Method and device for estimating the density of a collision object in a collision with a motor vehicle

Abstract

Method for estimating the density (D) of a collision object (11; 12; 13) for a protection system (2) of a motor vehicle (1), comprising the following steps: a) receiving environmental data from at least one environmental sensor (3, 3a) for detecting a collision object (11; 12; 13), b) Determining the extent of the collision object in at least two dimensions, namely width (W) and height (H) from the environmental data, d) detecting a collision based on measured values ​​from at least one collision sensor (4) present in and / or on the motor vehicle (1), e) estimating a mass (M) of the collision object (1; 12; 13) on the basis of measured values ​​of the collision sensor (4), characterized in that the method comprises the following steps: c) estimating a volume (V) of the collision object (11; 12; 13) based on the measured extent, f) calculating an estimated average density (D) of the collision object (11; 12; 13) from estimated volume (V) and estimated mass (M), g) forwarding the estimated density (D) in the protection system (2) for further assessment of the collision in relation to other parameters.

Description

State of the art

[0001] The present invention relates to a method and a device for estimating the density of a collision object in a collision with a motor vehicle. Detecting the type and / or severity of a collision between a vehicle, in particular a motor vehicle, and a collision object in order to trigger appropriate safety measures is of great importance for the safety of the vehicle's occupants and external road users.

[0002] Modern motor vehicles are equipped with extensive sensors and monitoring devices aimed at increasing safety for vehicle occupants and external road users. In the course of the development of autonomous vehicles, which participate in road traffic without driver intervention, increasingly improved systems for scanning the vehicle's surroundings have been and are being developed. The present invention assumes that a vehicle is equipped with extensive sensor technology and means for generating scanning data of its surroundings. In particular, these can be video cameras, radar systems, lidar systems, and / or ultrasound systems.

[0003] Such systems are also known to be used to anticipate collisions that can no longer be avoided as early as possible, to assess their course and severity, and to trigger vehicle safety systems, such as seat belt tensioners, seat adjustments, and / or airbags, in a timely manner when an accident occurs. With such systems, it is also common practice to classify or categorize accidents based on their severity and / or sequence, so that different sequences of safety measures can then be initiated. The same applies to the protection of external road users, especially pedestrians, who can be protected, for example, by pedestrian airbags or by raising the hood of a vehicle.

[0004] DE 103 36 638 A1 discloses a device that uses environmental sensors to determine and classify the shape and dimensions of an object in a vehicle's surroundings. The object's speed can also be taken into account in the classification. Depending on this classification, protective measures can be activated. Another combination of predictive environmental sensors and contact sensors is known from DE 10 2014 226 966 A1.

[0005] From the document DE 10 2005 006 763 A1 a method is known in which a relative speed of a first object on a collision course with a second moving object as well as the force effect of the first object on the second object are recorded in order to characterize the first object.

[0006] DE 10 2006 060 844 A1 discloses a vehicle safety device for initiating safety measures. This device receives crash-relevant data from a transmitter on a tire of a second vehicle. This crash-relevant data may include, among other things, the mass, height, and width of the second vehicle. Disclosure of the invention

[0007] Here, a method and a device are presented that make it possible to further improve the categorization of possible accidents or the classification of potential collision partners according to the type and / or severity of a collision.

[0008] Against this background, a method according to claim 1 for estimating the density of a collision object is proposed in order to be able to take this into account when detecting the type and / or severity of a collision between a vehicle and a collision object in an early phase of the collision, and to make the triggering of appropriate safety measures more reliable and safer with this additional parameter. A device for estimating the density according to claim 10 is also presented here.

[0009] A parameter that has so far been difficult to consider when categorizing and classifying accidents and (potential) collision partners between a vehicle and a collision object is the density of the collision object. Knowing the density is important for predicting the accident sequence as accurately as possible, but also for distinguishing between road users requiring protection (external persons) and objects or animals that could not be identified before the collision or were mistakenly identified, for example, as pedestrians.

