviated injury scale, MAIS.

20. The method of any one of the preceding claims, wherein estimating the severity in case of collision comprises: determining pre-crash parameters, particularly at a worst point, based on a respective mass and respective velocity of the at least one traffic participant; determining a collision configuration being indicative for a location of impact comprising at least one of a front impact, a rear impact, a driver-side impact, and a passenger-side impact; determining a speed difference (DeltaV) between a speed of the at least one traffic participant before and after the collision based on the pre-crash parameters and the collision configuration; or determining a closing speed corresponding to a relative velocity of one of the at least one traffic participant; and estimating the severity in case of collision based on the estimated speed difference or the closing speed and the collision configuration.

21 . The method of claim 20, further comprising: determining an injury level and/or a property damage level based on maximum abbreviated injury scale, MAIS, based on the speed difference (DeltaV) or the closing speed.

22. The method of any one of claims 1 to 18, wherein estimating the severity in case of collision comprises: determining a respective bounding box of the at least one traffic participant based on kinematic data and metadata of the at least one traffic participant; and determining at least one severity metric comprising at least one of an impact velocity indicating relative velocity between the centers of mass of the traffic participants involved, a near miss angle indicating an angle between longitudinal axes of the traffic participants involved at the time of near miss, an offset distance between longitudinal axes of the traffic participants involved at the time of near miss, an offset distance between two closest central axes of the traffic participants involved, an acceleration of the traffic participants involved, and an angular velocity of the traffic participants involved; and determining the severity in case of collision based on the at least one severity metric.

23. An apparatus (100) for assessing risk of deploying an at least one traffic participant (12, 14) at a specific location, the apparatus comprising: interface circuitry (110) configured to: receive trajectory data (112) indicating trajectory information of at least one traffic participant (12, 14) of a traffic scenario at a specific location; and processing circuitry (120) configured to: analyze the trajectory data (112) for occurrence of near miss events involving the at least one traffic participant (12, 14); determine a safety criticality of the near miss events based on the analysis, the safety criticality considering a closeness to becoming a collision and an estimated severity in case of such collision; and generate output data (122) indicating the safety criticality for the specific location.

24. A non-transitory machine-readable medium having stored thereon a program having a program code for performing the method according to any one of claims 1 to 22, when the program is executed on a processor or a programmable hardware.

25. A computer program having a program code for performing the method according to any one of claims 1 to 22, when the program is executed on a processor or a programmable hardware.