[0010] Safety systems in motor vehicles (e.g. stability control ESP or brake control) and comfort systems (e.g. lane keeping assistant) are increasingly being networked with passive safety systems.

[0011] In the "collision scenario" functional variant (also called "ID-x"), such a safety system (e.g., airbag deployment algorithm) is configured based on environmental sensors. Environmental sensors (mono / stereo camera, radar, lidar, ultrasound) monitor the environment and determine a possible impending collision and its type. An airbag control unit can be configured to be more sensitive in the event of an impending collision, allowing the restraint systems to be deployed more quickly.

[0012] For example, a radar sensor can be used to predict a frontal collision (also known as "ID-F") with a vehicle. During the predicted time at which the accident is expected to occur, the activation threshold for restraint systems is reduced (i.e., a more sensitive and earlier reaction). If a possible accident is then registered by the classic passive safety sensors (e.g., acceleration sensors, pressure sensors, particularly in the front of the vehicle), a faster response can be made, as the plausibility check time (the time the system needs to check whether there is a high probability of a collision) can be limited. Depending on the expansion level, a frontal collision ("ID-F") or side collision ("ID-S") can be predicted, or a response can be made to a rear collision.In collisions, a distinction can be made between different opponents, for example cars, trucks, pedestrians (“ID-P”) or a firmly anchored object.

[0013] The situation designated "ID-x" always requires that an accident has already occurred. Predictive systems simply shorten the reaction time, allowing vehicle occupants to prepare better for the accident (creating more space to dissipate kinetic energy and thus avoid acceleration peaks). The basic functionality of collision sensing with acceleration sensors, etc. remains. The goal with "ID-F" ("ID-S") is generally faster deployment of the restraint systems inside the vehicle. "ID-P", on the other hand, primarily aims to make the deployment of pedestrian protection systems more robust (safe but with a low rate of false deployments), since the acceleration signal in a collision with a pedestrian is very small and difficult to distinguish from other situations. In simulations, a so-called "leg impactor" (a special dummy leg) is used, which is only approx.6 kg and must trigger the alarm. A small animal, however, should not necessarily trigger the alarm (bird strike, impact from small animals such as rabbits, etc.). False triggering is undesirable, as pedestrian protection systems are often designed to be irreversible, and a faulty triggering of the system incurs costs.

[0014] Surrounding sensors have only a limited detection range. For example, a long-range radar has a detection range of less than + / - 10°, while a camera has a detection range of + / - 25°, for example. The angles specified here each define a conical area in front of the vehicle within which typical surrounding sensors can detect objects. Detecting objects takes a certain amount of time. To make the detection robust against noise, the system often only accepts objects that have been visible for a certain period of time.

[0015] Acceleration sensors designed to measure pedestrian impacts are connected to the vehicle body. For example, driving through a pothole can generate the same signal as a pedestrian being hit, because acceleration sensors provide only a small amount of information (measurement channels) compared to environmental sensors (e.g., a video sensor has 1,000,000 measurement points or "pixels").

[0016] "ID-P" is designed to increase robustness compared to conventional pedestrian protection systems. For example, the threshold for triggering pedestrian protection is raised (i.e., made less sensitive) if the surrounding sensor does not detect a pedestrian.

[0017] The article by SN Huang, JK Yang and F. Eklund “Analysis of Car-Pedestrian Impact Scenarios for the Evaluation of a Pedestrian Senso System Based on the Accident Data from Sweden” describes basic situations for potential collisions between pedestrians and vehicles as well as sensor systems for vehicles to detect such situations.

[0018] It should be noted that the processes and systems described here for pedestrians can also be applied, with certain limitations, to other road users, such as playing children, cyclists, wheelchair users, etc. Therefore, the following will occasionally refer to collision objects or road users requiring protection.

[0019] When pedestrian protection systems are triggered in the event of a collision between a pedestrian and a motor vehicle, the consequences for the pedestrian can often be mitigated. Examples of possible protection systems include pedestrian airbags or the raising of the hood. With most systems, the triggering of these systems results in the vehicle being unable to continue driving, and irreversible systems must first be replaced, which incurs costs.

[0020] It is therefore desirable that pedestrian protection systems are rarely triggered falsely when there is actually no collision with a pedestrian. Depending on the complexity of a pedestrian protection system, false triggering can have many causes, such as uneven road surfaces, collisions with small animals, falling rocks, or objects on the road.

[0021] Advanced assistance systems for motor vehicles are therefore designed to identify pedestrians in the vicinity of the vehicle using suitable sensors, if possible, even before a collision. They can, for example, anticipate potential collisions based on the relative speed between the pedestrian and the vehicle and, where possible, prevent them or mitigate the consequences for the pedestrian by triggering pedestrian protection systems in a timely manner when an impact is detected. In such cases, false activation is unlikely.

[0022] The method described here for estimating the density of a collision object uses environmental data obtained with at least one environmental sensor (step a)) to detect collision objects. The volume of the collision object is then estimated.

[0023] The extent or the volume corresponding to the three-dimensional extent of the collision object is recorded or estimated in steps b) and c). Often, the estimation is only possible in two dimensions if only two dimensions of the collision object are recognizable from the environmental data. These are usually the width (W) and the height (H). If necessary (for example, if three dimensions of the collision object are recognizable based on environmental data from several remotely arranged environmental sensors), a three-dimensional extent can also be estimated (including the depth of the collision object as a third dimension).

[0024] Typically, an environment sensor provides at least the dimensions of the collision object perpendicular to the sensor's beam direction, but often also information going far beyond that, since additional information, such as the depth or type of object, can be obtained through image recognition or successive observations of the object that necessarily approaches before a collision. Environment sensors can also measure the distance and often the relative speed between the vehicle and the collision object, which can be advantageous for subsequent calculations. In any case, it is possible to estimate the volume of the collision object based on the environment data. Using two or more environment sensors, it is even possible to create a stereo image of the surroundings.

[0025] If a collision occurs, it is registered by a collision sensor, which may also consist of several individual sensors. (Step d)) The strength of the signal from the collision sensor allows an estimate of the mass of the colliding object (Step e)). The average density of the colliding object can be determined by simple division of the volume and mass. (Step f)) This, along with other parameters, is an important additional parameter for classifying the type and / or severity of the collision.

[0026] Since an environment sensor can initially only measure the angle at which a collision object appears within its detection range, the described method preferably also includes the measurement of the distance (particularly by measuring the propagation time of the scanning signals) between the motor vehicle and the collision object to determine the absolute dimensions of the collision object. Accuracy can be increased by taking multiple consecutive measurements as the collision object approaches.

[0027] Preferred methods for scanning the environment include video surveillance, lidar surveillance, radar surveillance, ultrasonic surveillance, or combinations thereof. All of these systems can also be used as stereo systems with two sensors.

[0028] When using radar surveillance, which is often already present in motor vehicles as distance radar, reflection signals, especially reflections from a road surface that pass underneath a vehicle, especially a truck, can also be used to determine the depth of a collision object. A combination of video and radar signals, in particular, often allows even a distant object to be identified and its approximate volume to be determined.

[0029] Preferably, a comparison with stored scan data is performed to identify and / or classify collision objects in order to determine the height, width, and depth of a detected collision object. Accuracy is increased by determining the distance, allowing a large truck to be distinguished from a small one, or an adult pedestrian to be distinguished from a child.

[0030] When a collision actually occurs, collision sensors in the front of the vehicle are triggered. Preferably, one or more acceleration sensors or pressure sensors are used. If multiple sensors are used, information about the location of the collision on the vehicle is also available. For example, a hose with pressure sensors at its ends is installed behind a bumper in the front area. In the event of an impact, the bumper deforms, creating a pressure pulse in the hose, which can be detected by the pressure sensors at both ends. Different travel times for the pressure pulse can also be used to determine the location of the impact.

[0031] Measurements from such collision sensors can be used to estimate the mass of a collision object. These measurements contain, for example, information regarding pressure pulses and / or accelerations that occur on the vehicle as a result of the collision. These measurements can be used for mass estimation.

[0032] The estimated density is preferably used to verify a hypothesis about the type of collision object developed from other information, and to trigger or not trigger protective measures depending on this. For example, if other information points to a pedestrian, but the measured density is far below or above that typical for a human body, the "pedestrian" hypothesis can be rejected and pedestrian protection measures may not be triggered. In this way, for example, cardboard cutouts or garbage cans can be distinguished from actual pedestrians. It can also make it easier to distinguish plastic construction site barriers from children, even though they may appear similar.

[0033] The accuracy of the mass estimate described above can be further increased by taking into account the speed of the vehicle, or even better, the relative speed between the vehicle and the collision object. Since the strength of the signals from a collision sensor is proportional to the mass and impact velocity of a collision object, the mass can be estimated more accurately when the impact velocity is known.

[0034] As a result, with the described method, the protection system preferentially triggers a protective measure when, taking into account the estimated density of a collision object, the system detects a situation in which the protective measure appears necessary. Taking density into account increases safety for all occupants or external road users, but reduces the number of possible false triggers. The density can be estimated at an early stage of the collision, for example, within a few milliseconds after the initial contact between the collision object and the vehicle, so that the further course of events can be influenced in a timely manner by appropriate protective measures.

[0035] It should be noted that the described method steps a) to g) of the method typically run at least partially in parallel during operation of a motor vehicle. The environmental data from an environmental sensor used in step a) is preferably obtained by the environmental sensor by scanning the environment. The environmental scanning preferably takes place continuously throughout the entire operation of the motor vehicle. If a potential collision object is detected in step a), the subsequent steps are initiated, which then preferably take place in parallel to step a).

[0036] If necessary, the execution of the method is aborted as soon as it is determined that a collision with a previously identified potential collision object is not occurring. The execution of the method can also be aborted if it is determined that the estimated density D determined in step g) is no longer required for further evaluation of the collision in conjunction with other parameters. Therefore, it is possible and possibly advantageous for the method to include an abort option, which occasionally results in it being executed only partially.

[0037] A control device which is particularly designed to carry out the described method is also to be described here.

[0038] The control unit for a protection system of a motor vehicle for estimating the density of a collision object comprises the following features: - a first signal input for environmental data from at least one environmental sensor for detecting a collision object, - a first estimation module for estimating a volume of the collision object based on environmental data from the environmental sensor, - a second signal input for measured values ​​of a collision detection with at least one collision sensor arranged in and / or on the motor vehicle, - a second estimation module for estimating a mass of the collision object based on measured values ​​of the collision sensor, - a calculation unit for calculating an estimated average density of the collision object from estimated volume and estimated mass, - a forwarding in the protection system for the estimated density for further evaluation of the collision in relation to other parameters, - a signal output for outputting a signal to trigger at least one protective measure when the protection system detects a situation in which the protective measure appears necessary.

[0039] The device can, for example, be a control unit configured to carry out the described method. The necessary sensors for determining the data processed by the method (in particular, data from an environmental scan acquired with an environmental sensor) can be connected to this device. For this purpose, corresponding signal inputs can be provided on the device.

[0040] Also to be described here is a computer program which is set up to carry out the described method steps and a machine-readable storage medium on which this computer program is stored.

[0041] Further details of the invention and an embodiment, to which the invention is not limited, are explained in more detail with reference to the drawings. They show: Fig. 1 a schematic representation of a motor vehicle in a constellation with three different collision objects before a collision and Fig. 2 shows the schematic sequence of the method according to the invention.

[0042] Fig. 1 shows a schematic representation of a motor vehicle 1 moving at a speed L and approaching various collision objects 11, 12, 13. The collision objects shown as examples are a truck 11, a pedestrian 12, and a construction site barge 13. The motor vehicle 1 is equipped with an environment sensor 3, which provides data about the environment in front of it. The environment sensor 3 can also be supplemented by a second environment sensor 3a in order to generate a stereo image of the environment. The motor vehicle 1 is also equipped with a collision sensor 4, which can also consist of several individual sensors. The environment sensor 3 is connected to a first signal input 6, and the collision sensor 4 is connected to a second signal input 8 of a protection system 2, which in turn is connected via a signal output 15 to at least one protection component 5 (e.g., pedestrian airbag, passenger airbag).The safety system has a first estimation module 7 for estimating the volume V of a collision object 11, 12, 13 from data from the environment sensor 3 and a second estimation module 9 for estimating the mass M of a collision object from the data from the collision sensor 4. In a calculation unit 10, the estimated average density of the collision object 11, 12, 13 is determined from the estimated mass M by dividing it by the estimated volume V. This is forwarded within the protection system and, together with further information within the protection system, is used to classify the collision object 11, 12, 13 and / or to classify the type and / or severity of the collision. If necessary, the protection system 2 outputs a signal via a signal output 15 to trigger one or more protection components 5 at suitable times.

[0043] The environment sensor 3 scans the surroundings of the motor vehicle and can thus detect collision objects 11, 12, 13 with which the motor vehicle 1 will collide on its further path, even before the collision and collect information about them. Using the example of a slowly moving (or stationary) truck 11 as the collision object, it can be seen that at least the two dimensions, namely width B and height H, can be detected immediately. However, to determine the absolute values, the distance A between the motor vehicle 1 and the collision object 11 is still required, which is generally also measured by the environment sensor 3 or, for example, a supplementary radar system. In many cases, simply identifying the collision object 11 as a truck can provide an approximate estimate of its volume V.Often, however, in the preliminary phase of a collision, the third dimension as the depth T of the collision object is partially visible and can be measured, possibly even by radar beams reflected from a road surface beneath the truck. The situation is similar with a pedestrian 12 as a collision object. Height H and width B provide an initial indication as to whether it is a pedestrian. Depending on the quality of the available scan data, however, it can be difficult to distinguish, for example, a child from a construction site barge 13, as shown as the third collision object. In the event of a later impact of one of the collision objects 11, 12, 13, initially only the estimated volume V as the product of width B, height H and depth T is available. In addition, a previously measured relative speed RL between the motor vehicle and the collision object 11, 12, 13 is typically available.From the relative speed RL (or, to a first approximation, only from the speed L of the motor vehicle 1) and the strength of the signal from the collision sensor 4, the mass M can be estimated in the second estimation module 9, from which the average density D of the collision object 11, 12, 13 is then determined using the estimated volume V. This is processed as additional information in the protection system 2.

[0044] Fig.2 shows a schematic diagram of the method sequence in the protection system 2. A collision object 12, in this case a pedestrian, for example, is detected by an environment sensor 3, wherein its dimensions width B, height H and depth T and its relative speed RL to the motor vehicle 1 are scanned. This data is passed via the first signal input 6 to the first estimation module 7, where the width B, height H and depth T as well as the distance A and relative speed RL between the motor vehicle 1 and the collision object 12 are determined from the data. The volume can then be estimated from this, and the more accurately the better the type of collision object can be identified or classified. In an initial phase of the collision, data from the contact sensor 4 is passed via the second signal input 8 to the second estimation module 8, which uses this data to estimate the mass M of the collision object 12, for which purpose the relative speed RL from the environment sensor 3 orfrom the first estimation module 7 is taken into account. The estimated volume V and the estimated mass M are forwarded to the calculation unit 10, which calculates the estimated average density D from them. The density D is then passed on to an object classification 16, which has at least one further parameter input 17 for further parameters. Here, the density D can either be included in the classification or an object hypothesis can be formulated from the other parameters, which is then subjected to a plausibility check with the estimated density. As a result, the accuracy of the object classification is improved and the number of false triggers is increased without any loss of safety. A reaction decision maker 18 can then use the result of the classification to decide which protective measures should be triggered and when, and provide corresponding signals for at least one protective component 5 at the signal output 15.

Claims

[1] Method for estimating the density (D) of a collision object (11; 12; 13) for a protection system (2) of a motor vehicle (1), comprising the following steps: a) receiving environmental data from at least one environmental sensor (3, 3a) for detecting a collision object (11; 12; 13), b) Determining the extent of the collision object in at least two dimensions, namely width (W) and height (H) from the environmental data, d) detecting a collision based on measured values ​​from at least one collision sensor (4) present in and / or on the motor vehicle (1), e) estimating a mass (M) of the collision object (1; 12; 13) on the basis of measured values ​​of the collision sensor (4), characterized by that the procedure includes the following steps: c) estimating a volume (V) of the collision object (11; 12; 13) based on the measured extent, f) calculating an estimated average density (D) of the collision object (11; 12; 13) from estimated volume (V) and estimated mass (M), g) forwarding the estimated density (D) in the protection system (2) for further assessment of the collision in relation to other parameters. [2] Method according to claim 1, wherein the detection of the extent carried out in step b) is supplemented by one or more successive measurements of the distance (A) between the motor vehicle (1) and the collision object (11; 12; 13). [3] Method according to claim 1 or 2, wherein an environment scan for determining environment data is carried out by at least one of the following methods: video surveillance, lidar surveillance, radar surveillance, ultrasonic surveillance, in each case mono or in the case of at least two environment sensors (3, 3a) in stereo. [4] Method according to one of the preceding claims, wherein when using at least one environmental sensor (3, 3a) with radar, reflection signals are also used to determine a depth (T) of a collision object (11; 12; 13). [5] Method according to one of the preceding claims, wherein the determination of width (W), height (H) and / or depth (T) of the collision object (11; 12; 13) is supported by identification and / or classification of the collision object (11; 12; 13) on the basis of existing stored scanning data. [6] Method according to one of the preceding claims, wherein acceleration sensors and / or pressure sensors are used as collision sensors (4) in step d). [7] Method according to one of the preceding claims, wherein following step g) an object hypothesis about the collision object (11; 12; 13) is established as a function of the other parameters, which hypothesis is verified on the basis of the estimated density. [8] Method according to one of the preceding claims, wherein in step e) the speed (L) of the motor vehicle (1) or the measured relative speed of the motor vehicle (1) and the collision object (11; 12; 13) is used to estimate the mass (M) of the collision object (11; 12; 13). [9] Method according to one of the preceding claims, wherein step g) in the protection system (2) is followed by: h) triggering at least one protective measure when the protective system (2) detects a situation in which the protective measure appears necessary. [10] Control device for a protection system (2) of a motor vehicle (1) for estimating the density (D) of a collision object (11; 12; 13) comprising the following features: - a first signal input (6) for environmental data from at least one environmental sensor (3, 3a) for detecting a collision object (11; 12; 13), - a second signal input (8) for measured values ​​of a collision detection with at least one collision sensor (4) arranged in and / or on the motor vehicle, - a second estimation module (9) for estimating a mass (M) of the collision object (11; 12; 13) on the basis of measured values ​​of the collision sensor (4), characterized by that the control unit includes the following features: - a first estimation module (7) for estimating a volume (V) of the collision object (11; 12; 13) from the environmental data of the environmental sensor (3, 3a), - a calculation unit (10) for calculating an estimated average density (D) of the collision object (11; 12; 13) from estimated volume (V) and estimated mass (M), - a forwarding (14) in the protection system (2) for the estimated density (D) for further evaluation of the collision in relation to other parameters, - a signal output (15) for outputting a signal to trigger at least one protective measure when the protective system (2) detects a situation in which the protective measure appears necessary. [11] Computer program which is arranged to carry out all steps of the method according to one of claims 1 to 9. [12] A machine-readable storage medium on which the computer program according to claim 11 is stored.

Images